EngineX/xc-llm-ascend
3
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

[Lint]Style: Convert vllm-ascend/ to ruff format(Batch #10) (#6173)

Browse Source 
### What this PR does / why we need it?
**Scope of Changes**:
| File Path |
| :--- |
|`vllm_ascend/ops/layer_shard_linear.py`|
|`vllm_ascend/ops/linear.py`|
|`vllm_ascend/ops/linear_op.py`|
|`vllm_ascend/worker/worker.py`|
| ` vllm_ascend/patch/worker/patch_bert.py` |
| ` vllm_ascend/patch/worker/patch_deepseek.py` |
| ` vllm_ascend/patch/worker/patch_distributed.py` |
| ` vllm_ascend/patch/worker/patch_module.py` |
| ` vllm_ascend/patch/worker/patch_multimodal_merge.py` |
| ` vllm_ascend/patch/worker/patch_qwen3_next.py` |
| ` vllm_ascend/patch/worker/patch_qwen3_next_mtp.py` |
| ` vllm_ascend/patch/worker/patch_rejection_sampler.py` |
| ` vllm_ascend/patch/worker/patch_rope.py` |
| ` vllm_ascend/patch/worker/patch_triton.py` |
| ` vllm_ascend/patch/worker/patch_unquantized_gemm.py` |
| ` vllm_ascend/patch/worker/patch_v2_egale.py` |
|` vllm_ascend/worker/npu_input_batch.py`|
|` vllm_ascend/worker/v2/aclgraph_utils.py`|
|` vllm_ascend/worker/v2/attn_utils.py`|
|` vllm_ascend/worker/v2/model_runner.py`|
|` vllm_ascend/worker/v2/sample/gumbel.py`|
|` vllm_ascend/worker/v2/sample/penalties.py`|
|` vllm_ascend/worker/v2/sample/sampler.py`|
|` vllm_ascend/worker/v2/spec_decode/__init__.py`|
|` vllm_ascend/worker/v2/spec_decode/eagle.py`|
|` vllm_ascend/worker/v2/states.py`|
### Does this PR introduce _any_ user-facing change?

### How was this patch tested?

- vLLM version: v0.14.0
- vLLM main:
d68209402d

Signed-off-by: MrZ20 <2609716663@qq.com>
Signed-off-by: SILONG ZENG <2609716663@qq.com>
Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>
Co-authored-by: wangxiyuan <wangxiyuan1007@gmail.com>
This commit is contained in:
SILONG ZENG
2026-02-06 15:35:06 +08:00
committed by GitHub
parent 65b7f716e6
commit 19b5d44ea8
33 changed files with 938 additions and 1243 deletions
Download Patch File Download Diff File Expand all files Collapse all files

177
vllm_ascend/distributed/parallel_state.py
View File

@@ -1,35 +1,33 @@
from typing import Optional
import torch
from vllm.config import ParallelConfig, get_current_vllm_config
from vllm.distributed.parallel_state import (GroupCoordinator, get_tp_group,
get_world_group,
init_model_parallel_group)
from vllm.distributed.parallel_state import GroupCoordinator, get_tp_group, get_world_group, init_model_parallel_group
from vllm_ascend.ascend_config import get_ascend_config
from vllm_ascend.utils import enable_dsa_cp_with_layer_shard, flashcomm2_enable
# Currently, mc2 op need their own group coordinator.
_MC2: Optional[GroupCoordinator] = None
_MC2: GroupCoordinator | None = None
# Module specific tensor parallel groups
_MLP_TP: Optional[GroupCoordinator] = None
_OTP: Optional[GroupCoordinator] = None
_LMTP: Optional[GroupCoordinator] = None
_EMBED_TP: Optional[GroupCoordinator] = None
_MLP_TP: GroupCoordinator | None = None
_OTP: GroupCoordinator | None = None
_LMTP: GroupCoordinator | None = None
_EMBED_TP: GroupCoordinator | None = None
# flashcomm specific groups
_FLASHCOMM2_OTP: Optional[GroupCoordinator] = None
_FLASHCOMM2_ODP: Optional[GroupCoordinator] = None
_FC3_QUANT_X: Optional[GroupCoordinator] = None
_FLASHCOMM2_OTP: GroupCoordinator | None = None
_FLASHCOMM2_ODP: GroupCoordinator | None = None
_FC3_QUANT_X: GroupCoordinator | None = None
# shard_weight across rank groups
_SHARD_WEIGHT: Optional[GroupCoordinator] = None
_SHARD_WEIGHT: GroupCoordinator | None = None
_P_TP: Optional[GroupCoordinator] = None
_P_TP: GroupCoordinator | None = None
def init_ascend_model_parallel(parallel_config: ParallelConfig, ):
def init_ascend_model_parallel(
parallel_config: ParallelConfig,
):
if model_parallel_initialized():
return
assert torch.distributed.is_initialized()
@@ -43,9 +41,9 @@ def init_ascend_model_parallel(parallel_config: ParallelConfig, ):
# ExternalDP is the data parallel group that is not part of the model,
# every dp rank can generate independently (in verl integration).
all_ranks = torch.arange(world_size).reshape(
-1, global_dp_size * parallel_config.prefill_context_parallel_size *
global_tp_size)
#TODO: all_ranks should be the same as vllm_all_ranks, all_ranks needs to be removed in the future.
-1, global_dp_size * parallel_config.prefill_context_parallel_size * global_tp_size
)
# TODO: all_ranks should be the same as vllm_all_ranks, all_ranks needs to be removed in the future.
vllm_all_ranks = torch.arange(world_size).reshape(
-1,
global_dp_size,
@@ -57,49 +55,35 @@ def init_ascend_model_parallel(parallel_config: ParallelConfig, ):
pd_tp_ratio = get_ascend_config().pd_tp_ratio
pd_head_ratio = get_ascend_config().pd_head_ratio
global _P_TP
assert _P_TP is None, (
"distributed prefill tensor parallel group is already initialized")
assert _P_TP is None, "distributed prefill tensor parallel group is already initialized"
prefill_tensor_model_parallel_size = pd_tp_ratio
# divide alltoall groups
if pd_head_ratio > 1 and get_current_vllm_config(
).kv_transfer_config.is_kv_producer:
if pd_head_ratio > 1 and get_current_vllm_config().kv_transfer_config.is_kv_producer:
num_head_replica = get_ascend_config().num_head_replica
remote_tp_size = global_tp_size // pd_tp_ratio
if num_head_replica <= 1:
group_ranks = all_ranks.view(
-1, prefill_tensor_model_parallel_size).unbind(0)
group_ranks = all_ranks.view(-1, prefill_tensor_model_parallel_size).unbind(0)
else:
group_ranks = all_ranks.clone().view(
global_dp_size, -1,
num_head_replica) # [DP_size, num_head, num_head_replica]
global_dp_size, -1, num_head_replica
) # [DP_size, num_head, num_head_replica]
group_ranks = group_ranks.permute(0, 2, 1)
group_ranks = group_ranks.reshape(
-1,
group_ranks.size(-1)) # [DP_size * num_head_replica, num_head]
group_ranks = group_ranks.reshape(-1, group_ranks.size(-1)) # [DP_size * num_head_replica, num_head]
alltoall_group_size = group_ranks.size(-1) // remote_tp_size
group_ranks = group_ranks.unsqueeze(-1).view(
global_dp_size, num_head_replica, -1, alltoall_group_size
) # [DP_size, num_head_replica, num_alltoall_group, alltoall_group_size]
group_ranks = group_ranks.reshape(-1,
alltoall_group_size).unbind(0)
group_ranks = group_ranks.reshape(-1, alltoall_group_size).unbind(0)
group_ranks = [x.tolist() for x in group_ranks]
local_rank = get_world_group().local_rank
num = next(
(i for i, ranks in enumerate(group_ranks) if local_rank in ranks),
None)
_P_TP = init_model_parallel_group(group_ranks,
get_world_group().local_rank,
backend,
group_name=f"p_tp_{num}")
num = next((i for i, ranks in enumerate(group_ranks) if local_rank in ranks), None)
_P_TP = init_model_parallel_group(group_ranks, get_world_group().local_rank, backend, group_name=f"p_tp_{num}")
global _MC2
group_ranks = all_ranks.unbind(0)
group_ranks = [x.tolist() for x in group_ranks]
_MC2 = init_model_parallel_group(group_ranks,
get_world_group().local_rank,
backend,
group_name="mc2")
_MC2 = init_model_parallel_group(group_ranks, get_world_group().local_rank, backend, group_name="mc2")
# Initialize fine-grained TP process groups on Ascend for four components:
# 1. LM Head: output logits projection (`lmhead_tensor_parallel_size`)
@@ -108,39 +92,28 @@ def init_ascend_model_parallel(parallel_config: ParallelConfig, ):
# 4. MLP: feed-forward network in transformer blocks (`mlp_tensor_parallel_size`)
_group_cache = {}
def _create_or_get_group(group_size: int,
group_name: str) -> GroupCoordinator:
def _create_or_get_group(group_size: int, group_name: str) -> GroupCoordinator:
if group_size is None:
return None
if group_size not in _group_cache:
rank_grid = torch.arange(world_size).reshape(
global_pp_size, global_dp_size, global_tp_size)
rank_grid = torch.arange(world_size).reshape(global_pp_size, global_dp_size, global_tp_size)
num_chunks = global_dp_size // group_size
group_ranks = []
for pp_idx in range(global_pp_size):
stage_ranks = rank_grid[pp_idx] # (dp, tp)
for chunk in range(num_chunks):
for tp_idx in range(global_tp_size):
group = stage_ranks[chunk * group_size:(chunk + 1) *
group_size, tp_idx].tolist()
group = stage_ranks[chunk * group_size : (chunk + 1) * group_size, tp_idx].tolist()
group_ranks.append(group)
pg = init_model_parallel_group(group_ranks,
get_world_group().local_rank,
backend,
group_name=group_name)
pg = init_model_parallel_group(group_ranks, get_world_group().local_rank, backend, group_name=group_name)
_group_cache[group_size] = pg
return _group_cache[group_size]
otp_size = get_ascend_config(
).finegrained_tp_config.oproj_tensor_parallel_size
lmhead_tp_size = get_ascend_config(
).finegrained_tp_config.lmhead_tensor_parallel_size
embedding_tp_size = get_ascend_config(
).finegrained_tp_config.embedding_tensor_parallel_size
mlp_tp_size = get_ascend_config(
).finegrained_tp_config.mlp_tensor_parallel_size
otp_size = get_ascend_config().finegrained_tp_config.oproj_tensor_parallel_size
lmhead_tp_size = get_ascend_config().finegrained_tp_config.lmhead_tensor_parallel_size
embedding_tp_size = get_ascend_config().finegrained_tp_config.embedding_tensor_parallel_size
mlp_tp_size = get_ascend_config().finegrained_tp_config.mlp_tensor_parallel_size
global _OTP, _LMTP, _EMBED_TP, _MLP_TP
@@ -156,10 +129,8 @@ def init_ascend_model_parallel(parallel_config: ParallelConfig, ):
# TODO: Extract and unify the logic across different communication group.
flashcomm2_otp_group_ranks = []
if flashcomm2_enable():
flashcomm2_otp_size = get_ascend_config(
).flashcomm2_oproj_tensor_parallel_size
num_fc2_oproj_tensor_parallel_groups: int = (global_tp_size //
flashcomm2_otp_size)
flashcomm2_otp_size = get_ascend_config().flashcomm2_oproj_tensor_parallel_size
num_fc2_oproj_tensor_parallel_groups: int = global_tp_size // flashcomm2_otp_size
global _FLASHCOMM2_OTP
global _FLASHCOMM2_ODP
@@ -168,8 +139,7 @@ def init_ascend_model_parallel(parallel_config: ParallelConfig, ):
if flashcomm2_otp_size > 1:
odp_group_ranks: list[list[int]] = [
[] for _ in range(flashcomm2_otp_size * global_dp_size *
global_pp_size)
[] for _ in range(flashcomm2_otp_size * global_dp_size * global_pp_size)
]
for dp_group_index in range(global_dp_size):
for pp_group_index in range(global_pp_size):
@@ -186,31 +156,24 @@ def init_ascend_model_parallel(parallel_config: ParallelConfig, ):
ranks.append(global_rank)
odp_group_index = odp_base_index + j
odp_group_ranks[odp_group_index].append(
global_rank)
odp_group_ranks[odp_group_index].append(global_rank)
flashcomm2_otp_group_ranks.append(ranks)
_FLASHCOMM2_OTP = init_model_parallel_group(
flashcomm2_otp_group_ranks,
get_world_group().local_rank,
backend,
group_name="flashcomm2_otp")
flashcomm2_otp_group_ranks, get_world_group().local_rank, backend, group_name="flashcomm2_otp"
)
_FLASHCOMM2_ODP = init_model_parallel_group(
odp_group_ranks,
get_world_group().local_rank,
backend,
group_name="flashcomm2_odp")
odp_group_ranks, get_world_group().local_rank, backend, group_name="flashcomm2_odp"
)
def create_shard_weight_group(
module_tp_group_ranks: None) -> GroupCoordinator:
def create_shard_weight_group(module_tp_group_ranks: None) -> GroupCoordinator:
# Argument module_tp_group_ranks: The module specific tensor parallel group.
# There are three situations.
# 1. If it is None, then the TP_size of the specific module is 1 and is replicated linear layer.
# 2. If it is not None, and the module tp_group is same as the global tp_group.
# 3. If it is not None, and the module tp_group is different from the global tp_group.(eg. flashcomm2_otp)
group_ranks = []
pp_group_ranks = vllm_all_ranks.transpose(2, 4).reshape(
-1, global_pp_size)
pp_group_ranks = vllm_all_ranks.transpose(2, 4).reshape(-1, global_pp_size)
if module_tp_group_ranks is None:
# If it is None, then the TP_size of this shard weight is 1.
shard_weight_group_ranks = pp_group_ranks.transpose(0, 1).unbind(0)
@@ -219,14 +182,9 @@ def init_ascend_model_parallel(parallel_config: ParallelConfig, ):
# combine standard tp group and non-standard tp group to build shard_weight comm_group
module_tp_tanspose_ranks = module_tp_group_ranks.transpose(0, 1)
G = world_size // (global_pp_size * module_tp_group_ranks.size(1))
shard_weight_group_ranks = torch.stack(
[t.view(global_pp_size, G) for t in module_tp_tanspose_ranks],
dim=1)
shard_weight_group_ranks = torch.stack([t.view(global_pp_size, G) for t in module_tp_tanspose_ranks], dim=1)
group_ranks = shard_weight_group_ranks.view(-1, G).tolist()
return init_model_parallel_group(group_ranks,
get_world_group().local_rank,
backend,
group_name="shard_weight")
return init_model_parallel_group(group_ranks, get_world_group().local_rank, backend, group_name="shard_weight")
# Create shard weight group if enabled
if get_ascend_config().layer_sharding is not None:
@@ -235,8 +193,7 @@ def init_ascend_model_parallel(parallel_config: ParallelConfig, ):
if len(flashcomm2_otp_group_ranks) == 0:
FC2_group_ranks = None
else:
FC2_group_ranks = torch.tensor(
flashcomm2_otp_group_ranks).squeeze(0)
FC2_group_ranks = torch.tensor(flashcomm2_otp_group_ranks).squeeze(0)
_SHARD_WEIGHT = create_shard_weight_group(FC2_group_ranks)
elif enable_dsa_cp_with_layer_shard():
# For dsa_cp, all shard layers are replicated.
@@ -250,40 +207,37 @@ def init_ascend_model_parallel(parallel_config: ParallelConfig, ):
global _FC3_QUANT_X
group_ranks = all_ranks.unbind(0)
group_ranks = [x.tolist() for x in group_ranks]
_FC3_QUANT_X = init_model_parallel_group(group_ranks,
get_world_group().local_rank,
backend,
group_name="fc3_quant_x")
_FC3_QUANT_X = init_model_parallel_group(
group_ranks, get_world_group().local_rank, backend, group_name="fc3_quant_x"
)
def model_parallel_initialized():
return (_MC2 is not None)
return _MC2 is not None
def get_mc2_group() -> GroupCoordinator:
assert _MC2 is not None, ("mc2 group is not initialized")
assert _MC2 is not None, "mc2 group is not initialized"
return _MC2
def get_mlp_tp_group() -> GroupCoordinator:
assert _MLP_TP is not None, ("mlp group is not initialized")
assert _MLP_TP is not None, "mlp group is not initialized"
return _MLP_TP
def get_otp_group() -> GroupCoordinator:
assert _OTP is not None, (
"output tensor parallel group is not initialized")
assert _OTP is not None, "output tensor parallel group is not initialized"
return _OTP
def get_lmhead_tp_group() -> GroupCoordinator:
assert _LMTP is not None, (
"lm head tensor parallel group is not initialized")
assert _LMTP is not None, "lm head tensor parallel group is not initialized"
return _LMTP
def get_embed_tp_group() -> GroupCoordinator:
assert _EMBED_TP is not None, ("emtp group is not initialized")
assert _EMBED_TP is not None, "emtp group is not initialized"
return _EMBED_TP
@@ -292,25 +246,22 @@ def get_flashcomm2_otp_group() -> GroupCoordinator:
def get_flashcomm2_odp_group() -> GroupCoordinator:
assert _FLASHCOMM2_ODP is not None, (
"output data parallel group for flashcomm2 is not initialized")
assert _FLASHCOMM2_ODP is not None, "output data parallel group for flashcomm2 is not initialized"
return _FLASHCOMM2_ODP
def get_shard_weight_group() -> GroupCoordinator:
assert _SHARD_WEIGHT is not None, (
"output shard weight parallel group for flashcomm2 is not initialized")
assert _SHARD_WEIGHT is not None, "output shard weight parallel group for flashcomm2 is not initialized"
return _SHARD_WEIGHT
def get_p_tp_group() -> GroupCoordinator:
assert _P_TP is not None, (
"distributed prefill tensor parallel group is not initialized")
assert _P_TP is not None, "distributed prefill tensor parallel group is not initialized"
return _P_TP
def get_fc3_quant_x_group() -> GroupCoordinator:
assert _FC3_QUANT_X is not None, ("fc3 quant x group is not initialized")
assert _FC3_QUANT_X is not None, "fc3 quant x group is not initialized"
return _FC3_QUANT_X
@@ -346,14 +297,12 @@ def destroy_ascend_model_parallel():
_P_TP = None
global _FLASHCOMM2_OTP
if _FLASHCOMM2_OTP and get_ascend_config(
).flashcomm2_oproj_tensor_parallel_size != 1:
if _FLASHCOMM2_OTP and get_ascend_config().flashcomm2_oproj_tensor_parallel_size != 1:
_FLASHCOMM2_OTP.destroy()
_FLASHCOMM2_OTP = None
global _FLASHCOMM2_ODP
if _FLASHCOMM2_ODP and get_ascend_config(
).flashcomm2_oproj_tensor_parallel_size != 1:
if _FLASHCOMM2_ODP and get_ascend_config().flashcomm2_oproj_tensor_parallel_size != 1:
_FLASHCOMM2_ODP.destroy()
_FLASHCOMM2_ODP = None

30
vllm_ascend/distributed/utils.py
View File

@@ -1,5 +1,3 @@
from typing import Optional
import torch
import torch.distributed as dist
from vllm.distributed.parallel_state import GroupCoordinator, get_dp_group
@@ -8,7 +6,9 @@ from vllm.forward_context import get_forward_context
from vllm_ascend.distributed.parallel_state import get_fc3_quant_x_group
def fc3_all_gather_and_maybe_unpad_impl(x: torch.Tensor, ) -> torch.Tensor:
def fc3_all_gather_and_maybe_unpad_impl(
x: torch.Tensor,
) -> torch.Tensor:
try:
forward_context = get_forward_context()
except AssertionError:
@@ -22,34 +22,26 @@ def fc3_all_gather_and_maybe_unpad_impl(x: torch.Tensor, ) -> torch.Tensor:
else:
# unpad
num_tokens_across_dp_cpu = dp_metadata.num_tokens_across_dp_cpu
result = torch.empty((num_tokens_across_dp_cpu.sum(), *x.shape[1:]),
device=x.device,
dtype=x.dtype)
result = torch.empty((num_tokens_across_dp_cpu.sum(), *x.shape[1:]), device=x.device, dtype=x.dtype)
dp_size = get_dp_group().world_size
x = x.view(dp_size, forward_context.padded_length, *x.shape[1:])
offset = 0
for idx in range(dp_size):
num_tokens_dp = num_tokens_across_dp_cpu[idx]
result[offset:offset + num_tokens_dp] = x[idx, :num_tokens_dp]
result[offset : offset + num_tokens_dp] = x[idx, :num_tokens_dp]
offset += num_tokens_dp
x = result
return x
def all_gather_async(input: torch.Tensor,
group: GroupCoordinator,
output: Optional[torch.Tensor] = None,
async_op: bool = True):
def all_gather_async(
input: torch.Tensor, group: GroupCoordinator, output: torch.Tensor | None = None, async_op: bool = True
):
if group.world_size == 1:
return input, None
if output is None:
input_size = input.size()
output_size = (input_size[0] * group.world_size, ) + input_size[1:]
output = torch.empty(output_size,
dtype=input.dtype,
device=input.device)
return output, dist.all_gather_into_tensor(output,
input,
group=group.device_group,
async_op=async_op)
output_size = (input_size[0] * group.world_size,) + input_size[1:]
output = torch.empty(output_size, dtype=input.dtype, device=input.device)
return output, dist.all_gather_into_tensor(output, input, group=group.device_group, async_op=async_op)

104
vllm_ascend/ops/layer_shard_linear.py
View File

@@ -1,6 +1,6 @@
from collections.abc import Callable
from dataclasses import dataclass
from functools import lru_cache
from typing import Callable, List, Optional
import torch
import torch.distributed as dist
@@ -17,39 +17,38 @@ def dispose_tensor(x: torch.Tensor):
@dataclass
class LayerMetadata:
"""Metadata for a layer.
"""
"""Metadata for a layer."""
layer_idx: int # The index of the layer.
layer: LinearBase # The layer object.
post_method: Callable[[
torch.nn.Module
], None] # The `process_weights_after_loading` method from the quant method.
post_method: Callable[[torch.nn.Module], None] # The `process_weights_after_loading` method from the quant method.
weight: torch.Tensor # The weight tensor.
window_idx: int # The index of the window.
@dataclass
class ShardWindowMetadata:
"""Metadata for a shard window.
"""
"""Metadata for a shard window."""
weight: torch.Tensor # The weight tensor to be shard by layers.
data_layer_idx: int # The index of the layer this window's weight is equal to.
work: Optional[torch.distributed.Work] # The asynchronous broadcast work.
work: torch.distributed.Work | None # The asynchronous broadcast work.
@dataclass
class SeriesMetadata:
"""Metadata for a weight shard series.
"""
"""Metadata for a weight shard series."""
group: GroupCoordinator
start_layer: int
end_layer: int
num_layers: int
prefetch_step: int
dummy_weight: torch.Tensor # Dummy weight to replace the loaded weight matrix. All the layers in the series share the same dummy weight tensor.
dummy_weight: torch.Tensor # Dummy weight to replace the loaded weight matrix.
# All the layers in the series share the same dummy weight tensor.
layers: list[LayerMetadata]
shard_windows: list[
ShardWindowMetadata] # Shard windows for prefetching. The window size is (`prefetch_step` + 1), as only the weights for the next (`prefetch_step` + 1) layers need to be stored.
shard_windows: list[ShardWindowMetadata] # Shard windows for prefetching. The window size is (`prefetch_step` + 1),
# as only the weights for the next (`prefetch_step` + 1) layers need to be stored.
window_offset: int # The index of the window for the next coming layer.
def is_source(self, layer_idx) -> bool:
@@ -63,9 +62,9 @@ class SeriesMetadata:
self.layers.sort(key=lambda x: x.layer_idx)
self.num_layers = len(self.layers)
assert self.num_layers > 0, "No layers in the series"
assert self.prefetch_step >= 0 and self.prefetch_step <= max(
0, self.num_layers -
2), "prefetch_step must be in [0, num_layers - 2]"
assert self.prefetch_step >= 0 and self.prefetch_step <= max(0, self.num_layers - 2), (
"prefetch_step must be in [0, num_layers - 2]"
)
self.start_layer = self.layers[0].layer_idx
self.end_layer = self.layers[-1].layer_idx + 1
@@ -73,25 +72,27 @@ class SeriesMetadata:
layer = self.layers[layer_idx - self.start_layer]
assert layer.layer_idx == layer_idx, "layer_idx must be consecutive"
is_source = self.is_source(layer_idx)
# If the weight uses dummy weight, make a copy temporary such that the post method call won't affect other layers which also uses dummy weight.
# If the weight uses dummy weight, make a copy temporary such that the post method call
# won't affect other layers which also uses dummy weight.
if not is_source:
layer.weight.set_(torch.empty_like(self.dummy_weight))
# Broadcast to get the true weight.
dist.broadcast(layer.weight,
src=self.group.ranks[layer_idx %
self.group.world_size],
group=self.group.device_group)
dist.broadcast(
layer.weight, src=self.group.ranks[layer_idx % self.group.world_size], group=self.group.device_group
)
# Call `process_weights_after_loading` from the quant method.
layer.post_method(layer.layer)
step = layer_idx - self.start_layer
if step < self.prefetch_step:
# Build the windows for the first `prefetch_step` layers. The weights can be used for the first `prefetch_step` layers in `forward()`, so also clone the weights.
# Build the windows for the first `prefetch_step` layers. The weights can be used
# for the first `prefetch_step` layers in `forward()`, so also clone the weights.
self.shard_windows.append(
ShardWindowMetadata(
weight=layer.weight.clone().detach(),
data_layer_idx=layer_idx,
work=None,
))
)
)
layer.window_idx = step
# When the layer not intended to be stored in this device, link to the corresponding window's tensor.
if not is_source:
@@ -104,7 +105,8 @@ class SeriesMetadata:
weight=torch.empty_like(layer.weight),
data_layer_idx=-1,
work=None,
))
)
)
# When the layer not intended to be stored in this device, dispose the tensor.
if not is_source:
dispose_tensor(layer.weight)
@@ -113,8 +115,7 @@ class SeriesMetadata:
def reach_layer(self, layer_idx: int):
# The index of the layer to be prefetched.
next_layer_idx = (layer_idx + self.prefetch_step
) % self.num_layers + self.start_layer
next_layer_idx = (layer_idx + self.prefetch_step) % self.num_layers + self.start_layer
next_layer = self.layers[next_layer_idx - self.start_layer]
# The index of the window to store the weight for the coming layer.
next_layer.window_idx = self.window_offset
@@ -123,8 +124,7 @@ class SeriesMetadata:
if not self.is_source(next_layer_idx):
next_layer.weight.set_(window.weight)
# Update `window_offset` by rolling one step.
self.window_offset = (self.window_offset + 1) % (self.prefetch_step +
1)
self.window_offset = (self.window_offset + 1) % (self.prefetch_step + 1)
assert window.data_layer_idx != next_layer_idx
window.data_layer_idx = next_layer_idx
# Start asynchronous broadcast work.
@@ -132,13 +132,13 @@ class SeriesMetadata:
next_layer.weight,
src=self.group.ranks[next_layer_idx % self.group.world_size],
group=self.group.device_group,
async_op=True)
async_op=True,
)
def wait_weight(self, layer_idx: int):
# Find the asynchronous broadcast work and wait for it.
assert self.shard_windows
window = self.shard_windows[self.layers[layer_idx -
self.start_layer].window_idx]
window = self.shard_windows[self.layers[layer_idx - self.start_layer].window_idx]
# Make sure the data in the corresponding shard window is for the current layer.
assert window.data_layer_idx == layer_idx
if window.work is not None:
@@ -148,8 +148,8 @@ class SeriesMetadata:
@dataclass
class LayerExternalMetadata:
"""External metadata for a layer.
"""
"""External metadata for a layer."""
series: SeriesMetadata
layer_idx: int
@@ -159,9 +159,7 @@ _series_dict: dict[str, SeriesMetadata] = {}
_layer_external_dict: dict[int, LayerExternalMetadata] = {}
def _create_forward_wrapper(forward: Callable, series: SeriesMetadata,
layer_idx: int) -> Callable:
def _create_forward_wrapper(forward: Callable, series: SeriesMetadata, layer_idx: int) -> Callable:
def wrapped_forward(*args, **kwargs):
# Wait for the weight.
series.wait_weight(layer_idx)
@@ -173,23 +171,32 @@ def _create_forward_wrapper(forward: Callable, series: SeriesMetadata,
"""
Register linear layers into a shard storage series.
In a parallel group, each device stores a distinct, non-overlapping subset of layers from the series. All layers in a series must have the same structure (are isomorphic). The weight matrix for the i-th layer is stored on device (i % n), where n is the number of devices.
In a parallel group, each device stores a distinct, non-overlapping subset of layers from the series.
All layers in a series must have the same structure (are isomorphic). The weight matrix for the i-th layer
is stored on device (i % n), where n is the number of devices.
After loading the model, you must call `post_process_after_loading_for_shard_weight_series(layer)` on any layer of this series to complete the initialization.
After loading the model, you must call `post_process_after_loading_for_shard_weight_series(layer)`
on any layer of this series to complete the initialization.
During execution, each time a new layer is reached, you must call `reach_layer_for_shard_weight_series(layer)` for that layer to prefetch the weights. The argument `prefetch_step` is a non-negative integer k that manages asynchronous weight prefetching. Each call to `reach_layer_for_shard_weight_series(current_layer)` method will trigger an asynchronous prefetch for the weights of the k-th subsequent layer after `current_layer` within the series.
During execution, each time a new layer is reached, you must call `reach_layer_for_shard_weight_series(layer)`
for that layer to prefetch the weights. The argument `prefetch_step` is a non-negative integer k that manages
asynchronous weight prefetching. Each call to `reach_layer_for_shard_weight_series(current_layer)` method will
trigger an asynchronous prefetch for the weights of the k-th subsequent layer after `current_layer` within the series.
Note: The layers are managed as a circular buffer. The index of the layer to prefetch is determined by the formula:
- start_layer is the index of the first layer in the series (inclusive).
- end_layer is the index of the last layer in the series (exclusive). Thus, the series includes all layers with indices in the range [start_layer, end_layer).
- end_layer is the index of the last layer in the series (exclusive). Thus, the series includes all layers with
indices in the range [start_layer, end_layer).
- total_layers = end_layer - start_layer
- prefetch_layer_idx = (layer_idx + prefetch_step) % total_layers + start_layer
To hold the weights for the current layer and the k prefetched layers, a pool of (k + 1) shard tensor buffers will be created for this series.
To hold the weights for the current layer and the k prefetched layers, a pool of (k + 1) shard tensor buffers
will be created for this series.
Arguments:
series_name: This name identifies which series this layer belongs to.
group: The group coordinator for handling asynchronous communications. It is recommended to create a new group coordinator for each new series.
group: The group coordinator for handling asynchronous communications. It is recommended to create a new group
coordinator for each new series.
layer: The linear layer object to register.
prefetch_step: An integer that manages asynchronous weight prefetching. Setting it to 0 or 1 can cover most cases.
"""
@@ -224,7 +231,8 @@ def register_layer_to_shard_weight_series(
post_method=layer.quant_method.process_weights_after_loading,
weight=layer.weight,
window_idx=-1,
))
)
)
# Discard the original `process_weights_after_loading` method such that it won't be called by others.
layer.quant_method.process_weights_after_loading = lambda layer: None
# When the layer not intended to be stored in this device, dispose the tensor and skip weight loading.
@@ -257,6 +265,7 @@ def wait_layer_for_shard_weight_series(layer: LinearBase):
@lru_cache(maxsize=1)
def get_current_model_num_hidden_layers() -> int:
from vllm.config import get_current_vllm_config
vllm_config = get_current_vllm_config()
return vllm_config.model_config.get_total_num_hidden_layers()
@@ -268,10 +277,11 @@ def is_hidden_layer(layer: LinearBase) -> bool:
def register_all_layers_to_shard_weight_series(
layer_sharding: List[LinearBase], ):
for curr_layer in (layer_sharding or []):
layer_sharding: list[LinearBase],
):
for curr_layer in layer_sharding or []:
if is_hidden_layer(curr_layer):
layer_name = curr_layer.prefix.split('.')[-1]
layer_name = curr_layer.prefix.split(".")[-1]
register_layer_to_shard_weight_series(
series_name=layer_name,
group=get_shard_weight_group(),

285
vllm_ascend/ops/linear.py
View File

@@ -20,19 +20,23 @@ AscendMergedColumnParallelLinear, AscendMergedColumnParallelLinear,
AscendRowParallelLinear and AscendColumnParallelLinear.
"""
from typing import Optional, Union
import torch
import torch.nn as nn
from torch.nn.parameter import Parameter
from vllm.config import get_current_vllm_config
from vllm.distributed import divide
from vllm.model_executor.layers.linear import ( # noqa
WEIGHT_LOADER_V2_SUPPORTED, ColumnParallelLinear, LinearBase,
MergedColumnParallelLinear, QKVParallelLinear, QuantizeMethodBase,
ReplicatedLinear, RowParallelLinear, UnquantizedLinearMethod)
from vllm.model_executor.layers.quantization.base_config import \
QuantizationConfig
WEIGHT_LOADER_V2_SUPPORTED,
ColumnParallelLinear,
LinearBase,
MergedColumnParallelLinear,
QKVParallelLinear,
QuantizeMethodBase,
ReplicatedLinear,
RowParallelLinear,
UnquantizedLinearMethod,
)
from vllm.model_executor.layers.quantization.base_config import QuantizationConfig
from vllm.model_executor.utils import set_weight_attrs
from vllm_ascend.ops.linear_op import get_parallel_op, get_replicated_op
@@ -50,14 +54,13 @@ class AscendUnquantizedLinearMethod(UnquantizedLinearMethod):
# TODO(realliujiaxu): Remove this class after linear of vllm supports custom comm group
class AscendLinearBase(LinearBase):
def __init__(
self,
input_size: int,
output_size: int,
skip_bias_add: bool = False,
params_dtype: Optional[torch.dtype] = None,
quant_config: Optional[QuantizationConfig] = None,
params_dtype: torch.dtype | None = None,
quant_config: QuantizationConfig | None = None,
prefix: str = "",
*,
return_bias: bool = True,
@@ -75,11 +78,9 @@ class AscendLinearBase(LinearBase):
self.quant_config = quant_config
self.prefix = prefix
if quant_config is None:
self.quant_method: Optional[
QuantizeMethodBase] = AscendUnquantizedLinearMethod()
self.quant_method: QuantizeMethodBase | None = AscendUnquantizedLinearMethod()
else:
self.quant_method = quant_config.get_quant_method(self,
prefix=prefix)
self.quant_method = quant_config.get_quant_method(self, prefix=prefix)
self.return_bias = return_bias
self.disable_tp = disable_tp
@@ -100,11 +101,11 @@ class AscendQKVParallelLinear(QKVParallelLinear):
hidden_size: int,
head_size: int,
total_num_heads: int,
total_num_kv_heads: Optional[int] = None,
total_num_kv_heads: int | None = None,
bias: bool = True,
skip_bias_add: bool = False,
params_dtype: Optional[torch.dtype] = None,
quant_config: Optional[QuantizationConfig] = None,
params_dtype: torch.dtype | None = None,
quant_config: QuantizationConfig | None = None,
prefix: str = "",
*,
return_bias: bool = True,
@@ -112,9 +113,9 @@ class AscendQKVParallelLinear(QKVParallelLinear):
v_head_size: int | None = None,
):
self.v_head_size = v_head_size if v_head_size is not None else head_size
self.custom_op, _, tp_size = get_parallel_op(disable_tp, prefix, self,
"column")
# TODO(realliujiaxu): Replace the initialization code below with super().__init__ after linear of vllm supports custom comm group
self.custom_op, _, tp_size = get_parallel_op(disable_tp, prefix, self, "column")
# TODO(realliujiaxu): Replace the initialization code below with super().__init__ after
# linear of vllm supports custom comm group
self.hidden_size = hidden_size
self.head_size = head_size
self.total_num_heads = total_num_heads
@@ -125,35 +126,35 @@ class AscendQKVParallelLinear(QKVParallelLinear):
self.num_heads = divide(self.total_num_heads, tp_size)
if tp_size >= self.total_num_kv_heads:
self.num_kv_heads = 1
self.num_kv_head_replicas = divide(tp_size,
self.total_num_kv_heads)
self.num_kv_head_replicas = divide(tp_size, self.total_num_kv_heads)
else:
self.num_kv_heads = divide(self.total_num_kv_heads, tp_size)
self.num_kv_head_replicas = 1
input_size = self.hidden_size
output_size = (self.num_heads +
2 * self.num_kv_heads) * tp_size * self.head_size
output_size = (self.num_heads + 2 * self.num_kv_heads) * tp_size * self.head_size
self.output_sizes = [
self.num_heads * self.head_size * tp_size, # q_proj
self.num_kv_heads * self.head_size * tp_size, # k_proj
self.num_kv_heads * self.head_size * tp_size, # v_proj
]
AscendColumnParallelLinear.__init__(self,
input_size=input_size,
output_size=output_size,
bias=bias,
gather_output=False,
skip_bias_add=skip_bias_add,
params_dtype=params_dtype,
quant_config=quant_config,
prefix=prefix,
return_bias=return_bias,
disable_tp=disable_tp)
AscendColumnParallelLinear.__init__(
self,
input_size=input_size,
output_size=output_size,
bias=bias,
gather_output=False,
skip_bias_add=skip_bias_add,
params_dtype=params_dtype,
quant_config=quant_config,
prefix=prefix,
return_bias=return_bias,
disable_tp=disable_tp,
)
def forward(
self,
input_,
) -> Union[torch.Tensor, tuple[torch.Tensor, Optional[Parameter]]]:
) -> torch.Tensor | tuple[torch.Tensor, Parameter | None]:
if self.custom_op is not None:
return self.custom_op.apply(input_)
@@ -178,35 +179,36 @@ class AscendMergedColumnParallelLinear(MergedColumnParallelLinear):
bias: bool = True,
gather_output: bool = False,
skip_bias_add: bool = False,
params_dtype: Optional[torch.dtype] = None,
quant_config: Optional[QuantizationConfig] = None,
params_dtype: torch.dtype | None = None,
quant_config: QuantizationConfig | None = None,
prefix: str = "",
*,
return_bias: bool = True,
disable_tp: bool = False,
):
self.custom_op, self.tp_rank, self.tp_size = get_parallel_op(
disable_tp, prefix, self, "column")
# TODO(realliujiaxu): Replace the initialization code below with super().__init__ after linear of vllm supports custom comm group
self.custom_op, self.tp_rank, self.tp_size = get_parallel_op(disable_tp, prefix, self, "column")
# TODO(realliujiaxu): Replace the initialization code below with super().__init__ after
# linear of vllm supports custom comm group
self.output_sizes = output_sizes
assert all(output_size % self.tp_size == 0
for output_size in output_sizes)
AscendColumnParallelLinear.__init__(self,
input_size=input_size,
output_size=sum(output_sizes),
bias=bias,
gather_output=gather_output,
skip_bias_add=skip_bias_add,
params_dtype=params_dtype,
quant_config=quant_config,
prefix=prefix,
return_bias=return_bias,
disable_tp=disable_tp)
assert all(output_size % self.tp_size == 0 for output_size in output_sizes)
AscendColumnParallelLinear.__init__(
self,
input_size=input_size,
output_size=sum(output_sizes),
bias=bias,
gather_output=gather_output,
skip_bias_add=skip_bias_add,
params_dtype=params_dtype,
quant_config=quant_config,
prefix=prefix,
return_bias=return_bias,
disable_tp=disable_tp,
)
def forward(
self,
input_,
) -> Union[torch.Tensor, tuple[torch.Tensor, Optional[Parameter]]]:
) -> torch.Tensor | tuple[torch.Tensor, Parameter | None]:
if self.custom_op is not None:
return self.custom_op.apply(input_)
@@ -229,9 +231,9 @@ class AscendRowParallelLinear(RowParallelLinear):
bias: bool = True,
input_is_parallel: bool = True,
skip_bias_add: bool = False,
params_dtype: Optional[torch.dtype] = None,
params_dtype: torch.dtype | None = None,
reduce_results: bool = True,
quant_config: Optional[QuantizationConfig] = None,
quant_config: QuantizationConfig | None = None,
prefix: str = "",
*,
return_bias: bool = True,
@@ -247,23 +249,25 @@ class AscendRowParallelLinear(RowParallelLinear):
self.unique_prefix = unique_prefix
compilation_config.static_forward_context[unique_prefix] = self
self.custom_op, self.tp_rank, self.tp_size = get_parallel_op(
disable_tp, prefix, self, "row")
# TODO(realliujiaxu): Replace the initialization code below with super().__init__ after linear of vllm supports custom comm group
self.custom_op, self.tp_rank, self.tp_size = get_parallel_op(disable_tp, prefix, self, "row")
# TODO(realliujiaxu): Replace the initialization code below with super().__init__ after
# linear of vllm supports custom comm group
# Divide the weight matrix along the first dimension.
self.input_size_per_partition = divide(input_size, self.tp_size)
self.output_size_per_partition = output_size
self.output_partition_sizes = [output_size]
AscendLinearBase.__init__(self,
input_size,
output_size,
skip_bias_add,
params_dtype,
quant_config,
prefix,
return_bias=return_bias,
disable_tp=disable_tp)
AscendLinearBase.__init__(
self,
input_size,
output_size,
skip_bias_add,
params_dtype,
quant_config,
prefix,
return_bias=return_bias,
disable_tp=disable_tp,
)
self.input_is_parallel = input_is_parallel
self.reduce_results = reduce_results
@@ -277,19 +281,23 @@ class AscendRowParallelLinear(RowParallelLinear):
output_size=self.output_size,
params_dtype=self.params_dtype,
weight_loader=(
self.weight_loader_v2 if self.quant_method.__class__.__name__
in WEIGHT_LOADER_V2_SUPPORTED else self.weight_loader))
self.weight_loader_v2
if self.quant_method.__class__.__name__ in WEIGHT_LOADER_V2_SUPPORTED
else self.weight_loader
),
)
if not reduce_results and (bias and not skip_bias_add):
raise ValueError("When not reduce the results, adding bias to the "
"results can lead to incorrect results")
raise ValueError("When not reduce the results, adding bias to the results can lead to incorrect results")
if bias:
self.bias = Parameter(
torch.empty(self.output_size, dtype=params_dtype))
set_weight_attrs(self.bias, {
"output_dim": 0,
"weight_loader": self.weight_loader,
})
self.bias = Parameter(torch.empty(self.output_size, dtype=params_dtype))
set_weight_attrs(
self.bias,
{
"output_dim": 0,
"weight_loader": self.weight_loader,
},
)
else:
self.register_parameter("bias", None)
@@ -300,7 +308,7 @@ class AscendRowParallelLinear(RowParallelLinear):
self,
input_,
**kwargs,
) -> Union[torch.Tensor, tuple[torch.Tensor, Optional[Parameter]]]:
) -> torch.Tensor | tuple[torch.Tensor, Parameter | None]:
if self.custom_op is not None:
return self.custom_op.apply(input_)
@@ -321,36 +329,36 @@ class AscendColumnParallelLinear(ColumnParallelLinear):
bias: bool = True,
gather_output: bool = False,
skip_bias_add: bool = False,
params_dtype: Optional[torch.dtype] = None,
quant_config: Optional[QuantizationConfig] = None,
output_sizes: Optional[list[int]] = None,
params_dtype: torch.dtype | None = None,
quant_config: QuantizationConfig | None = None,
output_sizes: list[int] | None = None,
prefix: str = "",
*,
return_bias: bool = True,
disable_tp: bool = False,
):
self.custom_op, self.tp_rank, self.tp_size = get_parallel_op(
disable_tp, prefix, self, "column")
# TODO(realliujiaxu): Replace the initialization code below with super().__init__ after linear of vllm supports custom comm group
#
self.custom_op, self.tp_rank, self.tp_size = get_parallel_op(disable_tp, prefix, self, "column")
# TODO(realliujiaxu): Replace the initialization code below with super().__init__ after
# linear of vllm supports custom comm group
self.input_size_per_partition = input_size
self.output_size_per_partition = divide(output_size, self.tp_size)
self.output_partition_sizes = [self.output_size_per_partition]
# If QKV or MergedColumn, use output size of each partition.
if hasattr(self, "output_sizes"):
self.output_partition_sizes = [
divide(output_size, self.tp_size)
for output_size in self.output_sizes
]
self.output_partition_sizes = [divide(output_size, self.tp_size) for output_size in self.output_sizes]
AscendLinearBase.__init__(self,
input_size,
output_size,
skip_bias_add,
params_dtype,
quant_config,
prefix,
return_bias=return_bias,
disable_tp=disable_tp)
AscendLinearBase.__init__(
self,
input_size,
output_size,
skip_bias_add,
params_dtype,
quant_config,
prefix,
return_bias=return_bias,
disable_tp=disable_tp,
)
self.gather_output = gather_output
@@ -366,16 +374,20 @@ class AscendColumnParallelLinear(ColumnParallelLinear):
output_size=self.output_size,
params_dtype=self.params_dtype,
weight_loader=(
self.weight_loader_v2 if self.quant_method.__class__.__name__
in WEIGHT_LOADER_V2_SUPPORTED else self.weight_loader))
self.weight_loader_v2
if self.quant_method.__class__.__name__ in WEIGHT_LOADER_V2_SUPPORTED
else self.weight_loader
),
)
if bias:
self.bias = Parameter(
torch.empty(self.output_size_per_partition,
dtype=params_dtype))
set_weight_attrs(self.bias, {
"output_dim": 0,
"weight_loader": self.weight_loader,
})
self.bias = Parameter(torch.empty(self.output_size_per_partition, dtype=params_dtype))
set_weight_attrs(
self.bias,
{
"output_dim": 0,
"weight_loader": self.weight_loader,
},
)
else:
self.register_parameter("bias", None)
@@ -385,7 +397,7 @@ class AscendColumnParallelLinear(ColumnParallelLinear):
def forward(
self,
input_,
) -> Union[torch.Tensor, tuple[torch.Tensor, Optional[Parameter]]]:
) -> torch.Tensor | tuple[torch.Tensor, Parameter | None]:
if self.custom_op is not None:
return self.custom_op.apply(input_)
@@ -414,8 +426,8 @@ class AscendReplicatedLinear(ReplicatedLinear):
output_size: int,
bias: bool = True,
skip_bias_add: bool = False,
params_dtype: Optional[torch.dtype] = None,
quant_config: Optional[QuantizationConfig] = None,
params_dtype: torch.dtype | None = None,
quant_config: QuantizationConfig | None = None,
prefix: str = "",
*,
return_bias: bool = True,
@@ -428,32 +440,39 @@ class AscendReplicatedLinear(ReplicatedLinear):
else:
self.output_partition_sizes = [output_size]
AscendLinearBase.__init__(self,
input_size,
output_size,
skip_bias_add,
params_dtype,
quant_config,
prefix=prefix,
return_bias=return_bias,
disable_tp=disable_tp)
AscendLinearBase.__init__(
self,
input_size,
output_size,
skip_bias_add,
params_dtype,
quant_config,
prefix=prefix,
return_bias=return_bias,
disable_tp=disable_tp,
)
# All the linear layer supports quant method.
assert self.quant_method is not None
self.quant_method.create_weights(self,
self.input_size, [self.output_size],
self.input_size,
self.output_size,
self.params_dtype,
weight_loader=self.weight_loader)
self.quant_method.create_weights(
self,
self.input_size,
[self.output_size],
self.input_size,
self.output_size,
self.params_dtype,
weight_loader=self.weight_loader,
)
if bias:
self.bias = Parameter(
torch.empty(self.output_size, dtype=self.params_dtype))
set_weight_attrs(self.bias, {
"output_dim": 0,
"weight_loader": self.weight_loader,
})
self.bias = Parameter(torch.empty(self.output_size, dtype=self.params_dtype))
set_weight_attrs(
self.bias,
{
"output_dim": 0,
"weight_loader": self.weight_loader,
},
)
else:
self.register_parameter("bias", None)
@@ -463,7 +482,7 @@ class AscendReplicatedLinear(ReplicatedLinear):
def forward(
self,
input_,
) -> Union[torch.Tensor, tuple[torch.Tensor, Optional[Parameter]]]:
) -> torch.Tensor | tuple[torch.Tensor, Parameter | None]:
if self.custom_op is not None:
return self.custom_op.apply(input_)

310
vllm_ascend/ops/linear_op.py
View File

@@ -31,16 +31,18 @@ CustomLinearOp
└── CustomReplicatedOp
How to extend a new linear op? Taking column parallel op as an example:
1. Inherit from CustomColumnParallelOp and create a new class MyColumnParallelOp
2. [Optional] The default communication group is the TP group. If a custom communication group is needed, override the comm_group method
2. [Optional] The default communication group is the TP group. If a custom communication group is needed,
override the comm_group method
3. Override the apply method according to requirements, which will replace the original linear.forward
4. Add selection logic for MyColumnParallelOp in the get_column_parallel_op method, typically based on prefix and configuration judgments
Row parallel op follows a similar approach - inherit from RowColumnParallelOp and register the new class in get_row_parallel_op.
4. Add selection logic for MyColumnParallelOp in the get_column_parallel_op method, typically based on
prefix and configuration judgments
Row parallel op follows a similar approach - inherit from RowColumnParallelOp and register the new class in
get_row_parallel_op.
"""
import re
from functools import lru_cache
from types import SimpleNamespace
from typing import Optional, Union
import torch
import torch.distributed as dist
@@ -49,27 +51,37 @@ import torch_npu
from torch import nn
from torch.distributed import ProcessGroup
from torch.nn.parameter import Parameter
from vllm.distributed import (split_tensor_along_last_dim,
tensor_model_parallel_all_reduce,
tensor_model_parallel_reduce_scatter)
from vllm.distributed import (
split_tensor_along_last_dim,
tensor_model_parallel_all_reduce,
tensor_model_parallel_reduce_scatter,
)
from vllm.distributed.parallel_state import get_tp_group
from vllm.forward_context import get_forward_context
from vllm_ascend import envs as envs_ascend
from vllm_ascend.ascend_config import get_ascend_config
from vllm_ascend.distributed.parallel_state import (get_flashcomm2_odp_group,
get_flashcomm2_otp_group,
get_mlp_tp_group,
get_otp_group)
from vllm_ascend.distributed.parallel_state import (
get_flashcomm2_odp_group,
get_flashcomm2_otp_group,
get_mlp_tp_group,
get_otp_group,
)
from vllm_ascend.ops.flashcomm2_oshard_manager import flashcomm2_oshard_manager
from vllm_ascend.utils import (enable_dsa_cp, enable_dsa_cp_with_layer_shard, enable_sp, flashcomm2_enable,
get_flashcomm2_reorgnized_batch_ids,
matmul_allreduce_enable, mlp_tp_enable,
oproj_tp_enable, shared_expert_dp_enabled,
get_weight_prefetch_method)
from vllm_ascend.utils import (
enable_dsa_cp,
enable_dsa_cp_with_layer_shard,
enable_sp,
flashcomm2_enable,
get_flashcomm2_reorgnized_batch_ids,
get_weight_prefetch_method,
matmul_allreduce_enable,
mlp_tp_enable,
oproj_tp_enable,
shared_expert_dp_enabled,
)
class CustomLinearOp:
def __init__(self, layer):
self.layer = layer
self.bias = None
@@ -112,7 +124,6 @@ class CustomLinearOp:
class CustomColumnParallelOp(CustomLinearOp):
def __init__(self, layer):
super().__init__(layer)
self.gather_output = None
@@ -123,7 +134,6 @@ class CustomColumnParallelOp(CustomLinearOp):
class CustomRowParallelOp(CustomLinearOp):
def __init__(self, layer):
super().__init__(layer)
self.reduce_results = None
@@ -140,7 +150,9 @@ class CustomRowParallelOp(CustomLinearOp):
output, output_bias = self.apply_impl(input_)
weight_prefetch_method = get_weight_prefetch_method()
if weight_prefetch_method:
weight_prefetch_method.maybe_prefetch_mlp_weight_preprocess(weight_prefetch_method.MLP_GATE_UP, output, self.prefix)
weight_prefetch_method.maybe_prefetch_mlp_weight_preprocess(
weight_prefetch_method.MLP_GATE_UP, output, self.prefix
)
if not self.return_bias:
return output
@@ -148,7 +160,6 @@ class CustomRowParallelOp(CustomLinearOp):
class CustomReplicatedOp(CustomLinearOp):
def apply_impl(self, input_):
bias = self.bias if not self.skip_bias_add else None
assert self.quant_method is not None
@@ -160,7 +171,6 @@ class CustomReplicatedOp(CustomLinearOp):
class MLPColumnParallelOp(CustomColumnParallelOp):
def __init__(self, layer):
super().__init__(layer)
@@ -171,7 +181,7 @@ class MLPColumnParallelOp(CustomColumnParallelOp):
def apply_impl(
self,
input_: torch.Tensor,
) -> Union[torch.Tensor, tuple[torch.Tensor, Optional[Parameter]]]:
) -> torch.Tensor | tuple[torch.Tensor, Parameter | None]:
bias = self.bias if not self.skip_bias_add else None
# Matrix multiply.
assert self.quant_method is not None
@@ -183,7 +193,6 @@ class MLPColumnParallelOp(CustomColumnParallelOp):
class MLPRowParallelOp(CustomRowParallelOp):
def __init__(self, layer):
super().__init__(layer)
@@ -191,22 +200,16 @@ class MLPRowParallelOp(CustomRowParallelOp):
def comm_group(self):
return get_mlp_tp_group()
def apply_impl(
self, input_: torch.Tensor
) -> Union[torch.Tensor, tuple[torch.Tensor, Optional[Parameter]]]:
def apply_impl(self, input_: torch.Tensor) -> torch.Tensor | tuple[torch.Tensor, Parameter | None]:
if self.input_is_parallel:
input_parallel = input_
else:
splitted_input = split_tensor_along_last_dim(
input_, num_partitions=self.tp_size)
splitted_input = split_tensor_along_last_dim(input_, num_partitions=self.tp_size)
input_parallel = splitted_input[self.tp_rank].contiguous()
assert self.quant_method is not None
bias_ = None if (self.tp_rank > 0
or self.skip_bias_add) else self.layer.bias
output_parallel = self.quant_method.apply(self.layer,
input_parallel,
bias=bias_)
bias_ = None if (self.tp_rank > 0 or self.skip_bias_add) else self.layer.bias
output_parallel = self.quant_method.apply(self.layer, input_parallel, bias=bias_)
output = self.comm_group.reduce_scatter(output_parallel, 0)
output_bias = self.bias if self.skip_bias_add else None
@@ -214,7 +217,6 @@ class MLPRowParallelOp(CustomRowParallelOp):
class OProjRowParallelOp(CustomRowParallelOp):
def __init__(self, layer):
super().__init__(layer)
@@ -225,13 +227,11 @@ class OProjRowParallelOp(CustomRowParallelOp):
def apply_impl(
self,
input_: torch.Tensor,
) -> Union[torch.Tensor, tuple[torch.Tensor, Optional[Parameter]]]:
) -> torch.Tensor | tuple[torch.Tensor, Parameter | None]:
if self.input_is_parallel:
input_parallel = input_
else:
splitted_input = split_tensor_along_last_dim(
input_, num_partitions=self.tp_size)
splitted_input = split_tensor_along_last_dim(input_, num_partitions=self.tp_size)
input_parallel = splitted_input[self.tp_rank].contiguous()
# Prepare tensors for all-to-all communication
@@ -241,27 +241,19 @@ class OProjRowParallelOp(CustomRowParallelOp):
# Reshape tensor for efficient cross-device transfer:
# [batch, dim] -> [tp_size, batch, chunk] -> flattened
send_buf = (input_parallel.reshape(-1,
self.tp_size, chunk_size).transpose(
0, 1).contiguous().view(-1))
send_buf = input_parallel.reshape(-1, self.tp_size, chunk_size).transpose(0, 1).contiguous().view(-1)
# Create receive buffer
recv_buf = torch.empty(total_batch_size * chunk_size,
dtype=input_parallel.dtype,
device=input_parallel.device)
recv_buf = torch.empty(total_batch_size * chunk_size, dtype=input_parallel.dtype, device=input_parallel.device)
# Perform all-to-all communication
dist.all_to_all_single(recv_buf,
send_buf,
group=self.comm_group.device_group)
dist.all_to_all_single(recv_buf, send_buf, group=self.comm_group.device_group)
input_parallel = recv_buf.view(total_batch_size, chunk_size)
# Only fuse bias add for rank 0 to avoid duplicate bias addition in TP>1
bias_ = None if (self.tp_rank > 0 or self.skip_bias_add) else self.bias
assert self.quant_method is not None
output_parallel = self.quant_method.apply(self.layer,
input_parallel,
bias=bias_)
output_parallel = self.quant_method.apply(self.layer, input_parallel, bias=bias_)
# otp-specific: Combine partial results across devices
output = self.comm_group.reduce_scatter(output_parallel, dim=0)
@@ -278,14 +270,12 @@ class OProjRowParallelOp(CustomRowParallelOp):
class Flashcomm2OProjRowParallelOp(CustomRowParallelOp):
def __init__(self, layer):
super().__init__(layer)
self.odp_group = get_flashcomm2_odp_group()
self.odp_size = self.odp_group.world_size
self.otp_size = get_ascend_config().flashcomm2_oproj_tensor_parallel_size
self.reorgnized_batch_ids = get_flashcomm2_reorgnized_batch_ids(
get_tp_group().world_size)
self.reorgnized_batch_ids = get_flashcomm2_reorgnized_batch_ids(get_tp_group().world_size)
self.group_indices = torch.tensor(self.reorgnized_batch_ids).npu()
self.layer._quant_comm_config = {}
@@ -308,32 +298,28 @@ class Flashcomm2OProjRowParallelOp(CustomRowParallelOp):
def apply_impl(
self,
input_: torch.Tensor,
) -> Union[torch.Tensor, tuple[torch.Tensor, Optional[Parameter]]]:
) -> torch.Tensor | tuple[torch.Tensor, Parameter | None]:
"""Linear layer for Flashcomm2.
Input.ahspe = [batchsize*seqlength, headnum*headdim/TP]
Output.shape = [(batchsize*seqlength+padsize)/TP, hiddensize]
Input.ahspe = [batchsize*seqlength, headnum*headdim/TP]
Output.shape = [(batchsize*seqlength+padsize)/TP, hiddensize]
"""
# Handle input parallelism - split or use as-is
if self.input_is_parallel:
input_parallel = input_
else:
tp_rank = self.tp_rank
splitted_input = split_tensor_along_last_dim(
input_, num_partitions=self.tp_size)
splitted_input = split_tensor_along_last_dim(input_, num_partitions=self.tp_size)
input_parallel = splitted_input[tp_rank].contiguous()
# padding for all-to-all
forward_context = get_forward_context()
num_padding_tokens = forward_context.pad_size
if num_padding_tokens > 0:
input_parallel = nn.functional.pad(input_parallel,
(0, 0, 0, num_padding_tokens))
input_parallel = nn.functional.pad(input_parallel, (0, 0, 0, num_padding_tokens))
def otp_maybe_quant_comm(x):
# Reorganize the tensor so that the batch id and rank id correspond to each other.
chunk_num = len(self.reorgnized_batch_ids) * len(
self.reorgnized_batch_ids[0])
chunk_num = len(self.reorgnized_batch_ids) * len(self.reorgnized_batch_ids[0])
batch_size = x.size(0)
assert batch_size % chunk_num == 0, f"Batch_size({batch_size}) must be divisible by chunk_num({chunk_num})"
@@ -352,26 +338,19 @@ class Flashcomm2OProjRowParallelOp(CustomRowParallelOp):
total_intermediate_size = local_intermediate_size * all2all_tp_size
# Create receive buffer
recv_buf = torch.empty(total_intermediate_size * chunk_size,
dtype=x.dtype,
device=x.device)
recv_buf = torch.empty(total_intermediate_size * chunk_size, dtype=x.dtype, device=x.device)
# Perform all-to-all communication
dist.all_to_all_single(recv_buf,
send_buf,
group=self.odp_group.device_group)
dist.all_to_all_single(recv_buf, send_buf, group=self.odp_group.device_group)
return recv_buf.view(all2all_tp_size, chunk_size,
-1).transpose(0, 1).reshape(chunk_size, -1)
return recv_buf.view(all2all_tp_size, chunk_size, -1).transpose(0, 1).reshape(chunk_size, -1)
if not hasattr(self, "_quant_comm_config"):
self.layer._quant_comm_config = {}
self.layer._quant_comm_config[
"communication_fn"] = otp_maybe_quant_comm
actual_quant_method = getattr(self.quant_method, 'quant_method',
self.quant_method)
from vllm_ascend.quantization.methods.w8a8_static import \
AscendW8A8LinearMethod
self.layer._quant_comm_config["communication_fn"] = otp_maybe_quant_comm
actual_quant_method = getattr(self.quant_method, "quant_method", self.quant_method)
from vllm_ascend.quantization.methods.w8a8_static import AscendW8A8LinearMethod
if not isinstance(actual_quant_method, AscendW8A8LinearMethod):
# Check if w8a8 quantization is enabled. If not, communicate immediately.
input_parallel = otp_maybe_quant_comm(input_parallel)
@@ -382,9 +361,7 @@ class Flashcomm2OProjRowParallelOp(CustomRowParallelOp):
# bias will not get added more than once in TP>1 case)
bias_ = None if (self.tp_rank > 0 or self.skip_bias_add) else self.bias
output_parallel = self.quant_method.apply(self.layer,
input_parallel,
bias=bias_)
output_parallel = self.quant_method.apply(self.layer, input_parallel, bias=bias_)
# output_parallel shape: [bs/(TP/flashcomm2_otp_size), hiddenstate]
if self.tp_size > 1:
# flashcomm2 with reduce-scatter
@@ -408,8 +385,7 @@ class Flashcomm2OProjRowParallelOp(CustomRowParallelOp):
self.input_is_parallel = self.layer.input_is_parallel
self.input_size_per_partition = self.layer.input_size_per_partition
if flashcomm2_oshard_manager.flashcomm2_oshard_enable():
flashcomm2_oshard_manager.register_layer(self.layer,
prefetch_step=1)
flashcomm2_oshard_manager.register_layer(self.layer, prefetch_step=1)
class MatmulAllreduceRowParallelOp(CustomRowParallelOp):
@@ -419,28 +395,22 @@ class MatmulAllreduceRowParallelOp(CustomRowParallelOp):
super().__init__(layer)
self.hcomm_info = self.get_hcomm_info(self.comm_group.device_group)
def apply_impl(
self, input_: torch.Tensor
) -> Union[torch.Tensor, tuple[torch.Tensor, Optional[Parameter]]]:
def apply_impl(self, input_: torch.Tensor) -> torch.Tensor | tuple[torch.Tensor, Parameter | None]:
if self.input_is_parallel:
input_parallel = input_
else:
splitted_input = split_tensor_along_last_dim(
input_, num_partitions=self.tp_size)
splitted_input = split_tensor_along_last_dim(input_, num_partitions=self.tp_size)
input_parallel = splitted_input[self.tp_rank].contiguous()
"""Calculate the output tensor of forward by considering
fusing communication and computation."""
bias_ = None if (self.tp_rank > 0 or self.skip_bias_add) else self.bias
if self.reduce_results and self.tp_size > 1:
output = torch_npu.npu_mm_all_reduce_base(input_parallel,
self.layer.weight.t(),
self.hcomm_info,
bias=bias_)
output = torch_npu.npu_mm_all_reduce_base(
input_parallel, self.layer.weight.t(), self.hcomm_info, bias=bias_
)
else:
assert self.quant_method is not None
output = self.quant_method.apply(self.layer,
input_parallel,
bias=bias_)
output = self.quant_method.apply(self.layer, input_parallel, bias=bias_)
output_bias = self.bias if self.skip_bias_add else None
return output, output_bias
@@ -454,18 +424,14 @@ class MatmulAllreduceRowParallelOp(CustomRowParallelOp):
rank = torch.distributed.get_rank(group)
if torch.__version__ > "2.0":
global_rank = torch.distributed.get_global_rank(group, rank)
cls._HCOMM_INFO = group._get_backend(
torch.device("npu")).get_hccl_comm_name(global_rank)
cls._HCOMM_INFO = group._get_backend(torch.device("npu")).get_hccl_comm_name(global_rank)
else:
cls._HCOMM_INFO = group.get_hccl_comm_name(rank)
return cls._HCOMM_INFO
class SequenceColumnParallelOp(CustomColumnParallelOp):
def apply_impl(
self, input_: torch.Tensor
) -> Union[torch.Tensor, tuple[torch.Tensor, Optional[Parameter]]]:
def apply_impl(self, input_: torch.Tensor) -> torch.Tensor | tuple[torch.Tensor, Parameter | None]:
"""Linear layer with column parallelism.
Implemented multiple optimization projects for dense models, such as FlashComm and
@@ -490,13 +456,10 @@ class SequenceColumnParallelOp(CustomColumnParallelOp):
class Flashcomm2OshardQKVParallelOp(CustomColumnParallelOp):
def __init__(self, layer):
super().__init__(layer)
def apply_impl(
self, input_: torch.Tensor
) -> Union[torch.Tensor, tuple[torch.Tensor, Optional[Parameter]]]:
def apply_impl(self, input_: torch.Tensor) -> torch.Tensor | tuple[torch.Tensor, Parameter | None]:
"""Column-parallel linear with FlashComm2 OShard optimization."""
bias = self.bias if not self.skip_bias_add else None
@@ -505,12 +468,10 @@ class Flashcomm2OshardQKVParallelOp(CustomColumnParallelOp):
assert self.quant_method is not None
if enable_sp():
input_ = torch.ops.vllm.maybe_all_gather_and_maybe_unpad(
input_, True)
input_ = torch.ops.vllm.maybe_all_gather_and_maybe_unpad(input_, True)
# Trigger async broadcast before matmul to overlap communication.
flashcomm2_oshard_manager.trigger_broadcast_for_layer(
self.layer.prefix)
flashcomm2_oshard_manager.trigger_broadcast_for_layer(self.layer.prefix)
output_parallel = self.quant_method.apply(self.layer, input_, bias)
if self.gather_output and self.tp_size > 1:
@@ -523,14 +484,11 @@ class Flashcomm2OshardQKVParallelOp(CustomColumnParallelOp):
class SequenceRowParallelOp(CustomRowParallelOp):
def __init__(self, layer):
super().__init__(layer)
self.unique_prefix = None
def apply_impl(
self, input_: torch.Tensor
) -> Union[torch.Tensor, tuple[torch.Tensor, Optional[Parameter]]]:
def apply_impl(self, input_: torch.Tensor) -> torch.Tensor | tuple[torch.Tensor, Parameter | None]:
"""Linear layer with column parallelism.
Implemented multiple optimization projects for dense models, such as FlashComm and
@@ -540,26 +498,21 @@ class SequenceRowParallelOp(CustomRowParallelOp):
if self.input_is_parallel:
input_parallel = input_
else:
splitted_input = split_tensor_along_last_dim(
input_, num_partitions=self.tp_size)
splitted_input = split_tensor_along_last_dim(input_, num_partitions=self.tp_size)
input_parallel = splitted_input[self.tp_rank].contiguous()
assert self.quant_method is not None
bias_ = None if (self.tp_rank > 0 or self.skip_bias_add) else self.bias
if self.tp_size == 1 or not self.reduce_results:
output = self.quant_method.apply(self.layer,
input_parallel,
bias=bias_)
output = self.quant_method.apply(self.layer, input_parallel, bias=bias_)
else:
output = torch.ops.vllm.matmul_and_reduce(input_parallel,
self.unique_prefix)
output = torch.ops.vllm.matmul_and_reduce(input_parallel, self.unique_prefix)
output_bias = self.bias if self.skip_bias_add else None
return output, output_bias
def matmul_and_reduce(self, input_parallel: torch.Tensor,
bias_: Optional[Parameter]) -> torch.Tensor:
def matmul_and_reduce(self, input_parallel: torch.Tensor, bias_: Parameter | None) -> torch.Tensor:
assert self.quant_method is not None
try:
forward_context = get_forward_context()
@@ -572,29 +525,24 @@ class SequenceRowParallelOp(CustomRowParallelOp):
x = input_parallel
if not sp_enabled:
output_parallel = self.layer.quant_method.apply(self.layer,
x,
bias=bias_)
output_parallel = self.layer.quant_method.apply(self.layer, x, bias=bias_)
return tensor_model_parallel_all_reduce(output_parallel)
pad_size = forward_context.pad_size
if pad_size > 0 and not (enable_dsa_cp()
and "o_proj" in self.layer.prefix):
if pad_size > 0 and not (enable_dsa_cp() and "o_proj" in self.layer.prefix):
x = F.pad(x, (0, 0, 0, pad_size))
world_size = self.layer.tp_size
comm_mode = "aiv"
hcom_name = get_tp_group().device_group._get_backend(
torch.device('npu')).get_hccl_comm_name(self.layer.tp_rank)
hcom_name = get_tp_group().device_group._get_backend(torch.device("npu")).get_hccl_comm_name(self.layer.tp_rank)
from vllm.model_executor.layers.linear import UnquantizedLinearMethod
from vllm_ascend.quantization.methods import AscendW8A8LinearMethod
from vllm_ascend.quantization.method_adapters import AscendLinearMethod
from vllm_ascend.quantization.methods import AscendW8A8LinearMethod
# For unquant
if mmrs_fusion and isinstance(self.layer.quant_method,
UnquantizedLinearMethod):
if mmrs_fusion and isinstance(self.layer.quant_method, UnquantizedLinearMethod):
output = torch_npu.npu_mm_reduce_scatter_base(
x,
self.layer.weight.t(),
@@ -603,19 +551,22 @@ class SequenceRowParallelOp(CustomRowParallelOp):
reduce_op="sum",
bias=None,
comm_turn=0,
comm_mode=comm_mode)
comm_mode=comm_mode,
)
if bias_ is not None:
output.add_(bias_)
# For w8a8 quant
elif mmrs_fusion and (
isinstance(self.layer.quant_method, AscendLinearMethod)
and isinstance(self.layer.quant_method.quant_method,
AscendW8A8LinearMethod)):
isinstance(self.layer.quant_method, AscendLinearMethod)
and isinstance(self.layer.quant_method.quant_method, AscendW8A8LinearMethod)
):
if x.dtype != torch.int8:
x_quant = torch.ops.vllm.quantize(
x, self.layer.aclnn_input_scale,
x,
self.layer.aclnn_input_scale,
self.layer.aclnn_input_scale_reciprocal,
self.layer.aclnn_input_offset)
self.layer.aclnn_input_offset,
)
else:
x_quant = x
quant_bias = self.layer.quant_bias
@@ -631,14 +582,11 @@ class SequenceRowParallelOp(CustomRowParallelOp):
comm_turn=0,
x2_scale=deq_scale,
output_dtype=output_dtype,
comm_mode=comm_mode)
output = torch.add(
output,
torch.mul(quant_bias, deq_scale).to(self.layer.params_dtype))
comm_mode=comm_mode,
)
output = torch.add(output, torch.mul(quant_bias, deq_scale).to(self.layer.params_dtype))
else:
output_parallel = self.layer.quant_method.apply(self.layer,
x,
bias=bias_)
output_parallel = self.layer.quant_method.apply(self.layer, x, bias=bias_)
output = tensor_model_parallel_reduce_scatter(output_parallel, 0)
return output
@@ -651,13 +599,10 @@ class SequenceRowParallelOp(CustomRowParallelOp):
class ShardedCPRowParallelOp(CustomRowParallelOp):
@property
def comm_group(self):
# fake comm group to bypass tp logic
return SimpleNamespace(world_size=1,
rank_in_group=0,
device_group=None)
return SimpleNamespace(world_size=1, rank_in_group=0, device_group=None)
def apply_impl(
self,
@@ -677,13 +622,10 @@ class ShardedCPRowParallelOp(CustomRowParallelOp):
class ShardedCPColumnParallelOp(CustomColumnParallelOp):
@property
def comm_group(self):
# fake comm group to bypass tp logic
return SimpleNamespace(world_size=1,
rank_in_group=0,
device_group=None)
return SimpleNamespace(world_size=1, rank_in_group=0, device_group=None)
def apply_impl(
self,
@@ -700,12 +642,10 @@ class ShardedCPColumnParallelOp(CustomColumnParallelOp):
def _get_column_parallel_op(
prefix, layer
) -> Optional[Union[MLPColumnParallelOp, SequenceColumnParallelOp,
ShardedCPColumnParallelOp, Flashcomm2OshardQKVParallelOp]]:
) -> MLPColumnParallelOp | SequenceColumnParallelOp | ShardedCPColumnParallelOp | Flashcomm2OshardQKVParallelOp | None:
if enable_dsa_cp() and ("q_b_proj" in prefix or "kv_b_proj" in prefix):
return ShardedCPColumnParallelOp(layer)
if "gate_up_proj" in prefix and mlp_tp_enable(
) and not is_moe_layer(prefix):
if "gate_up_proj" in prefix and mlp_tp_enable() and not is_moe_layer(prefix):
return MLPColumnParallelOp(layer)
if flashcomm2_oshard_manager.flashcomm2_oshard_enable():
if any(p in prefix for p in ("qkv_proj", "conv1d", "query_key_value")):
@@ -714,7 +654,7 @@ def _get_column_parallel_op(
if "shared_expert" in prefix:
return None
sp_column_prefix = [
"gate_up_proj", # first MLP of most LLMs
"gate_up_proj", # first MLP of most LLMs
"in_proj", # gated deltanet of Qwen3 Next
"qkv_proj", # qkv linear of most LLMs
"conv1d", # gated deltanet of Qwen3 Next
@@ -729,9 +669,15 @@ def _get_column_parallel_op(
def _get_row_parallel_op(
prefix, layer
) -> Optional[Union[MLPRowParallelOp, OProjRowParallelOp,
Flashcomm2OProjRowParallelOp, MatmulAllreduceRowParallelOp,
SequenceRowParallelOp, ShardedCPRowParallelOp]]:
) -> (
MLPRowParallelOp
| OProjRowParallelOp
| Flashcomm2OProjRowParallelOp
| MatmulAllreduceRowParallelOp
| SequenceRowParallelOp
| ShardedCPRowParallelOp
| None
):
if enable_dsa_cp_with_layer_shard() and "o_proj" in prefix:
return ShardedCPRowParallelOp(layer)
if "down_proj" in prefix and mlp_tp_enable() and not is_moe_layer(prefix):
@@ -760,16 +706,21 @@ def _get_row_parallel_op(
def get_parallel_op(disable_tp, prefix, layer, direct):
if disable_tp or ("shared_experts" in prefix
and shared_expert_dp_enabled()):
if disable_tp or ("shared_experts" in prefix and shared_expert_dp_enabled()):
return None, 0, 1
custom_op: Optional[Union[MLPColumnParallelOp, SequenceColumnParallelOp,
MLPRowParallelOp, OProjRowParallelOp,
Flashcomm2OProjRowParallelOp,
Flashcomm2OshardQKVParallelOp,
MatmulAllreduceRowParallelOp,
SequenceRowParallelOp, ShardedCPRowParallelOp,
ShardedCPColumnParallelOp]] = None
custom_op: (
MLPColumnParallelOp
| SequenceColumnParallelOp
| MLPRowParallelOp
| OProjRowParallelOp
| Flashcomm2OProjRowParallelOp
| Flashcomm2OshardQKVParallelOp
| MatmulAllreduceRowParallelOp
| SequenceRowParallelOp
| ShardedCPRowParallelOp
| ShardedCPColumnParallelOp
| None
) = None
if direct == "row":
custom_op = _get_row_parallel_op(prefix, layer)
@@ -782,8 +733,7 @@ def get_parallel_op(disable_tp, prefix, layer, direct):
return None, get_tp_group().rank_in_group, get_tp_group().world_size
def get_replicated_op(disable_tp, prefix,
layer) -> Optional[Union[CustomReplicatedOp]]:
def get_replicated_op(disable_tp, prefix, layer) -> CustomReplicatedOp | None:
if disable_tp:
return None
@@ -791,24 +741,22 @@ def get_replicated_op(disable_tp, prefix,
def is_moe_layer(prefix: str) -> bool:
@lru_cache(maxsize=1)
def get_moe_params():
from vllm.config import get_current_vllm_config
vllm_config = get_current_vllm_config()
config = vllm_config.model_config.hf_text_config
n_routed_experts = getattr(config, 'n_routed_experts', 0)
first_k_dense_replace = getattr(config, 'first_k_dense_replace',
float('inf'))
moe_layer_freq = getattr(config, 'moe_layer_freq', 1)
n_routed_experts = getattr(config, "n_routed_experts", 0)
first_k_dense_replace = getattr(config, "first_k_dense_replace", float("inf"))
moe_layer_freq = getattr(config, "moe_layer_freq", 1)
return n_routed_experts, first_k_dense_replace, moe_layer_freq
match = re.search(r'layers\.(\d+)\.', prefix)
match = re.search(r"layers\.(\d+)\.", prefix)
if match is None:
return False
layer_idx = int(match.group(1))
n_routed_experts, first_k_dense_replace, moe_layer_freq = get_moe_params()
return (n_routed_experts is not None and layer_idx >= first_k_dense_replace
and layer_idx % moe_layer_freq == 0)
return n_routed_experts is not None and layer_idx >= first_k_dense_replace and layer_idx % moe_layer_freq == 0

302
vllm_ascend/ops/rotary_embedding.py
View File

@@ -17,13 +17,15 @@
import math
import os
from typing import Optional, Tuple
import torch
import torch_npu
from vllm.model_executor.layers.rotary_embedding import (
DeepseekScalingRotaryEmbedding, MRotaryEmbedding, RotaryEmbedding,
YaRNScalingRotaryEmbedding)
DeepseekScalingRotaryEmbedding,
MRotaryEmbedding,
RotaryEmbedding,
YaRNScalingRotaryEmbedding,
)
from vllm.model_executor.layers.rotary_embedding.common import ApplyRotaryEmb
from vllm.triton_utils import HAS_TRITON
@@ -31,8 +33,7 @@ if HAS_TRITON:
from vllm.model_executor.layers.rotary_embedding.mrope import triton_mrope
from vllm_ascend.platform import NPUPlatform
from vllm_ascend.utils import (AscendDeviceType, enable_custom_op,
get_ascend_device_type, has_rope, is_vl_model)
from vllm_ascend.utils import AscendDeviceType, enable_custom_op, get_ascend_device_type, has_rope, is_vl_model
# Currently, rope ops used on npu requires detached cos && sin as inputs.
# However, RotaryEmbedding in vllm use cos_sin_cache as a whole variable.
@@ -54,17 +55,13 @@ _cos_slice: torch.Tensor = None
_sin_slice: torch.Tensor = None
def set_cos_and_sin(vllm_config, max_num_reqs, decode_token_per_req, dtype,
device):
def set_cos_and_sin(vllm_config, max_num_reqs, decode_token_per_req, dtype, device):
global _cos_mla
global _sin_mla
global _cos
global _sin
if _cos_mla is not None or \
_sin_mla is not None or \
_cos is not None or \
_sin is not None:
if _cos_mla is not None or _sin_mla is not None or _cos is not None or _sin is not None:
return
model_config = vllm_config.model_config
@@ -72,36 +69,15 @@ def set_cos_and_sin(vllm_config, max_num_reqs, decode_token_per_req, dtype,
if model_config.use_mla:
rope_dim = model_config.hf_text_config.qk_rope_head_dim
_cos_mla = torch.ones(max_num_batched_tokens,
1,
1,
rope_dim,
dtype=dtype,
device=device)
_sin_mla = torch.zeros(max_num_batched_tokens,
1,
1,
rope_dim,
dtype=dtype,
device=device)
_cos_mla = torch.ones(max_num_batched_tokens, 1, 1, rope_dim, dtype=dtype, device=device)
_sin_mla = torch.zeros(max_num_batched_tokens, 1, 1, rope_dim, dtype=dtype, device=device)
elif not is_vl_model(vllm_config) and has_rope(vllm_config):
rope_dim = model_config.get_head_size()
# For models using partial rope like Qwen3-Next.
if hasattr(model_config.hf_text_config, "partial_rotary_factor"):
rope_dim = int(rope_dim *
model_config.hf_text_config.partial_rotary_factor)
_cos = torch.ones(1,
max_num_batched_tokens,
1,
rope_dim,
dtype=dtype,
device=device)
_sin = torch.zeros(1,
max_num_batched_tokens,
1,
rope_dim,
dtype=dtype,
device=device)
rope_dim = int(rope_dim * model_config.hf_text_config.partial_rotary_factor)
_cos = torch.ones(1, max_num_batched_tokens, 1, rope_dim, dtype=dtype, device=device)
_sin = torch.zeros(1, max_num_batched_tokens, 1, rope_dim, dtype=dtype, device=device)
def get_cos_and_sin_mla(positions, use_cache=False):
@@ -139,8 +115,7 @@ def _record_cos_and_sin_cache_interleaved(cos_sin_cache):
if _cos_cache is not None or _sin_cache is not None:
return
hidden_dim = cos_sin_cache.shape[-1] // 2
cos_cache, sin_cache = cos_sin_cache.view(-1, 2, hidden_dim).repeat(
1, 1, 2).chunk(2, dim=1)
cos_cache, sin_cache = cos_sin_cache.view(-1, 2, hidden_dim).repeat(1, 1, 2).chunk(2, dim=1)
_cos_cache = cos_cache.squeeze(1)
_sin_cache = sin_cache.squeeze(1)
@@ -151,16 +126,16 @@ def update_cos_sin(positions):
global _cos_slice
global _sin_slice
if _cos_sin_cache is None or \
_cos is None or \
_sin is None:
if _cos_sin_cache is None or _cos is None or _sin is None:
return
num_tokens = positions.size(0)
_cos[:, :num_tokens] = _cos_sin_cache.index_select(0, positions).view(
num_tokens, 2, -1).repeat(1, 1, 2).chunk(2, dim=-2)[0]
_sin[:, :num_tokens] = _cos_sin_cache.index_select(0, positions).view(
num_tokens, 2, -1).repeat(1, 1, 2).chunk(2, dim=-2)[1]
_cos[:, :num_tokens] = (
_cos_sin_cache.index_select(0, positions).view(num_tokens, 2, -1).repeat(1, 1, 2).chunk(2, dim=-2)[0]
)
_sin[:, :num_tokens] = (
_cos_sin_cache.index_select(0, positions).view(num_tokens, 2, -1).repeat(1, 1, 2).chunk(2, dim=-2)[1]
)
_cos_slice = _cos[:, :num_tokens]
_sin_slice = _sin[:, :num_tokens]
@@ -170,8 +145,7 @@ def get_cos_and_sin_slice():
def _custom_rotary_embedding_enabled(query, neox_style, head_size):
return query.dtype == torch.float16 and neox_style and head_size % 32 == 0 and enable_custom_op(
)
return query.dtype == torch.float16 and neox_style and head_size % 32 == 0 and enable_custom_op()
def _rope_forward_oot(
@@ -180,8 +154,8 @@ def _rope_forward_oot(
query: torch.Tensor,
key: torch.Tensor,
is_neox_style: bool,
offsets: Optional[torch.Tensor] = None
) -> Tuple[torch.Tensor, torch.Tensor]:
offsets: torch.Tensor | None = None,
) -> tuple[torch.Tensor, torch.Tensor]:
query_shape, key_shape = query.shape, key.shape
if self.cos_sin_cache.device != query.device:
self.cos_sin_cache = self.cos_sin_cache.to(query.device)
@@ -189,8 +163,7 @@ def _rope_forward_oot(
self.cos_sin_cache = self.cos_sin_cache.to(query.dtype)
cos, sin = get_cos_and_sin_slice()
# adopt custom kernel path for rotary_embedding
if _custom_rotary_embedding_enabled(
query, is_neox_style, self.head_size):
if _custom_rotary_embedding_enabled(query, is_neox_style, self.head_size):
query, key = torch.ops._C_ascend.rotary_embedding(
positions,
query,
@@ -201,43 +174,40 @@ def _rope_forward_oot(
)
return query.view(query_shape), key.view(key_shape)
if offsets is not None:
raise NotImplementedError(
"Batched rotary embedding is currently not supported on NPU.")
raise NotImplementedError("Batched rotary embedding is currently not supported on NPU.")
else:
if is_neox_style and self.head_size == 128 and self.cos_sin_cache.shape[
-1] == 128 and cos is not None and sin is not None:
if (
is_neox_style
and self.head_size == 128
and self.cos_sin_cache.shape[-1] == 128
and cos is not None
and sin is not None
):
# If cos and sin are generated outside, use npu_apply_rotary_pos_emb to avoid redundant calculation.
# This method requires head_size and rotary_dim equal 128 and neox_style is True
query = query.contiguous().view(1, query.shape[0], -1,
self.head_size)
query = query.contiguous().view(1, query.shape[0], -1, self.head_size)
key = key.contiguous().view(1, key.shape[0], -1, self.head_size)
# Although this function modifies in-place, please retain the function's return value.
# Otherwise, the graph fusion operation may fail.
query, key = torch_npu.npu_apply_rotary_pos_emb(
query, key, cos, sin)
query, key = torch_npu.npu_apply_rotary_pos_emb(query, key, cos, sin)
elif self.rotary_dim < self.head_size:
if HAS_TRITON:
if HAS_TRITON:
cos = cos.view(-1, self.rotary_dim)
sin = sin.view(-1, self.rotary_dim)
q = query.contiguous().view(query.shape[0], -1,
self.head_size)
q = query.contiguous().view(query.shape[0], -1, self.head_size)
k = key.contiguous().view(key.shape[0], -1, self.head_size)
query, key = torch.ops.vllm.rope_forward_triton(q,
k,
cos,
sin,
rope_dim=self.rotary_dim,
is_neox_style=True)
query, key = torch.ops.vllm.rope_forward_triton(
q, k, cos, sin, rope_dim=self.rotary_dim, is_neox_style=True
)
return query.view(query_shape), key.view(key_shape)
else:
num_tokens = query.shape[0]
query = query.view(num_tokens, -1, self.head_size)
key = key.view(num_tokens, -1, self.head_size)
q_rot = query[..., :self.rotary_dim]
q_pass = query[..., self.rotary_dim:]
k_rot = key[..., :self.rotary_dim]
k_pass = key[..., self.rotary_dim:]
q_rot = query[..., : self.rotary_dim]
q_pass = query[..., self.rotary_dim :]
k_rot = key[..., : self.rotary_dim]
k_pass = key[..., self.rotary_dim :]
q_rot = q_rot.contiguous().view(num_tokens, -1)
k_rot = k_rot.contiguous().view(num_tokens, -1)
# only the rotary part is processed here,
@@ -271,7 +241,6 @@ def _rope_forward_oot(
class AscendRotaryEmbedding(RotaryEmbedding):
def __init__(
self,
head_size: int,
@@ -281,8 +250,7 @@ class AscendRotaryEmbedding(RotaryEmbedding):
is_neox_style: bool,
dtype: torch.dtype,
) -> None:
super().__init__(head_size, rotary_dim, max_position_embeddings, base,
is_neox_style, dtype)
super().__init__(head_size, rotary_dim, max_position_embeddings, base, is_neox_style, dtype)
_record_cos_sin_cache(self.cos_sin_cache)
_record_cos_and_sin_cache_interleaved(self.cos_sin_cache)
@@ -291,18 +259,16 @@ class AscendRotaryEmbedding(RotaryEmbedding):
positions: torch.Tensor,
query: torch.Tensor,
key: torch.Tensor,
offsets: Optional[torch.Tensor] = None,
is_neox_style_override: Optional[bool] = None,
offsets: torch.Tensor | None = None,
is_neox_style_override: bool | None = None,
):
is_neox_style = self.is_neox_style
if is_neox_style_override is not None:
is_neox_style = is_neox_style_override
return _rope_forward_oot(self, positions, query, key, is_neox_style,
offsets)
return _rope_forward_oot(self, positions, query, key, is_neox_style, offsets)
class AscendYaRNRotaryEmbedding(YaRNScalingRotaryEmbedding):
def __init__(
self,
head_size: int,
@@ -322,10 +288,11 @@ class AscendYaRNRotaryEmbedding(YaRNScalingRotaryEmbedding):
"extrapolation_factor": extrapolation_factor,
"attn_factor": attn_factor,
"beta_fast": beta_fast,
"beta_slow": beta_slow
"beta_slow": beta_slow,
}
super().__init__(head_size, rotary_dim, max_position_embeddings, base,
is_neox_style, scaling_factor, dtype, **extra_kwargs)
super().__init__(
head_size, rotary_dim, max_position_embeddings, base, is_neox_style, scaling_factor, dtype, **extra_kwargs
)
_record_cos_sin_cache(self.cos_sin_cache)
def forward_oot(
@@ -333,16 +300,13 @@ class AscendYaRNRotaryEmbedding(YaRNScalingRotaryEmbedding):
positions: torch.Tensor,
query: torch.Tensor,
key: torch.Tensor,
offsets: Optional[torch.Tensor] = None,
is_neox_style_override: Optional[bool] = None,
offsets: torch.Tensor | None = None,
is_neox_style_override: bool | None = None,
):
return AscendRotaryEmbedding.forward_oot(self, positions, query, key,
offsets,
is_neox_style_override)
return AscendRotaryEmbedding.forward_oot(self, positions, query, key, offsets, is_neox_style_override)
class AscendDeepseekScalingRotaryEmbedding(DeepseekScalingRotaryEmbedding):
def __init__(
self,
head_size: int,
@@ -370,18 +334,17 @@ class AscendDeepseekScalingRotaryEmbedding(DeepseekScalingRotaryEmbedding):
self.beta_slow = beta_slow
# Get n-d magnitude scaling corrected for interpolation.
self.mscale = float(
self._yarn_get_mscale(self.scaling_factor, float(mscale)) /
self._yarn_get_mscale(self.scaling_factor, float(mscale_all_dim)) *
attn_factor)
super(DeepseekScalingRotaryEmbedding,
self).__init__(head_size, rotary_dim, max_position_embeddings,
base, is_neox_style, dtype)
self._yarn_get_mscale(self.scaling_factor, float(mscale))
/ self._yarn_get_mscale(self.scaling_factor, float(mscale_all_dim))
* attn_factor
)
super(DeepseekScalingRotaryEmbedding, self).__init__(
head_size, rotary_dim, max_position_embeddings, base, is_neox_style, dtype
)
# NOTE: For ascend friendly computing, reorder sin and cos cache
self.max_seq_len = math.ceil(max_position_embeddings * scaling_factor)
self._set_cos_sin_cache(self.max_seq_len,
device=NPUPlatform.device_type,
dtype=dtype)
self._set_cos_sin_cache(self.max_seq_len, device=NPUPlatform.device_type, dtype=dtype)
def _yarn_get_mscale(self, scale: float = 1, mscale: float = 1) -> float:
if scale <= 1:
@@ -390,56 +353,35 @@ class AscendDeepseekScalingRotaryEmbedding(DeepseekScalingRotaryEmbedding):
def _rotate_half(self, x):
"""Rotates half the hidden dims of the input."""
x1 = x[..., :x.shape[-1] // 2]
x2 = x[..., x.shape[-1] // 2:]
x1 = x[..., : x.shape[-1] // 2]
x2 = x[..., x.shape[-1] // 2 :]
return torch.cat((-x2, x1), dim=-1)
def _yarn_linear_ramp_mask(self, min_value, max_value, dim):
# Note: The if conditional branch is not used here
# to solve MTP compilation error.
max_value += (min_value == max_value).float() * 0.001
linear_func = (torch.arange(dim, dtype=torch.float32) -
min_value) / (max_value - min_value)
linear_func = (torch.arange(dim, dtype=torch.float32) - min_value) / (max_value - min_value)
ramp_func = torch.clamp(linear_func, 0, 1)
return ramp_func
# Inverse dim formula to find dim based on number of rotations
def _yarn_find_correction_dim(self,
num_rotations,
dim,
base=10000,
max_position_embeddings=2048):
def _yarn_find_correction_dim(self, num_rotations, dim, base=10000, max_position_embeddings=2048):
# Note: use torch instead of math to solve MTP compilation error.
return (dim * torch.log(
torch.tensor(max_position_embeddings) /
(num_rotations * 2 * torch.pi))) / (2 *
torch.log(torch.tensor(base)))
return (dim * torch.log(torch.tensor(max_position_embeddings) / (num_rotations * 2 * torch.pi))) / (
2 * torch.log(torch.tensor(base))
)
# Find dim range bounds based on rotations
def _yarn_find_correction_range(self,
low_rot,
high_rot,
dim,
base=10000,
max_position_embeddings=2048):
def _yarn_find_correction_range(self, low_rot, high_rot, dim, base=10000, max_position_embeddings=2048):
# Note: use torch instead of math to solve MTP compilation error.
low = torch.floor(
self._yarn_find_correction_dim(low_rot, dim, base,
max_position_embeddings))
high = torch.ceil(
self._yarn_find_correction_dim(high_rot, dim, base,
max_position_embeddings))
low = torch.floor(self._yarn_find_correction_dim(low_rot, dim, base, max_position_embeddings))
high = torch.ceil(self._yarn_find_correction_dim(high_rot, dim, base, max_position_embeddings))
# Note: use torch instead of max/min to solve MTP compilation error.
return torch.clamp(low, min=0), torch.clamp(high, max=dim - 1)
# Copied from transformers.models.llama.modeling_llama.apply_rotary_pos_emb
def _apply_rotary_pos_emb(self,
q,
k,
cos,
sin,
position_ids,
unsqueeze_dim=1):
def _apply_rotary_pos_emb(self, q, k, cos, sin, position_ids, unsqueeze_dim=1):
"""Applies Rotary Position Embedding to the query and key tensors.
Args:
q (`torch.Tensor`): The query tensor.
@@ -451,11 +393,11 @@ class AscendDeepseekScalingRotaryEmbedding(DeepseekScalingRotaryEmbedding):
used to pass offsetted position ids when working with a KV-cache.
unsqueeze_dim (`int`, *optional*, defaults to 1):
The 'unsqueeze_dim' argument specifies the dimension along which to unsqueeze cos[position_ids] and
sin[position_ids] so that they can be properly broadcasted to the dimensions of q and k. For example, note
that cos[position_ids] and sin[position_ids] have the shape [batch_size, seq_len, head_dim]. Then, if q and
k have the shape [batch_size, heads, seq_len, head_dim], then setting unsqueeze_dim=1 makes
cos[position_ids] and sin[position_ids] broadcastable to the shapes of q and k. Similarly, if q and k have
the shape [batch_size, seq_len, heads, head_dim], then set unsqueeze_dim=2.
sin[position_ids] so that they can be properly broadcasted to the dimensions of q and k. For example,
note that cos[position_ids] and sin[position_ids] have the shape [batch_size, seq_len, head_dim].
Then, if q and k have the shape [batch_size, heads, seq_len, head_dim], then setting unsqueeze_dim=1
makes cos[position_ids] and sin[position_ids] broadcastable to the shapes of q and k. Similarly,
if q and k have the shape [batch_size, seq_len, heads, head_dim], then set unsqueeze_dim=2.
Returns:
`tuple(torch.Tensor)` comprising of the query and key tensors rotated using the Rotary Position Embedding.
"""
@@ -488,10 +430,10 @@ class AscendDeepseekScalingRotaryEmbedding(DeepseekScalingRotaryEmbedding):
def _set_cos_sin_cache(self, max_seq_len, device, dtype):
dim = self.rotary_dim
freq_extra = 1.0 / (self.base**(
torch.arange(0, dim, 2, dtype=torch.float32, device=device) / dim))
freq_inter = 1.0 / (self.scaling_factor * self.base**(
torch.arange(0, dim, 2, dtype=torch.float32, device=device) / dim))
freq_extra = 1.0 / (self.base ** (torch.arange(0, dim, 2, dtype=torch.float32, device=device) / dim))
freq_inter = 1.0 / (
self.scaling_factor * self.base ** (torch.arange(0, dim, 2, dtype=torch.float32, device=device) / dim)
)
low, high = self._yarn_find_correction_range(
self.beta_fast,
@@ -500,10 +442,8 @@ class AscendDeepseekScalingRotaryEmbedding(DeepseekScalingRotaryEmbedding):
self.base,
self.max_position_embeddings,
)
inv_freq_mask = 1.0 - self._yarn_linear_ramp_mask(
low, high, dim // 2).to(device=device, dtype=torch.float32)
inv_freq = freq_inter * (1 -
inv_freq_mask) + freq_extra * inv_freq_mask
inv_freq_mask = 1.0 - self._yarn_linear_ramp_mask(low, high, dim // 2).to(device=device, dtype=torch.float32)
inv_freq = freq_inter * (1 - inv_freq_mask) + freq_extra * inv_freq_mask
self.register_buffer("inv_freq", inv_freq, persistent=False)
t = torch.arange(max_seq_len, device=device, dtype=torch.float32)
@@ -513,20 +453,16 @@ class AscendDeepseekScalingRotaryEmbedding(DeepseekScalingRotaryEmbedding):
sin_cached = torch.cat([freqs, freqs], dim=-1).sin() * self.mscale
cos_cached = cos_cached.to(dtype)
sin_cached = sin_cached.to(dtype)
cache = torch.cat(
[freqs.cos() * self.mscale,
freqs.sin() * self.mscale], dim=-1).to(dtype)
cache = torch.cat([freqs.cos() * self.mscale, freqs.sin() * self.mscale], dim=-1).to(dtype)
self.register_buffer("cos_sin_cache", cache, persistent=False)
self.register_buffer("cos_cached", cos_cached, persistent=False)
self.register_buffer("sin_cached", sin_cached, persistent=False)
_record_cos_sin_cache(cache)
_record_cos_and_sin_cache(cos_cached, sin_cached)
def forward(self,
positions: torch.Tensor,
query: torch.Tensor,
key: torch.Tensor,
offsets: Optional[torch.Tensor] = None):
def forward(
self, positions: torch.Tensor, query: torch.Tensor, key: torch.Tensor, offsets: torch.Tensor | None = None
):
if len(key.shape) == 2:
key = key[:, None, :]
# Note: we implement the non neox_style method with shuffle the last dim and neox style
@@ -535,26 +471,24 @@ class AscendDeepseekScalingRotaryEmbedding(DeepseekScalingRotaryEmbedding):
is_neox_style = True
if self.is_neox_style is False:
b, h_q, d = query.shape
query = query.view(b, h_q, d // 2,
2).transpose(3, 2).reshape(b, h_q, d)
query = query.view(b, h_q, d // 2, 2).transpose(3, 2).reshape(b, h_q, d)
b, h_k, d = key.shape
key = key.view(b, h_k, d // 2, 2).transpose(3,
2).reshape(b, h_k, d)
q_pe, k_pe = _rope_forward_oot(self, positions, query, key,
is_neox_style, offsets)
key = key.view(b, h_k, d // 2, 2).transpose(3, 2).reshape(b, h_k, d)
q_pe, k_pe = _rope_forward_oot(self, positions, query, key, is_neox_style, offsets)
return q_pe, k_pe
class AscendMRotaryEmbedding(MRotaryEmbedding):
# Empirical safety threshold for large Triton grids on Ascend NPU
_ASCEND_TRITON_GRID_LIMIT = 65535
def forward_triton(self,
positions: torch.Tensor,
query: torch.Tensor,
key: torch.Tensor | None = None,
offsets: torch.Tensor | None = None):
def forward_triton(
self,
positions: torch.Tensor,
query: torch.Tensor,
key: torch.Tensor | None = None,
offsets: torch.Tensor | None = None,
):
assert positions.ndim == 2
assert key is not None
@@ -571,10 +505,9 @@ class AscendMRotaryEmbedding(MRotaryEmbedding):
assert self.mrope_section
# When the grid becomes large, enable TRITON_ALL_BLOCKS_PARALLEL
# When the grid becomes large, enable TRITON_ALL_BLOCKS_PARALLEL
# to avoid scheduler/runtime failures.
if (query_shape[0] > self._ASCEND_TRITON_GRID_LIMIT and
os.environ.get("TRITON_ALL_BLOCKS_PARALLEL") != "1"):
if query_shape[0] > self._ASCEND_TRITON_GRID_LIMIT and os.environ.get("TRITON_ALL_BLOCKS_PARALLEL") != "1":
os.environ["TRITON_ALL_BLOCKS_PARALLEL"] = "1"
q, k = triton_mrope(
@@ -600,35 +533,37 @@ class AscendMRotaryEmbedding(MRotaryEmbedding):
# todo: need cann update in 8.5.0
return self.forward_triton(positions, query, key)
if self.mrope_section != [16, 24, 24] or \
get_ascend_device_type() == AscendDeviceType.A5:
if self.mrope_section != [16, 24, 24] or get_ascend_device_type() == AscendDeviceType.A5:
return super().forward_oot(positions, query, key)
import torch_npu
mrope_section = [0, 0, 0
] if positions.ndim == 1 else self.mrope_section
mrope_section = [0, 0, 0] if positions.ndim == 1 else self.mrope_section
if self.cos_sin_cache.device != query.device: # type: ignore
self.cos_sin_cache = self.cos_sin_cache.to( # type: ignore
query.device) # type: ignore
query.device
) # type: ignore
if self.cos_sin_cache.dtype != query.dtype: # type: ignore
self.cos_sin_cache = self.cos_sin_cache.to( # type: ignore
query.dtype) # type: ignore
query.dtype
) # type: ignore
query, key = torch_npu.npu_mrope(positions.contiguous(),
query.contiguous(),
key.contiguous(),
self.cos_sin_cache.contiguous(),
self.head_size,
mrope_section=mrope_section,
rotary_mode='half')
query, key = torch_npu.npu_mrope(
positions.contiguous(),
query.contiguous(),
key.contiguous(),
self.cos_sin_cache.contiguous(),
self.head_size,
mrope_section=mrope_section,
rotary_mode="half",
)
return query, key
class AscendApplyRotaryEmb(ApplyRotaryEmb):
def __init__(
self,
enforce_enable: bool = False,
@@ -647,8 +582,7 @@ class AscendApplyRotaryEmb(ApplyRotaryEmb):
cos: torch.Tensor,
sin: torch.Tensor,
) -> torch.Tensor:
x, cos, sin, origin_shape, origin_dtype = self._pre_process(
x, cos, sin)
x, cos, sin, origin_shape, origin_dtype = self._pre_process(x, cos, sin)
head_dim = x.shape[-1]
# cos, sin: [seq_len, head_dim // 2]

7
vllm_ascend/patch/worker/patch_bert.py
View File

@@ -24,15 +24,12 @@ TOKEN_TYPE_MULTIPLIER = 1 << 30
TOKEN_MASK = TOKEN_TYPE_MULTIPLIER - 1
def _encode_token_type_ids(input_ids: torch.Tensor,
token_type_ids: torch.Tensor) -> None:
def _encode_token_type_ids(input_ids: torch.Tensor, token_type_ids: torch.Tensor) -> None:
# input_ids can be padded to the right
input_ids[:token_type_ids.shape[0]].bitwise_or_(token_type_ids *
TOKEN_TYPE_MULTIPLIER)
input_ids[: token_type_ids.shape[0]].bitwise_or_(token_type_ids * TOKEN_TYPE_MULTIPLIER)
def _decode_token_type_ids(input_ids: torch.Tensor) -> torch.Tensor:
token_type_ids = input_ids // TOKEN_TYPE_MULTIPLIER
input_ids.bitwise_and_(TOKEN_MASK)

54
vllm_ascend/patch/worker/patch_deepseek.py Normal file
View File

@@ -0,0 +1,54 @@
from itertools import islice
import torch
from vllm.distributed import get_pp_group
from vllm.model_executor.models.deepseek_v2 import DeepseekV2Model, _get_llama_4_scaling
from vllm.sequence import IntermediateTensors
def forward(
self,
input_ids,
positions,
intermediate_tensors,
inputs_embeds,
):
if get_pp_group().is_first_rank:
if inputs_embeds is not None:
hidden_states = inputs_embeds
else:
hidden_states = self.embed_input_ids(input_ids)
residual = None
else:
assert intermediate_tensors is not None
hidden_states = intermediate_tensors["hidden_states"]
residual = intermediate_tensors["residual"]
# Compute llama 4 scaling once per forward pass if enabled
# Note(wxy): This is a hack fix to avoid graph mode error for torch 2.8
# We'll find a better way to remove this patch.
try:
llama_4_scaling_config = self.config.llama_4_scaling
except AttributeError:
llama_4_scaling_config = None
llama_4_scaling: torch.Tensor | None
if llama_4_scaling_config is not None:
llama_4_scaling = _get_llama_4_scaling(
original_max_position_embeddings=llama_4_scaling_config["original_max_position_embeddings"],
scaling_beta=llama_4_scaling_config["beta"],
positions=positions,
)
else:
llama_4_scaling = None
for layer in islice(self.layers, self.start_layer, self.end_layer):
hidden_states, residual = layer(positions, hidden_states, residual, llama_4_scaling)
if not get_pp_group().is_last_rank:
return IntermediateTensors({"hidden_states": hidden_states, "residual": residual})
hidden_states, _ = self.norm(hidden_states, residual)
return hidden_states
DeepseekV2Model.forward = forward

44
vllm_ascend/patch/worker/patch_distributed.py
View File

@@ -15,29 +15,25 @@
# limitations under the License.
#
from typing import List, Optional, Union
import torch
import vllm
from torch.distributed import Backend
from vllm.distributed.parallel_state import (GroupCoordinator,
_get_unique_name, _register_group)
from vllm.distributed.parallel_state import GroupCoordinator, _get_unique_name, _register_group
from vllm_ascend.distributed.device_communicators.npu_communicator import \
NPUCommunicator
from vllm_ascend.distributed.device_communicators.npu_communicator import NPUCommunicator
from vllm_ascend.utils import create_hccl_pg_options
class GroupCoordinatorPatch(GroupCoordinator):
def __init__(
self,
group_ranks: list[list[int]],
local_rank: int,
torch_distributed_backend: Union[str, Backend],
torch_distributed_backend: str | Backend,
use_device_communicator: bool, # whether to use device communicator
use_message_queue_broadcaster: bool = False,
group_name: Optional[str] = None,
group_name: str | None = None,
):
group_name = group_name or "anonymous"
self.unique_name = _get_unique_name(group_name)
@@ -52,9 +48,8 @@ class GroupCoordinatorPatch(GroupCoordinator):
for ranks in group_ranks:
device_group = torch.distributed.new_group(
ranks,
backend=torch_distributed_backend,
pg_options=hccl_pg_options)
ranks, backend=torch_distributed_backend, pg_options=hccl_pg_options
)
# a group with `gloo` backend, to allow direct coordination between
# processes through the CPU.
@@ -83,22 +78,23 @@ class GroupCoordinatorPatch(GroupCoordinator):
unique_name=self.unique_name,
)
from vllm.distributed.device_communicators.shm_broadcast import \
MessageQueue
self.mq_broadcaster: Optional[MessageQueue] = None
from vllm.distributed.device_communicators.shm_broadcast import MessageQueue
self.mq_broadcaster: MessageQueue | None = None
if use_message_queue_broadcaster and self.world_size > 1:
self.mq_broadcaster = MessageQueue.create_from_process_group(
self.cpu_group, 1 << 22, 6)
self.mq_broadcaster = MessageQueue.create_from_process_group(self.cpu_group, 1 << 22, 6)
self.use_custom_op_call = False
self.use_cpu_custom_send_recv = False
def all_to_all(self,
input_: torch.Tensor,
scatter_dim: int = 0,
gather_dim: int = -1,
scatter_sizes: Optional[List[int]] = None,
gather_sizes: Optional[List[int]] = None) -> torch.Tensor:
def all_to_all(
self,
input_: torch.Tensor,
scatter_dim: int = 0,
gather_dim: int = -1,
scatter_sizes: list[int] | None = None,
gather_sizes: list[int] | None = None,
) -> torch.Tensor:
if self.world_size == 1:
return input_
assert -input_.dim() <= scatter_dim < input_.dim(), (
@@ -108,9 +104,7 @@ class GroupCoordinatorPatch(GroupCoordinator):
f"Invalid gather dim ({gather_dim}) for input tensor with shape {input_.size()}"
)
assert self.device_communicator is not None, "device_communicator should be initialized when world_size > 1"
return self.device_communicator.all_to_all(input_, scatter_dim,
gather_dim, scatter_sizes,
gather_sizes)
return self.device_communicator.all_to_all(input_, scatter_dim, gather_dim, scatter_sizes, gather_sizes)
def all_reduce(self, input_):
if self.world_size == 1:

6
vllm_ascend/patch/worker/patch_huanyuan_vl.py
View File

@@ -13,15 +13,17 @@
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#from collections.abc import Iterable
# from collections.abc import Iterable
from vllm.transformers_utils.processors.hunyuan_vl import HunYuanVLProcessor
_original_call = HunYuanVLProcessor.__call__
def _patched_call(self, images=None, text=None, videos=None, **kwargs):
"""Remove add_special_tokens requirement."""
kwargs.pop("add_special_tokens", None)
return _original_call(self, images=images, text=text, videos=videos, **kwargs)
HunYuanVLProcessor.__call__ = _patched_call
HunYuanVLProcessor.__call__ = _patched_call

2
vllm_ascend/patch/worker/patch_module.py
View File

@@ -13,7 +13,6 @@ def _argsort(tensor, *args, **kwargs):
class _TorchWrapper:
def __init__(self):
self._raw_torch = torch
@@ -32,5 +31,6 @@ def patch_torch_npu_argsort():
global _is_patched
if not _is_patched:
import vllm.v1.attention.backends.gdn_attn as gdn_attn
gdn_attn.torch = _TorchWrapper()
_is_patched = True

3
vllm_ascend/patch/worker/patch_multimodal_merge.py
View File

@@ -18,8 +18,7 @@
import torch
import vllm
from vllm.model_executor.models.utils import (_embedding_count_expression,
_flatten_embeddings)
from vllm.model_executor.models.utils import _embedding_count_expression, _flatten_embeddings
from vllm.multimodal import NestedTensors

94
vllm_ascend/patch/worker/patch_qwen3_next.py
View File

@@ -13,33 +13,27 @@
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#from collections.abc import Iterable
# from collections.abc import Iterable
import torch
from einops import rearrange
from torch import nn
from vllm.config import CUDAGraphMode
from vllm.forward_context import get_forward_context
from vllm.model_executor.layers.fla.ops import (
chunk_gated_delta_rule, fused_recurrent_gated_delta_rule)
from vllm.model_executor.layers.fla.ops import chunk_gated_delta_rule, fused_recurrent_gated_delta_rule
from vllm.model_executor.layers.mamba.abstract import MambaBase
from vllm.model_executor.layers.mamba.ops.causal_conv1d import (
causal_conv1d_fn, causal_conv1d_update)
from vllm.model_executor.models.qwen3_next import (Qwen3NextGatedDeltaNet,
fused_gdn_gating)
from vllm.model_executor.layers.mamba.ops.causal_conv1d import causal_conv1d_fn, causal_conv1d_update
from vllm.model_executor.models.qwen3_next import Qwen3NextGatedDeltaNet
from vllm.triton_utils import triton
from vllm.v1.attention.backend import AttentionMetadata # type: ignore
from vllm.v1.attention.backends.gdn_attn import GDNAttentionMetadata
from vllm_ascend.ops.triton.fla.fused_qkvzba_split_reshape import \
fused_qkvzba_split_reshape_cat
from vllm_ascend.ops.triton.fla.sigmoid_gating import \
fused_sigmoid_gating_delta_rule_update
from vllm_ascend.ops.triton.fla.fused_qkvzba_split_reshape import fused_qkvzba_split_reshape_cat
from vllm_ascend.ops.triton.fla.sigmoid_gating import fused_sigmoid_gating_delta_rule_update
from vllm_ascend.ops.triton.fused_gdn_gating import fused_gdn_gating_patch
class AscendQwen3Next_GatedDeltaNet(nn.Module, MambaBase):
def forward(
self,
hidden_states: torch.Tensor,
@@ -61,10 +55,8 @@ class AscendQwen3Next_GatedDeltaNet(nn.Module, MambaBase):
forward_context = get_forward_context()
is_cuda_graph = forward_context.cudagraph_runtime_mode != CUDAGraphMode.NONE
# triton grid should be less than 66536
divide_grid = projected_states_qkvz.shape[0] * triton.cdiv(
self.num_k_heads, self.tp_size)
if self.num_v_heads // self.num_k_heads in [1, 2, 4] and \
is_cuda_graph and divide_grid < 65536:
divide_grid = projected_states_qkvz.shape[0] * triton.cdiv(self.num_k_heads, self.tp_size)
if self.num_v_heads // self.num_k_heads in [1, 2, 4] and is_cuda_graph and divide_grid < 65536:
mixed_qkv, z, b, a = fused_qkvzba_split_reshape_cat(
projected_states_qkvz,
projected_states_ba,
@@ -74,10 +66,8 @@ class AscendQwen3Next_GatedDeltaNet(nn.Module, MambaBase):
self.head_v_dim,
)
else:
query, key, value, z, b, a = self.fix_query_key_value_ordering(
projected_states_qkvz, projected_states_ba)
query, key, value = map(lambda x: rearrange(x, 'l p d -> l (p d)'),
(query, key, value))
query, key, value, z, b, a = self.fix_query_key_value_ordering(projected_states_qkvz, projected_states_ba)
query, key, value = map(lambda x: rearrange(x, "l p d -> l (p d)"), (query, key, value))
mixed_qkv = torch.cat((query, key, value), dim=-1)
# ============================================================
@@ -150,16 +140,14 @@ class AscendQwen3Next_GatedDeltaNet(nn.Module, MambaBase):
a = a[:num_actual_tokens]
# 1. Convolution sequence transformation
conv_weights = self.conv1d.weight.view(self.conv1d.weight.size(0),
self.conv1d.weight.size(2))
conv_weights = self.conv1d.weight.view(self.conv1d.weight.size(0), self.conv1d.weight.size(2))
if spec_sequence_masks is not None:
if attn_metadata.num_prefills == 0 and attn_metadata.num_decodes == 0:
mixed_qkv_spec = mixed_qkv
mixed_qkv_non_spec = None
else:
mixed_qkv_spec = mixed_qkv.index_select(0, spec_token_indx)
mixed_qkv_non_spec = mixed_qkv.index_select(
0, non_spec_token_indx)
mixed_qkv_non_spec = mixed_qkv.index_select(0, non_spec_token_indx)
else:
mixed_qkv_spec = None
mixed_qkv_non_spec = mixed_qkv
@@ -172,8 +160,7 @@ class AscendQwen3Next_GatedDeltaNet(nn.Module, MambaBase):
conv_weights,
self.conv1d.bias,
self.activation,
conv_state_indices=spec_state_indices_tensor[:, 0]
[:attn_metadata.num_spec_decodes],
conv_state_indices=spec_state_indices_tensor[:, 0][: attn_metadata.num_spec_decodes],
num_accepted_tokens=num_accepted_tokens,
query_start_loc=spec_query_start_loc,
max_query_len=spec_state_indices_tensor.size(-1),
@@ -204,21 +191,16 @@ class AscendQwen3Next_GatedDeltaNet(nn.Module, MambaBase):
conv_weights,
self.conv1d.bias,
self.activation,
conv_state_indices=
non_spec_state_indices_tensor[:attn_metadata.
num_actual_tokens],
conv_state_indices=non_spec_state_indices_tensor[: attn_metadata.num_actual_tokens],
validate_data=True,
)
else:
mixed_qkv_non_spec = None
query_spec, key_spec, value_spec = self.rearrange_mixed_qkv(
mixed_qkv_spec)
query_non_spec, key_non_spec, value_non_spec = self.rearrange_mixed_qkv(
mixed_qkv_non_spec)
query_spec, key_spec, value_spec = self.rearrange_mixed_qkv(mixed_qkv_spec)
query_non_spec, key_non_spec, value_non_spec = self.rearrange_mixed_qkv(mixed_qkv_non_spec)
if attn_metadata.num_prefills > 0 or spec_sequence_masks is not None:
g, beta = fused_gdn_gating_patch(self.A_log, a, b,
self.dt_bias)
g, beta = fused_gdn_gating_patch(self.A_log, a, b, self.dt_bias)
if spec_sequence_masks is not None:
if attn_metadata.num_prefills == 0 and attn_metadata.num_decodes == 0:
g_spec = g
@@ -248,8 +230,7 @@ class AscendQwen3Next_GatedDeltaNet(nn.Module, MambaBase):
beta=beta_spec,
initial_state=ssm_state,
inplace_final_state=True,
cu_seqlens=spec_query_start_loc[:attn_metadata.
num_spec_decodes + 1],
cu_seqlens=spec_query_start_loc[: attn_metadata.num_spec_decodes + 1],
ssm_state_indices=spec_state_indices_tensor,
num_accepted_tokens=num_accepted_tokens,
use_qk_l2norm_in_kernel=True,
@@ -259,8 +240,7 @@ class AscendQwen3Next_GatedDeltaNet(nn.Module, MambaBase):
# 2.2: Process the remaining part
if attn_metadata.num_prefills > 0:
initial_state = ssm_state[
non_spec_state_indices_tensor].contiguous()
initial_state = ssm_state[non_spec_state_indices_tensor].contiguous()
initial_state[~has_initial_state, ...] = 0
(
core_attn_out_non_spec,
@@ -278,24 +258,20 @@ class AscendQwen3Next_GatedDeltaNet(nn.Module, MambaBase):
use_qk_l2norm_in_kernel=True,
)
# Init cache
ssm_state[
non_spec_state_indices_tensor] = last_recurrent_state.to(
ssm_state.dtype)
ssm_state[non_spec_state_indices_tensor] = last_recurrent_state.to(ssm_state.dtype)
elif attn_metadata.num_decodes > 0:
core_attn_out_non_spec, last_recurrent_state = (
fused_recurrent_gated_delta_rule(
q=query_non_spec,
k=key_non_spec,
v=value_non_spec,
g=g_non_spec,
beta=beta_non_spec,
initial_state=ssm_state,
inplace_final_state=True,
cu_seqlens=non_spec_query_start_loc[:attn_metadata.
num_decodes + 1],
ssm_state_indices=non_spec_state_indices_tensor,
use_qk_l2norm_in_kernel=True,
))
core_attn_out_non_spec, last_recurrent_state = fused_recurrent_gated_delta_rule(
q=query_non_spec,
k=key_non_spec,
v=value_non_spec,
g=g_non_spec,
beta=beta_non_spec,
initial_state=ssm_state,
inplace_final_state=True,
cu_seqlens=non_spec_query_start_loc[: attn_metadata.num_decodes + 1],
ssm_state_indices=non_spec_state_indices_tensor,
use_qk_l2norm_in_kernel=True,
)
else:
core_attn_out_non_spec, last_recurrent_state = None, None
@@ -324,14 +300,12 @@ class AscendQwen3Next_GatedDeltaNet(nn.Module, MambaBase):
device=core_attn_out_non_spec.device,
)
merged_out.index_copy_(1, spec_token_indx, core_attn_out_spec)
merged_out.index_copy_(1, non_spec_token_indx,
core_attn_out_non_spec)
merged_out.index_copy_(1, non_spec_token_indx, core_attn_out_non_spec)
core_attn_out[:num_actual_tokens] = merged_out.squeeze(0)
elif spec_sequence_masks is not None:
core_attn_out[:num_actual_tokens] = core_attn_out_spec.squeeze(0)
else:
core_attn_out[:num_actual_tokens] = core_attn_out_non_spec.squeeze(
0)
core_attn_out[:num_actual_tokens] = core_attn_out_non_spec.squeeze(0)
Qwen3NextGatedDeltaNet.forward = AscendQwen3Next_GatedDeltaNet.forward

7
vllm_ascend/patch/worker/patch_qwen3_next_mtp.py
View File

@@ -1,13 +1,15 @@
import torch
import vllm.v1.worker.utils as utils
from vllm.v1.worker.utils import defaultdict, extract_layer_index
from vllm_ascend.utils import vllm_version_is
if vllm_version_is("v0.15.0"):
from vllm.attention.layer import Attention # type: ignore
from vllm.attention.layer import Attention # type: ignore
else:
from vllm.model_executor.layers.attention import Attention
# Without this patch, it will raise an exception when initialize kv_cache.
# TODO To remove the patch, we need check why the original bind_kv_cache raises an NotImplementedError.
def bind_kv_cache(
@@ -38,8 +40,7 @@ def bind_kv_cache(
# Convert kv_caches dict to a list of tensors in the order of layer_index.
index2name = defaultdict(list)
for layer_name in kv_caches:
index2name[extract_layer_index(layer_name,
num_attn_module)].append(layer_name)
index2name[extract_layer_index(layer_name, num_attn_module)].append(layer_name)
for layer_index in sorted(index2name.keys()):
layer_names = index2name[layer_index]

4
vllm_ascend/patch/worker/patch_rejection_sampler.py
View File

@@ -1,8 +1,6 @@
import vllm.v1.sample.rejection_sampler as rs
from vllm_ascend.sample.rejection_sampler import (apply_sampling_constraints,
expand_batch_to_tokens,
rejection_sample)
from vllm_ascend.sample.rejection_sampler import apply_sampling_constraints, expand_batch_to_tokens, rejection_sample
# TODO: delete this patch after apply_sampling_constraints and rejection_sample
# are extracted to as class func of RejectionSampler

4
vllm_ascend/patch/worker/patch_rope.py
View File

@@ -17,12 +17,10 @@
import torch
import torch.nn as nn
from vllm.model_executor.layers.rotary_embedding.base import \
RotaryEmbeddingBase
from vllm.model_executor.layers.rotary_embedding.base import RotaryEmbeddingBase
class AscendRotaryEmbeddingBase(nn.Module):
def get_cos_sin(self, seqlen: int) -> tuple[torch.Tensor, torch.Tensor]:
cos_sin = self.cos_sin_cache[:seqlen]
cos, sin = cos_sin.chunk(2, dim=-1)

13
vllm_ascend/patch/worker/patch_triton.py
View File

@@ -3,16 +3,15 @@ import vllm.v1.worker.gpu.sample.gumbel
from vllm_ascend.ops.triton.fla.chunk import chunk_gated_delta_rule
from vllm_ascend.ops.triton.fla.layernorm_guard import LayerNormFn
from vllm_ascend.ops.triton.fla.sigmoid_gating import \
fused_recurrent_gated_delta_rule_fwd_kernel
from vllm_ascend.ops.triton.mamba.causal_conv1d import (
causal_conv1d_fn, causal_conv1d_update_npu)
from vllm_ascend.worker.v2.sample.gumbel import \
gumbel_sample as ascend_gumbel_sample
from vllm_ascend.ops.triton.fla.sigmoid_gating import fused_recurrent_gated_delta_rule_fwd_kernel
from vllm_ascend.ops.triton.mamba.causal_conv1d import causal_conv1d_fn, causal_conv1d_update_npu
from vllm_ascend.worker.v2.sample.gumbel import gumbel_sample as ascend_gumbel_sample
vllm.model_executor.layers.mamba.ops.causal_conv1d.causal_conv1d_update = causal_conv1d_update_npu
vllm.model_executor.layers.mamba.ops.causal_conv1d.causal_conv1d_fn = causal_conv1d_fn
vllm.model_executor.layers.fla.ops.fused_recurrent.fused_recurrent_gated_delta_rule_fwd_kernel = fused_recurrent_gated_delta_rule_fwd_kernel
vllm.model_executor.layers.fla.ops.fused_recurrent.fused_recurrent_gated_delta_rule_fwd_kernel = (
fused_recurrent_gated_delta_rule_fwd_kernel
)
vllm.model_executor.layers.fla.ops.layernorm_guard.LayerNormFn = LayerNormFn
vllm.model_executor.layers.fla.ops.chunk_gated_delta_rule = chunk_gated_delta_rule
vllm.v1.worker.gpu.sample.gumbel.gumbel_sample = ascend_gumbel_sample

13
vllm_ascend/patch/worker/patch_unquantized_gemm.py
View File

@@ -36,11 +36,14 @@ def unquantized_gemm_fake(
return torch.empty(output_shape, dtype=x.dtype, device=x.device)
direct_register_custom_op(op_name="unquantized_gemm",
op_func=unquantized_gemm,
fake_impl=unquantized_gemm_fake,
mutates_args=[],
dispatch_key="PrivateUse1")
direct_register_custom_op(
op_name="unquantized_gemm",
op_func=unquantized_gemm,
fake_impl=unquantized_gemm_fake,
mutates_args=[],
dispatch_key="PrivateUse1",
)
def default_unquantized_gemm(
layer: torch.nn.Module,

31
vllm_ascend/patch/worker/patch_v2_egale.py
View File

@@ -19,11 +19,10 @@
import numpy as np
import torch
import vllm
from vllm.v1.sample.metadata import SamplingMetadata
from vllm.v1.worker.gpu.input_batch import InputBatch
from vllm.v1.worker.gpu.sample.gumbel import gumbel_sample
from vllm.v1.sample.metadata import SamplingMetadata
from vllm.v1.worker.gpu.spec_decode.eagle import (prepare_eagle_decode,
prepare_eagle_inputs)
from vllm.v1.worker.gpu.spec_decode.eagle import prepare_eagle_decode, prepare_eagle_inputs
from vllm_ascend.worker.v2.attn_utils import build_attn_metadata
@@ -54,8 +53,7 @@ def propose(
# seq_lens) of the target model.
if aux_hidden_states:
assert self.method == "eagle3"
hidden_states = self.model.combine_hidden_states(
torch.cat(aux_hidden_states, dim=-1))
hidden_states = self.model.combine_hidden_states(torch.cat(aux_hidden_states, dim=-1))
else:
hidden_states = last_hidden_states
num_tokens = input_batch.num_tokens_after_padding
@@ -95,19 +93,12 @@ def propose(
seeds = self.seeds[:num_reqs].clone()
pos = self.input_buffers.positions[:num_reqs].clone()
# Gather the values and copy them to the pre-allocated buffers.
torch.gather(sampling_metadata.temperature,
0,
cu_num_logits,
out=temperature)
torch.gather(sampling_metadata.temperature, 0, cu_num_logits, out=temperature)
torch.gather(sampling_metadata.seeds, 0, cu_num_logits, out=seeds)
torch.gather(input_batch.positions, 0, last_token_indices, out=pos)
# NOTE(woosuk): We must add 1 to the positions to match the Gumbel noise
# used for draft and target sampling.
draft_tokens = gumbel_sample(logits,
temperature,
seeds,
pos + 1,
apply_temperature=True)
draft_tokens = gumbel_sample(logits, temperature, seeds, pos + 1, apply_temperature=True)
if self.num_speculative_steps == 1:
# Early exit.
return draft_tokens.view(-1, 1)
@@ -127,9 +118,8 @@ def propose(
self.max_num_reqs,
)
query_start_loc = self.input_buffers.query_start_loc
query_start_loc_gpu = query_start_loc.gpu[:num_reqs + 1]
slot_mappings = self.block_tables.compute_slot_mappings(
query_start_loc_gpu, pos)
query_start_loc_gpu = query_start_loc.gpu[: num_reqs + 1]
slot_mappings = self.block_tables.compute_slot_mappings(query_start_loc_gpu, pos)
cudagraph_size = self.cudagraph_manager.get_cudagraph_size(num_reqs)
if cudagraph_size is not None:
@@ -138,8 +128,8 @@ def propose(
return self.draft_tokens[:num_reqs]
# Run eager mode.
query_start_loc.np[:num_reqs + 1] = np.arange(num_reqs + 1)
query_start_loc_cpu = query_start_loc.cpu[:num_reqs + 1]
query_start_loc.np[: num_reqs + 1] = np.arange(num_reqs + 1)
query_start_loc_cpu = query_start_loc.cpu[: num_reqs + 1]
# HACK(woosuk)
seq_lens_np = np.full(num_reqs, self.max_model_len, dtype=np.int32)
block_tables = [x[:num_reqs] for x in self.block_tables.input_block_tables]
@@ -158,8 +148,7 @@ def propose(
slot_mappings=slot_mappings,
kv_cache_config=self.kv_cache_config,
)
self.generate_draft(num_reqs, attn_metadata,
num_tokens_across_dp=None) # FIXME
self.generate_draft(num_reqs, attn_metadata, num_tokens_across_dp=None) # FIXME
return self.draft_tokens[:num_reqs]

96
vllm_ascend/worker/npu_input_batch.py
View File

@@ -23,15 +23,13 @@ from vllm.lora.request import LoRARequest
from vllm.pooling_params import PoolingParams
from vllm.v1.outputs import LogprobsTensors
from vllm.v1.pool.metadata import PoolingStates
from vllm.v1.sample.logits_processor import (BatchUpdateBuilder,
LogitsProcessors)
from vllm.v1.sample.logits_processor import BatchUpdateBuilder, LogitsProcessors
from vllm.v1.worker.gpu_input_batch import InputBatch
from vllm_ascend.worker.block_table import MultiGroupBlockTable
class NPUInputBatch(InputBatch):
def __init__(
self,
max_num_reqs: int,
@@ -72,10 +70,9 @@ class NPUInputBatch(InputBatch):
pin_memory=False,
)
self.token_ids_cpu = self.token_ids_cpu_tensor.numpy()
self.is_token_ids_tensor = torch.zeros((max_num_reqs, max_model_len),
device="cpu",
dtype=bool,
pin_memory=False)
self.is_token_ids_tensor = torch.zeros(
(max_num_reqs, max_model_len), device="cpu", dtype=bool, pin_memory=False
)
self.is_token_ids = self.is_token_ids_tensor.numpy()
# Store prompt embeddings per request to avoid OOM from large upfront
# allocation if max_model_len is big.
@@ -85,13 +82,12 @@ class NPUInputBatch(InputBatch):
self.num_tokens_no_spec = np.zeros(max_num_reqs, dtype=np.int32)
self.num_prompt_tokens = np.zeros(max_num_reqs, dtype=np.int32)
self.num_computed_tokens_cpu_tensor = torch.zeros(
(max_num_reqs, ),
(max_num_reqs,),
device="cpu",
dtype=torch.int32,
pin_memory=pin_memory,
)
self.num_computed_tokens_cpu = self.num_computed_tokens_cpu_tensor.numpy(
)
self.num_computed_tokens_cpu = self.num_computed_tokens_cpu_tensor.numpy()
# Block table.
self.block_table = MultiGroupBlockTable(
@@ -107,34 +103,21 @@ class NPUInputBatch(InputBatch):
)
# Sampling-related.
self.temperature = torch.empty((max_num_reqs, ),
dtype=torch.float32,
device=device)
self.temperature_cpu_tensor = torch.empty((max_num_reqs, ),
dtype=torch.float32,
device="cpu",
pin_memory=pin_memory)
self.temperature = torch.empty((max_num_reqs,), dtype=torch.float32, device=device)
self.temperature_cpu_tensor = torch.empty(
(max_num_reqs,), dtype=torch.float32, device="cpu", pin_memory=pin_memory
)
self.temperature_cpu = self.temperature_cpu_tensor.numpy()
self.greedy_reqs: set[str] = set()
self.random_reqs: set[str] = set()
self.top_p = torch.empty((max_num_reqs, ),
dtype=torch.float32,
device=device)
self.top_p_cpu_tensor = torch.empty((max_num_reqs, ),
dtype=torch.float32,
device="cpu",
pin_memory=pin_memory)
self.top_p = torch.empty((max_num_reqs,), dtype=torch.float32, device=device)
self.top_p_cpu_tensor = torch.empty((max_num_reqs,), dtype=torch.float32, device="cpu", pin_memory=pin_memory)
self.top_p_cpu = self.top_p_cpu_tensor.numpy()
self.top_p_reqs: set[str] = set()
self.top_k = torch.empty((max_num_reqs, ),
dtype=torch.int32,
device=device)
self.top_k_cpu_tensor = torch.empty((max_num_reqs, ),
dtype=torch.int32,
device="cpu",
pin_memory=pin_memory)
self.top_k = torch.empty((max_num_reqs,), dtype=torch.int32, device=device)
self.top_k_cpu_tensor = torch.empty((max_num_reqs,), dtype=torch.int32, device="cpu", pin_memory=pin_memory)
self.top_k_cpu = self.top_k_cpu_tensor.numpy()
self.top_k_reqs: set[str] = set()
@@ -142,54 +125,37 @@ class NPUInputBatch(InputBatch):
self.spec_decode_unsupported_reqs: set[str] = set()
# Frequency penalty related data structures
self.frequency_penalties = torch.empty((max_num_reqs, ),
dtype=torch.float,
device=device)
self.frequency_penalties = torch.empty((max_num_reqs,), dtype=torch.float, device=device)
self.frequency_penalties_cpu_tensor = torch.empty(
(max_num_reqs, ),
dtype=torch.float,
device="cpu",
pin_memory=pin_memory)
self.frequency_penalties_cpu = self.frequency_penalties_cpu_tensor.numpy(
(max_num_reqs,), dtype=torch.float, device="cpu", pin_memory=pin_memory
)
self.frequency_penalties_cpu = self.frequency_penalties_cpu_tensor.numpy()
self.frequency_penalties_reqs: set[str] = set()
# Presence penalty related data structures
self.presence_penalties = torch.empty((max_num_reqs, ),
dtype=torch.float,
device=device)
self.presence_penalties_cpu_tensor = torch.empty((max_num_reqs, ),
dtype=torch.float,
device="cpu",
pin_memory=pin_memory)
self.presence_penalties_cpu = self.presence_penalties_cpu_tensor.numpy(
self.presence_penalties = torch.empty((max_num_reqs,), dtype=torch.float, device=device)
self.presence_penalties_cpu_tensor = torch.empty(
(max_num_reqs,), dtype=torch.float, device="cpu", pin_memory=pin_memory
)
self.presence_penalties_cpu = self.presence_penalties_cpu_tensor.numpy()
self.presence_penalties_reqs: set[str] = set()
# Repetition penalty related data structures
self.repetition_penalties = torch.empty((max_num_reqs, ),
dtype=torch.float,
device=device)
self.repetition_penalties = torch.empty((max_num_reqs,), dtype=torch.float, device=device)
self.repetition_penalties_cpu_tensor = torch.empty(
(max_num_reqs, ),
dtype=torch.float,
device="cpu",
pin_memory=pin_memory)
self.repetition_penalties_cpu = self.repetition_penalties_cpu_tensor.numpy(
(max_num_reqs,), dtype=torch.float, device="cpu", pin_memory=pin_memory
)
self.repetition_penalties_cpu = self.repetition_penalties_cpu_tensor.numpy()
self.repetition_penalties_reqs: set[str] = set()
# Speculative decoding
self.num_accepted_tokens_cpu_tensor = torch.ones((max_num_reqs, ),
dtype=torch.int64,
device="cpu",
pin_memory=pin_memory)
self.num_accepted_tokens_cpu = self.num_accepted_tokens_cpu_tensor.numpy(
self.num_accepted_tokens_cpu_tensor = torch.ones(
(max_num_reqs,), dtype=torch.int64, device="cpu", pin_memory=pin_memory
)
self.num_accepted_tokens_cpu = self.num_accepted_tokens_cpu_tensor.numpy()
# lora related
self.request_lora_mapping = np.zeros((self.max_num_reqs, ),
dtype=np.int64)
self.request_lora_mapping = np.zeros((self.max_num_reqs,), dtype=np.int64)
self.lora_id_to_request_ids: dict[int, set[str]] = {}
self.lora_id_to_lora_request: dict[int, LoRARequest] = {}
@@ -218,8 +184,7 @@ class NPUInputBatch(InputBatch):
# req_index -> bad_words_token_ids
self.bad_words_token_ids: dict[int, list[list[int]]] = {}
self.logits_processing_needs_token_ids = np.zeros(max_num_reqs,
dtype=bool)
self.logits_processing_needs_token_ids = np.zeros(max_num_reqs, dtype=bool)
self.req_output_token_ids: list[list[int] | None] = []
@@ -229,8 +194,7 @@ class NPUInputBatch(InputBatch):
self.logitsprocs_need_output_token_ids = logitsprocs_need_output_token_ids
# Store last speculative tokens for sampler.
self.spec_token_ids: list[list[int]] = [[]
for _ in range(max_num_reqs)]
self.spec_token_ids: list[list[int]] = [[] for _ in range(max_num_reqs)]
# This is updated each time the batch constituents change.
self.sampling_metadata = self._make_sampling_metadata()

9
vllm_ascend/worker/v2/aclgraph_utils.py
View File

@@ -22,19 +22,16 @@ from typing import Any
import torch
import torch.nn as nn
from vllm.config import VllmConfig
from vllm.v1.attention.backend import AttentionMetadataBuilder
from vllm.v1.kv_cache_interface import KVCacheConfig
from vllm.v1.worker.gpu.block_table import BlockTables
from vllm.v1.worker.gpu.cudagraph_utils import CudaGraphManager
from vllm.v1.worker.gpu.cudagraph_utils import \
prepare_inputs_to_capture as prepare_inputs_to_capture_gpu
from vllm.v1.worker.gpu.cudagraph_utils import prepare_inputs_to_capture as prepare_inputs_to_capture_gpu
from vllm.v1.worker.gpu.input_batch import InputBuffers
from vllm.v1.attention.backend import AttentionMetadataBuilder
from vllm_ascend.worker.v2.utils import torch_cuda_wrapper
class AclGraphManager(CudaGraphManager):
"""ACL Graph Manager for Ascend NPUs."""
@@ -51,7 +48,7 @@ class AclGraphManager(CudaGraphManager):
attn_metadata_builders: list[AttentionMetadataBuilder],
kv_cache_config: KVCacheConfig,
) -> None:
with (torch_cuda_wrapper(), prepare_capture_inputs_wrapper()):
with torch_cuda_wrapper(), prepare_capture_inputs_wrapper():
super().capture_graph(
num_tokens,
model,

24
vllm_ascend/worker/v2/attn_utils.py
View File

@@ -18,19 +18,17 @@
#
from collections.abc import Sequence
from typing import Any, Tuple
from typing import Any
import numpy as np
import torch
from vllm.config import VllmConfig
from vllm.v1.kv_cache_interface import EncoderOnlyAttentionSpec, KVCacheConfig
from vllm.v1.attention.backend import AttentionMetadataBuilder
from vllm.v1.kv_cache_interface import EncoderOnlyAttentionSpec, KVCacheConfig
from vllm_ascend.attention.attention_mask import AttentionMaskBuilder
from vllm_ascend.attention.attention_v1 import AscendAttentionState
from vllm_ascend.attention.utils import (AscendCommonAttentionMetadata,
AscendPrefillContextParallelMetadata)
from vllm_ascend.attention.utils import AscendCommonAttentionMetadata, AscendPrefillContextParallelMetadata
_ATTENTION_MASK_BUILDER = None
@@ -59,8 +57,7 @@ def build_attn_metadata(
attn_state: Any | None = None,
graph_pad_size: int = -1,
num_input_tokens: int = 0,
prefill_context_parallel_metadata: AscendPrefillContextParallelMetadata
| None = None,
prefill_context_parallel_metadata: AscendPrefillContextParallelMetadata | None = None,
) -> dict[str, Any]:
"""Build attention metadata for Ascend NPUs."""
# TODO(Ronald1995): optimize AscendCommonAttentionMetadata.
@@ -92,7 +89,8 @@ def build_attn_metadata(
graph_pad_size=graph_pad_size,
num_input_tokens=num_input_tokens,
prefill_context_parallel_metadata=prefill_context_parallel_metadata,
max_seq_len=max_seq_len)
max_seq_len=max_seq_len,
)
attn_metadata_builder = attn_metadata_builders[i]
metadata = attn_metadata_builder.build(
@@ -114,8 +112,8 @@ def build_attn_state(
"""Build attention state for npu's attention backend."""
if vllm_config.model_config.runner_type == "pooling":
if isinstance(
vllm_config.kv_cache_config.kv_cache_groups[0].kv_cache_spec,
EncoderOnlyAttentionSpec,
vllm_config.kv_cache_config.kv_cache_groups[0].kv_cache_spec,
EncoderOnlyAttentionSpec,
):
attn_state = AscendAttentionState.PrefillNoCache
else:
@@ -126,16 +124,14 @@ def build_attn_state(
# but only one token is not hit in cache.
elif np.all(num_scheduled_tokens == 1):
attn_state = AscendAttentionState.DecodeOnly
if (vllm_config.speculative_config
and vllm_config.speculative_config.method == 'mtp'):
if vllm_config.speculative_config and vllm_config.speculative_config.method == "mtp":
# SpecDecoding now supports seq_len=1 and seq_len=2
# In Prefilling Decoding Disaggregation scenario, SpecDecoding
# need to supports seq_len=1
attn_state = AscendAttentionState.SpecDecoding
# Speculative decoding.
elif np.all(num_valid_tokens == 1):
if (vllm_config.speculative_config
and vllm_config.speculative_config.method == 'mtp'):
if vllm_config.speculative_config and vllm_config.speculative_config.method == "mtp":
attn_state = AscendAttentionState.SpecDecoding
else:
attn_state = AscendAttentionState.ChunkedPrefill

74
vllm_ascend/worker/v2/model_runner.py
View File

@@ -22,15 +22,16 @@ import torch
from vllm.config import VllmConfig
from vllm.logger import init_logger
from vllm.v1.core.sched.output import SchedulerOutput
from vllm.v1.worker.gpu.input_batch import (InputBatch,
combine_sampled_and_draft_tokens,
prepare_pos_seq_lens,
prepare_prefill_inputs)
from vllm.v1.worker.gpu.input_batch import (
InputBatch,
combine_sampled_and_draft_tokens,
prepare_pos_seq_lens,
prepare_prefill_inputs,
)
from vllm.v1.worker.gpu.model_runner import GPUModelRunner
from vllm_ascend.worker.v2.aclgraph_utils import AclGraphManager
from vllm_ascend.worker.v2.attn_utils import (build_attn_metadata,
build_attn_state)
from vllm_ascend.worker.v2.attn_utils import build_attn_metadata, build_attn_state
from vllm_ascend.worker.v2.input_batch import AscendInputBuffers
from vllm_ascend.worker.v2.sample.sampler import AscendSampler
from vllm_ascend.worker.v2.spec_decode import init_speculator
@@ -45,7 +46,7 @@ class NPUModelRunner(GPUModelRunner):
"""Model runner for Ascend NPUs."""
def __init__(self, vllm_config: VllmConfig, device: torch.device):
with (torch_cuda_wrapper(), uva_wrapper()):
with torch_cuda_wrapper(), uva_wrapper():
super().__init__(vllm_config, device)
# because we will override these attribute, delete these attribute to
@@ -94,7 +95,8 @@ class NPUModelRunner(GPUModelRunner):
# we need to adjust triton operators in sampler,
# so reinitialize sampler here.
self.sampler: AscendSampler = AscendSampler(
logprobs_mode=self.model_config.logprobs_mode, )
logprobs_mode=self.model_config.logprobs_mode,
)
# we need to copy num_computed_tokens back to cpu to help
# update actual seq_lens_cpu. gpu attention backend doesn't need these
@@ -131,16 +133,12 @@ class NPUModelRunner(GPUModelRunner):
self._update_seq_lens_cpu(scheduler_output, req_ids)
num_scheduled_tokens = np.array(
[scheduler_output.num_scheduled_tokens[i] for i in req_ids],
dtype=np.int32)
num_scheduled_tokens = np.array([scheduler_output.num_scheduled_tokens[i] for i in req_ids], dtype=np.int32)
num_valid_tokens = num_scheduled_tokens
if scheduler_output.scheduled_spec_decode_tokens:
num_valid_tokens = np.array(
[
num_tokens - len(
scheduler_output.scheduled_spec_decode_tokens.get(
i, []))
num_tokens - len(scheduler_output.scheduled_spec_decode_tokens.get(i, []))
for num_tokens, i in zip(num_scheduled_tokens, req_ids)
],
dtype=np.int32,
@@ -153,9 +151,7 @@ class NPUModelRunner(GPUModelRunner):
num_valid_tokens,
)
idx_mapping_list = [
self.req_states.req_id_to_index[req_id] for req_id in req_ids
]
idx_mapping_list = [self.req_states.req_id_to_index[req_id] for req_id in req_ids]
idx_mapping = self.input_buffers.idx_mapping
idx_mapping.np[:num_reqs] = idx_mapping_list
idx_mapping_np = idx_mapping.np[:num_reqs]
@@ -167,16 +163,11 @@ class NPUModelRunner(GPUModelRunner):
# No draft token scheduled (common case).
total_num_draft_tokens = 0
total_num_logits = num_reqs
cu_num_logits = torch.arange(num_reqs + 1,
device=self.device,
dtype=torch.int32)
cu_num_logits = torch.arange(num_reqs + 1, device=self.device, dtype=torch.int32)
else:
draft_tokens = scheduler_output.scheduled_spec_decode_tokens
num_draft_tokens = np.array(
[
len(draft_tokens[req_id]) if req_id in draft_tokens else 0
for req_id in req_ids
],
[len(draft_tokens[req_id]) if req_id in draft_tokens else 0 for req_id in req_ids],
dtype=np.int32,
)
total_num_draft_tokens = int(num_draft_tokens.sum())
@@ -184,10 +175,9 @@ class NPUModelRunner(GPUModelRunner):
np.cumsum(
num_draft_tokens + 1,
out=self.input_buffers.cu_num_logits.np[1:num_reqs + 1],
out=self.input_buffers.cu_num_logits.np[1 : num_reqs + 1],
)
cu_num_logits = self.input_buffers.cu_num_logits.copy_to_gpu(
num_reqs + 1)
cu_num_logits = self.input_buffers.cu_num_logits.copy_to_gpu(num_reqs + 1)
# Block tables: num_kv_cache_groups x [num_reqs, max_num_blocks]
block_tables = self.block_tables.gather_block_tables(idx_mapping_npu)
@@ -195,20 +185,15 @@ class NPUModelRunner(GPUModelRunner):
# Get query_start_loc.
np.cumsum(
num_scheduled_tokens,
out=self.input_buffers.query_start_loc.np[1:num_reqs + 1],
out=self.input_buffers.query_start_loc.np[1 : num_reqs + 1],
)
# Pad for full CUDA graph mode.
# Some attention backends like FA3 require query_start_loc to be non-decreasing.
self.input_buffers.query_start_loc.np[num_reqs + 1:] = num_tokens
self.input_buffers.query_start_loc.np[num_reqs + 1 :] = num_tokens
self.input_buffers.query_start_loc.copy_to_gpu()
query_start_loc_gpu = self.input_buffers.query_start_loc.gpu[:
num_reqs +
1]
query_start_loc_cpu = self.input_buffers.query_start_loc.cpu[:
num_reqs +
1]
query_start_loc_np = self.input_buffers.query_start_loc.np[:num_reqs +
1]
query_start_loc_gpu = self.input_buffers.query_start_loc.gpu[: num_reqs + 1]
query_start_loc_cpu = self.input_buffers.query_start_loc.cpu[: num_reqs + 1]
query_start_loc_np = self.input_buffers.query_start_loc.np[: num_reqs + 1]
# Get prefill tokens.
prepare_prefill_inputs(
@@ -249,7 +234,8 @@ class NPUModelRunner(GPUModelRunner):
# Compute slot mappings: [num_kv_cache_groups, num_tokens]
slot_mappings = self.block_tables.compute_slot_mappings(
query_start_loc_gpu, self.input_buffers.positions[:num_tokens])
query_start_loc_gpu, self.input_buffers.positions[:num_tokens]
)
# Layer name -> attention metadata.
# TODO(Ronald1995): try to add a new method `build_attn_metadata` in
@@ -263,8 +249,7 @@ class NPUModelRunner(GPUModelRunner):
query_start_loc_cpu=query_start_loc_cpu,
seq_lens=self.input_buffers.seq_lens,
seq_lens_np=self.input_buffers.seq_lens_np,
num_computed_tokens_cpu=self.req_states.
num_computed_tokens_cpu[idx_mapping_cpu],
num_computed_tokens_cpu=self.req_states.num_computed_tokens_cpu[idx_mapping_cpu],
block_tables=block_tables,
slot_mappings=slot_mappings,
kv_cache_config=self.kv_cache_config,
@@ -335,16 +320,13 @@ class NPUModelRunner(GPUModelRunner):
req_index = self.req_states.req_id_to_index[req_id]
# num_computed_tokens_cpu has reverted by num_rejected_tokens already.
# in super postprocess method.
self.req_states.num_computed_tokens_cpu[
req_index] = self.num_computed_tokens_cpu[req_index]
self.req_states.num_computed_tokens_cpu[req_index] = self.num_computed_tokens_cpu[req_index]
# update seq_lens_cpu
for i, req_id in enumerate(req_ids):
req_index = self.req_states.req_id_to_index[req_id]
num_computed_tokens = self.req_states.num_computed_tokens_cpu[
req_index]
self.input_buffers.seq_lens_cpu[
i] = num_computed_tokens + num_scheduled_tokens[req_id]
num_computed_tokens = self.req_states.num_computed_tokens_cpu[req_index]
self.input_buffers.seq_lens_cpu[i] = num_computed_tokens + num_scheduled_tokens[req_id]
def eplb_warmup(self):
# TODO(Ronald1995): just define the method in case calling error in

3
vllm_ascend/worker/v2/sample/gumbel.py
View File

@@ -76,8 +76,7 @@ def _gumbel_sample_kernel(
idx = tl.argmax(logits, axis=0)
token_id = block_idx * BLOCK_SIZE + idx
value = tl.max(logits, axis=0)
tl.store(local_argmax_ptr + req_idx * local_argmax_stride + block_idx,
token_id)
tl.store(local_argmax_ptr + req_idx * local_argmax_stride + block_idx, token_id)
tl.store(local_max_ptr + req_idx * local_max_stride + block_idx, value)

12
vllm_ascend/worker/v2/sample/penalties.py
View File

@@ -68,8 +68,7 @@ def _penalties_and_temperature_kernel(
if use_penalty:
req_state_idx = tl.load(idx_mapping_ptr + batch_idx)
output_bin_counts = tl.load(
output_bin_counts_ptr + req_state_idx * output_bin_counts_stride +
block,
output_bin_counts_ptr + req_state_idx * output_bin_counts_stride + block,
mask=mask,
)
# to use vector core, if use > 0 will use scalar to slow down performance
@@ -77,11 +76,9 @@ def _penalties_and_temperature_kernel(
# Apply repetition penalties.
if use_rep_penalty:
packed_block = block_idx * BLOCK_SIZE // 32 + tl.arange(
0, BLOCK_SIZE // 32)
packed_block = block_idx * BLOCK_SIZE // 32 + tl.arange(0, BLOCK_SIZE // 32)
packed_mask = tl.load(
prompt_bin_mask_ptr + req_state_idx * prompt_bin_mask_stride +
packed_block,
prompt_bin_mask_ptr + req_state_idx * prompt_bin_mask_stride + packed_block,
mask=packed_block < tl.cdiv(vocab_size, 32),
)
# the compiler itself does not optimize right-shift operations, so we change the same func
@@ -97,8 +94,7 @@ def _penalties_and_temperature_kernel(
prompt_bin_mask = prompt_bin_mask.reshape(BLOCK_SIZE)
# If token appears in prompt or output, apply, otherwise use 1.0 for no-op.
scale = tl.where(prompt_bin_mask | output_bin_mask, rep_penalty,
1.0)
scale = tl.where(prompt_bin_mask | output_bin_mask, rep_penalty, 1.0)
# If logits are positive, divide by penalty, otherwise multiply by penalty.
logits *= tl.where(logits > 0, 1.0 / scale, scale)

9
vllm_ascend/worker/v2/sample/sampler.py
View File

@@ -16,18 +16,16 @@
#
import torch
from vllm.v1.sample.ops.topk_topp_sampler import apply_top_k_top_p
from vllm.v1.sample.metadata import SamplingMetadata
from vllm.v1.sample.ops.topk_topp_sampler import apply_top_k_top_p
from vllm.v1.worker.gpu.sample.min_p import apply_min_p
from vllm.v1.worker.gpu.sample.sampler import Sampler
from vllm_ascend.worker.v2.sample.gumbel import gumbel_sample
from vllm_ascend.worker.v2.sample.penalties import \
apply_penalties_and_temperature
from vllm_ascend.worker.v2.sample.penalties import apply_penalties_and_temperature
class AscendSampler(Sampler):
def sample(
self,
logits: torch.Tensor,
@@ -45,8 +43,7 @@ class AscendSampler(Sampler):
if sampling_metadata.min_p is not None:
apply_min_p(logits, sampling_metadata.min_p)
# Apply top_k and/or top_p. This might return a new tensor.
logits = apply_top_k_top_p(logits, sampling_metadata.top_k,
sampling_metadata.top_p)
logits = apply_top_k_top_p(logits, sampling_metadata.top_k, sampling_metadata.top_p)
sampled = gumbel_sample(
logits,

6
vllm_ascend/worker/v2/spec_decode/__init__.py
View File

@@ -30,9 +30,7 @@ def init_speculator(
speculative_config = vllm_config.speculative_config
assert speculative_config is not None
if speculative_config.use_eagle():
from vllm_ascend.worker.v2.spec_decode.eagle import \
AscendEagleSpeculator
from vllm_ascend.worker.v2.spec_decode.eagle import AscendEagleSpeculator
return AscendEagleSpeculator(vllm_config, device)
raise NotImplementedError(
f"{speculative_config.method} is not supported yet.")
raise NotImplementedError(f"{speculative_config.method} is not supported yet.")

1
vllm_ascend/worker/v2/spec_decode/eagle.py
View File

@@ -30,7 +30,6 @@ from vllm_ascend.worker.v2.attn_utils import build_attn_metadata
class AscendEagleSpeculator(EagleSpeculator):
def __init__(self, vllm_config: VllmConfig, device: torch.device):
"""Override GPU EagleSpeculator.__init__ for Ascend NPUs.
attnention metadata building in Ascend backend needs more information,

6
vllm_ascend/worker/v2/states.py
View File

@@ -63,8 +63,8 @@ class AscendRequestState(RequestState):
# NOTE(Ronald1995): Ascend NPUs do not support UVA yet,
# so we use CpuGpuBuffer to allocate prefill_token_ids buffer.
self.prefill_token_ids: CpuGpuBuffer = self._make_buffer( # type: ignore
(self.max_num_reqs, self.max_model_len),
dtype=torch.int32)
(self.max_num_reqs, self.max_model_len), dtype=torch.int32
)
def add_request(
self,
@@ -75,7 +75,6 @@ class AscendRequestState(RequestState):
sampling_params,
lora_request,
):
super().add_request(
req_id,
prompt_len,
@@ -93,7 +92,6 @@ def uva_wrapper():
"""Context manager to disable UVA for Ascend NPUs."""
class UvaBufferWrapper:
def __init__(self, *args, **kwargs):
pass

243
vllm_ascend/worker/worker.py
View File

@@ -20,7 +20,6 @@
import copy
import gc
from types import NoneType
from typing import Optional
import torch
import torch.nn as nn
@@ -29,12 +28,9 @@ import vllm.envs as envs_vllm
from torch_npu.op_plugin.atb._atb_ops import _register_atb_extensions
from torch_npu.profiler import dynamic_profile as dp
from vllm.config import CUDAGraphMode, VllmConfig, set_current_vllm_config
from vllm.distributed import (ensure_model_parallel_initialized,
init_distributed_environment)
from vllm.distributed import ensure_model_parallel_initialized, init_distributed_environment
from vllm.distributed.ec_transfer import ensure_ec_transfer_initialized
from vllm.distributed.kv_transfer import (ensure_kv_transfer_initialized,
get_kv_transfer_group,
has_kv_transfer_group)
from vllm.distributed.kv_transfer import ensure_kv_transfer_initialized, get_kv_transfer_group, has_kv_transfer_group
from vllm.distributed.parallel_state import get_pp_group, get_tp_group
from vllm.logger import logger
from vllm.lora.request import LoRARequest
@@ -44,8 +40,7 @@ from vllm.utils.mem_constants import GiB_bytes
from vllm.utils.torch_utils import STR_DTYPE_TO_TORCH_DTYPE
from vllm.v1.core.sched.output import GrammarOutput, SchedulerOutput
from vllm.v1.kv_cache_interface import KVCacheConfig, KVCacheSpec
from vllm.v1.outputs import (EMPTY_MODEL_RUNNER_OUTPUT, AsyncModelRunnerOutput,
DraftTokenIds, ModelRunnerOutput)
from vllm.v1.outputs import EMPTY_MODEL_RUNNER_OUTPUT, AsyncModelRunnerOutput, DraftTokenIds, ModelRunnerOutput
from vllm.v1.worker.worker_base import WorkerBase
from vllm.v1.worker.workspace import init_workspace_manager
@@ -56,37 +51,38 @@ from vllm_ascend.cpu_binding import bind_cpus
from vllm_ascend.device_allocator.camem import CaMemAllocator
from vllm_ascend.distributed.parallel_state import init_ascend_model_parallel
from vllm_ascend.ops.triton.triton_utils import init_device_properties_triton
from vllm_ascend.utils import (AscendDeviceType, check_ascend_device_type,
enable_sp, get_ascend_device_type,
register_ascend_customop)
from vllm_ascend.utils import (
AscendDeviceType,
check_ascend_device_type,
enable_sp,
get_ascend_device_type,
register_ascend_customop,
)
from vllm_ascend.worker.model_runner_v1 import NPUModelRunner
torch._dynamo.trace_rules.clear_lru_cache() # noqa: E402
from torch._dynamo.variables import TorchInGraphFunctionVariable # noqa: E402
from vllm.utils.torch_utils import set_random_seed
from vllm.utils.torch_utils import set_random_seed # noqa: E402
torch_non_c_binding_in_graph_functions_npu = dict.fromkeys(
["torch.npu.current_stream"],
TorchInGraphFunctionVariable,
) # noqa: E402
torch_non_c_binding_in_graph_functions_npu[
"torch.npu.stream"] = TorchInGraphFunctionVariable # noqa: E402
torch._dynamo.trace_rules.torch_name_rule_map.append(
torch_non_c_binding_in_graph_functions_npu) # noqa: E402
torch_non_c_binding_in_graph_functions_npu["torch.npu.stream"] = TorchInGraphFunctionVariable # noqa: E402
torch._dynamo.trace_rules.torch_name_rule_map.append(torch_non_c_binding_in_graph_functions_npu) # noqa: E402
class NPUWorker(WorkerBase):
def __init__(
self,
vllm_config: VllmConfig,
local_rank: int,
rank: int,
distributed_init_method: str,
is_driver_worker: bool = False,
# Additional parameters for compatibility with vllm
**kwargs):
self,
vllm_config: VllmConfig,
local_rank: int,
rank: int,
distributed_init_method: str,
is_driver_worker: bool = False,
# Additional parameters for compatibility with vllm
**kwargs,
):
"""Initialize the worker for Ascend."""
if not envs_ascend.COMPILE_CUSTOM_KERNELS:
logger.warning(
@@ -96,14 +92,17 @@ class NPUWorker(WorkerBase):
# register patch for vllm
from vllm_ascend.utils import adapt_patch
adapt_patch()
# Import _inductor for graph mode execution with triton
# This lazy import avoids torch_npu re-initialization in patch
from vllm.triton_utils import HAS_TRITON
if HAS_TRITON:
import torch_npu._inductor # noqa: F401
# Register ops when worker init.
from vllm_ascend import ops
ops.register_dummy_fusion_op()
if get_ascend_device_type() != AscendDeviceType.A5:
_register_atb_extensions()
@@ -112,17 +111,18 @@ class NPUWorker(WorkerBase):
init_ascend_config(vllm_config)
check_ascend_device_type()
super().__init__(vllm_config=vllm_config,
local_rank=local_rank,
rank=rank,
distributed_init_method=distributed_init_method,
is_driver_worker=is_driver_worker)
super().__init__(
vllm_config=vllm_config,
local_rank=local_rank,
rank=rank,
distributed_init_method=distributed_init_method,
is_driver_worker=is_driver_worker,
)
if self.cache_config.cache_dtype == "auto":
self.cache_dtype = self.model_config.dtype
else:
self.cache_dtype = STR_DTYPE_TO_TORCH_DTYPE[
self.cache_config.cache_dtype]
self.cache_dtype = STR_DTYPE_TO_TORCH_DTYPE[self.cache_config.cache_dtype]
self.profiler = self._init_profiler()
if vllm_config.model_config and vllm_config.model_config.enable_sleep_mode:
@@ -130,8 +130,8 @@ class NPUWorker(WorkerBase):
self._sleep_saved_buffers: dict[str, torch.Tensor] = {}
# FixMe: this is a patch to fix the issue cause by https://github.com/vllm-project/vllm/commit/de94289a98d7ec52a5ef02719e01a1db8b505170
from vllm.model_executor.layers.linear import \
WEIGHT_LOADER_V2_SUPPORTED
from vllm.model_executor.layers.linear import WEIGHT_LOADER_V2_SUPPORTED
if "UnquantizedLinearMethod" in WEIGHT_LOADER_V2_SUPPORTED:
WEIGHT_LOADER_V2_SUPPORTED.remove("UnquantizedLinearMethod")
@@ -151,33 +151,33 @@ class NPUWorker(WorkerBase):
# Either SIGTERM or SIGINT will terminate the worker
import signal
signal.signal(signal.SIGTERM, signal_handler)
signal.signal(signal.SIGINT, signal_handler)
def uninstall_static_kernel(self):
import os
import fcntl
import os
import subprocess
ascend_home_path = os.environ["ASCEND_HOME_PATH"]
static_kernel_dir_path = os.path.join(ascend_home_path, 'opp/static_kernel')
uninstall_script_path = os.path.join(static_kernel_dir_path, 'ai_core/uninstall.sh')
lock_file_path = os.path.join(static_kernel_dir_path, 'uninstall.lock')
static_kernel_dir_path = os.path.join(ascend_home_path, "opp/static_kernel")
uninstall_script_path = os.path.join(static_kernel_dir_path, "ai_core/uninstall.sh")
lock_file_path = os.path.join(static_kernel_dir_path, "uninstall.lock")
if not os.path.exists(uninstall_script_path):
return
with open(lock_file_path, 'w') as lock_fd:
with open(lock_file_path, "w") as lock_fd:
try:
fcntl.flock(lock_fd, fcntl.LOCK_EX | fcntl.LOCK_NB)
subprocess.Popen(
['bash', uninstall_script_path],
["bash", uninstall_script_path],
stdin=subprocess.DEVNULL,
stdout=subprocess.DEVNULL,
stderr=subprocess.DEVNULL,
start_new_session=True
start_new_session=True,
)
except (BlockingIOError, OSError) as e:
except (BlockingIOError, OSError):
return
finally:
try:
@@ -187,32 +187,30 @@ class NPUWorker(WorkerBase):
except Exception:
return
def sleep(self, level: int = 1) -> None:
free_bytes_before_sleep = torch.npu.mem_get_info()[0]
# Save the buffers before level 2 sleep
if level == 2:
model = self.model_runner.model
self._sleep_saved_buffers = {
name: buffer.cpu().clone()
for name, buffer in model.named_buffers()
}
self._sleep_saved_buffers = {name: buffer.cpu().clone() for name, buffer in model.named_buffers()}
allocator = CaMemAllocator.get_instance()
allocator.sleep(offload_tags=("weights", ) if level == 1 else tuple())
allocator.sleep(offload_tags=("weights",) if level == 1 else tuple())
free_bytes_after_sleep, total = torch.npu.mem_get_info()
freed_bytes = free_bytes_after_sleep - free_bytes_before_sleep
used_bytes = total - free_bytes_after_sleep
assert freed_bytes >= 0, "Memory usage increased after sleeping."
logger.info(
"Sleep mode freed %.2f GiB memory, "
"%.2f GiB memory is still in use.", freed_bytes / GiB_bytes,
used_bytes / GiB_bytes)
"Sleep mode freed %.2f GiB memory, %.2f GiB memory is still in use.",
freed_bytes / GiB_bytes,
used_bytes / GiB_bytes,
)
def wake_up(self, tags: Optional[list[str]] = None) -> None:
def wake_up(self, tags: list[str] | None = None) -> None:
if envs_ascend.VLLM_ASCEND_ENABLE_NZ:
raise ValueError(
"FRACTAL_NZ mode is enabled. This may cause model parameter precision issues "
"in the RL scenarios. Please set VLLM_ASCEND_ENABLE_NZ=0.")
"in the RL scenarios. Please set VLLM_ASCEND_ENABLE_NZ=0."
)
allocator = CaMemAllocator.get_instance()
allocator.wake_up(tags=tags)
@@ -220,22 +218,21 @@ class NPUWorker(WorkerBase):
model = self.model_runner.model
if tags is None or "weights" in tags:
for name, param in model.named_parameters():
if 'w2_weight' in name and param.shape[2] == hidden_size:
parts = name.split('.')
if "w2_weight" in name and param.shape[2] == hidden_size:
parts = name.split(".")
param_name = parts[-1]
parent_module = model.get_submodule(".".join(parts[:-1]))
w2_data = param.transpose(1, 2)
w2_data = torch.nn.Parameter(w2_data, requires_grad=False)
setattr(parent_module, param_name, w2_data)
elif 'w13_weight' in name and param.shape[1] == hidden_size:
parts = name.split('.')
elif "w13_weight" in name and param.shape[1] == hidden_size:
parts = name.split(".")
param_name = parts[-1]
parent_module = model.get_submodule(".".join(parts[:-1]))
w13_data = param.transpose(1, 2)
w13_data = torch.nn.Parameter(w13_data,
requires_grad=False)
w13_data = torch.nn.Parameter(w13_data, requires_grad=False)
setattr(parent_module, param_name, w13_data)
# Restore the buffers after level 2 sleep
@@ -245,8 +242,7 @@ class NPUWorker(WorkerBase):
buffer.data.copy_(self._sleep_saved_buffers[name].data)
self._sleep_saved_buffers = {}
def initialize_cache(self, num_gpu_blocks: int,
num_cpu_blocks: int) -> None:
def initialize_cache(self, num_gpu_blocks: int, num_cpu_blocks: int) -> None:
self.cache_config.num_gpu_blocks = num_gpu_blocks
self.cache_config.num_cpu_blocks = num_cpu_blocks
@@ -255,18 +251,19 @@ class NPUWorker(WorkerBase):
torch.npu.set_device(device)
torch.npu.empty_cache()
if (self.parallel_config.data_parallel_size > 1
and self.parallel_config.data_parallel_size_local > 0
and self.parallel_config.distributed_executor_backend
not in ["ray", "external_launcher"] and
self.vllm_config.parallel_config.data_parallel_backend != "ray"
and self.vllm_config.parallel_config.nnodes_within_dp == 1):
visible_device_count = (torch.npu.device_count()
if torch.npu.is_available() else 0)
if (
self.parallel_config.data_parallel_size > 1
and self.parallel_config.data_parallel_size_local > 0
and self.parallel_config.distributed_executor_backend not in ["ray", "external_launcher"]
and self.vllm_config.parallel_config.data_parallel_backend != "ray"
and self.vllm_config.parallel_config.nnodes_within_dp == 1
):
visible_device_count = torch.npu.device_count() if torch.npu.is_available() else 0
assert self.parallel_config.local_world_size <= visible_device_count, (
f"local_world_size ({self.parallel_config.local_world_size}) must "
f"be less than or equal to the number of visible devices "
f"({visible_device_count}).")
f"({visible_device_count})."
)
self.init_npu_memory = torch.npu.mem_get_info()[0]
# Initialize the distributed environment.
@@ -281,9 +278,7 @@ class NPUWorker(WorkerBase):
try:
bind_cpus(self.local_rank)
except Exception as e:
logger.warning(
f"Bind cpus failed in rank{self.local_rank}: {e} Skip binding cpu."
)
logger.warning(f"Bind cpus failed in rank{self.local_rank}: {e} Skip binding cpu.")
return device
def init_device(self):
@@ -296,11 +291,9 @@ class NPUWorker(WorkerBase):
init_workspace_manager(self.device, num_ubatches)
# Init ModelRunner here, so that we have access to self.device.
if self.use_v2_model_runner:
logger.warning(
"npu model runner v2 is in developing, some features doesn't work for now."
)
from vllm_ascend.worker.v2.model_runner import \
NPUModelRunner as NPUModelRunnerV2
logger.warning("npu model runner v2 is in developing, some features doesn't work for now.")
from vllm_ascend.worker.v2.model_runner import NPUModelRunner as NPUModelRunnerV2
self.model_runner = NPUModelRunnerV2(self.vllm_config, self.device)
else:
self.model_runner = NPUModelRunner(self.vllm_config, self.device)
@@ -327,27 +320,22 @@ class NPUWorker(WorkerBase):
"Error in memory profiling. "
f"Initial free memory {self.init_npu_memory}, current free memory"
f" {free_npu_memory}. This happens when the NPU memory was "
"not properly cleaned up before initializing the vLLM instance.")
"not properly cleaned up before initializing the vLLM instance."
)
# Get the peak memory allocation recorded by torch
peak_memory = torch_npu.npu.memory_stats()["allocated_bytes.all.peak"]
# TODO: don`t need impl this func after empty_cache in
# Worker.determine_num_available_blocks() unified`
torch.npu.empty_cache()
torch_allocated_bytes = torch_npu.npu.memory_stats(
)["allocated_bytes.all.current"]
total_allocated_bytes = torch_npu.npu.mem_get_info(
)[1] - torch_npu.npu.mem_get_info()[0]
torch_allocated_bytes = torch_npu.npu.memory_stats()["allocated_bytes.all.current"]
total_allocated_bytes = torch_npu.npu.mem_get_info()[1] - torch_npu.npu.mem_get_info()[0]
non_torch_allocations = total_allocated_bytes - torch_allocated_bytes
if non_torch_allocations > 0:
peak_memory += non_torch_allocations
available_kv_cache_memory = int(
total_npu_memory * self.cache_config.gpu_memory_utilization -
peak_memory)
available_kv_cache_memory = int(total_npu_memory * self.cache_config.gpu_memory_utilization - peak_memory)
available_kv_cache_memory = int(max(available_kv_cache_memory, 0))
logger.info(
f"Available memory: {available_kv_cache_memory}, total memory: {total_npu_memory}"
)
logger.info(f"Available memory: {available_kv_cache_memory}, total memory: {total_npu_memory}")
return available_kv_cache_memory
def execute_model(
@@ -361,32 +349,30 @@ class NPUWorker(WorkerBase):
intermediate_tensors = None
forward_pass = scheduler_output.total_num_scheduled_tokens > 0
if forward_pass and not get_pp_group().is_first_rank:
# If flashcomm1 is used, this all_gather_group parameter needs to be removed, otherwise it will conflict with the all-gather operation in flashcomm1.
# If flashcomm1 is used, this all_gather_group parameter needs to be removed, otherwise
# it will conflict with the all-gather operation in flashcomm1.
if enable_sp():
all_gather_group = None
else:
all_gather_group = get_tp_group()
intermediate_tensors = IntermediateTensors(
get_pp_group().recv_tensor_dict(
all_gather_group=all_gather_group))
get_pp_group().recv_tensor_dict(all_gather_group=all_gather_group)
)
output = self.model_runner.execute_model(scheduler_output,
intermediate_tensors)
if isinstance(output,
(ModelRunnerOutput, AsyncModelRunnerOutput, NoneType)):
output = self.model_runner.execute_model(scheduler_output, intermediate_tensors)
if isinstance(output, (ModelRunnerOutput, AsyncModelRunnerOutput, NoneType)):
return output
assert isinstance(output, IntermediateTensors)
parallel_config = self.vllm_config.parallel_config
assert parallel_config.distributed_executor_backend != (
"external_launcher") and not get_pp_group().is_last_rank
# If flashcomm1 is used, this all_gather_group parameter needs to be removed, otherwise it will conflict with the all-gather operation in flashcomm1.
assert parallel_config.distributed_executor_backend != ("external_launcher") and not get_pp_group().is_last_rank
# If flashcomm1 is used, this all_gather_group parameter needs to be removed, otherwise
# it will conflict with the all-gather operation in flashcomm1.
if enable_sp():
all_gather_group = None
else:
all_gather_group = get_tp_group()
get_pp_group().send_tensor_dict(output.tensors,
all_gather_group=all_gather_group)
get_pp_group().send_tensor_dict(output.tensors, all_gather_group=all_gather_group)
kv_connector_output = output.kv_connector_output
if not kv_connector_output:
@@ -394,28 +380,24 @@ class NPUWorker(WorkerBase):
# In case of PP with kv transfer, we need to pass through the
# kv_connector_output
if (not kv_connector_output.finished_sending
and not kv_connector_output.finished_recving):
if not kv_connector_output.finished_sending and not kv_connector_output.finished_recving:
return EMPTY_MODEL_RUNNER_OUTPUT
output = copy.copy(EMPTY_MODEL_RUNNER_OUTPUT)
output.kv_connector_output = kv_connector_output
return output
@torch.inference_mode()
def sample_tokens(
self, grammar_output: "GrammarOutput"
) -> ModelRunnerOutput | AsyncModelRunnerOutput:
def sample_tokens(self, grammar_output: "GrammarOutput") -> ModelRunnerOutput | AsyncModelRunnerOutput:
return self.model_runner.sample_tokens(grammar_output)
def load_model(self) -> None:
if self.vllm_config.model_config.enable_sleep_mode:
allocator = CaMemAllocator.get_instance()
assert allocator.get_current_usage() == 0, (
"Sleep mode can only be "
"used for one instance per process.")
assert allocator.get_current_usage() == 0, "Sleep mode can only be used for one instance per process."
context = allocator.use_memory_pool(tag="weights")
else:
from contextlib import nullcontext
context = nullcontext() # type: ignore
with context, set_current_vllm_config(self.vllm_config):
@@ -423,19 +405,15 @@ class NPUWorker(WorkerBase):
def compile_or_warm_up_model(self) -> None:
# Note: need to adapt for graph mode.
warmup_sizes = (self.vllm_config.compilation_config.compile_sizes
or []).copy()
warmup_sizes = (self.vllm_config.compilation_config.compile_sizes or []).copy()
if not self.model_config.enforce_eager:
cg_capture_sizes: list[int] = []
if self.vllm_config.compilation_config.cudagraph_mode != CUDAGraphMode.NONE:
cg_sizes = self.vllm_config.compilation_config.cudagraph_capture_sizes
cg_capture_sizes = [] if cg_sizes is None else cg_sizes
warmup_sizes = [
x for x in warmup_sizes if x not in cg_capture_sizes
]
warmup_sizes = [x for x in warmup_sizes if x not in cg_capture_sizes]
compile_ranges = self.vllm_config.compilation_config.get_compile_ranges(
)
compile_ranges = self.vllm_config.compilation_config.get_compile_ranges()
# For each compile_range, if none of the batch sizes
# in warmup_sizes or cudagraph_capture_sizes are in the range,
# add the end of the range to ensure compilation/warmup.
@@ -467,7 +445,7 @@ class NPUWorker(WorkerBase):
def get_model(self) -> nn.Module:
return self.model_runner.get_model()
def get_kv_connector_handshake_metadata(self) -> Optional[dict]:
def get_kv_connector_handshake_metadata(self) -> dict | None:
"""Get KV connector metadata from this worker if available."""
if not has_kv_transfer_group():
return None
@@ -503,6 +481,7 @@ class NPUWorker(WorkerBase):
context = allocator.use_memory_pool(tag="kv_cache")
else:
from contextlib import nullcontext
context = nullcontext() # type: ignore
with context:
self.model_runner.initialize_kv_cache(kv_cache_config)
@@ -528,21 +507,20 @@ class NPUWorker(WorkerBase):
return self.model_runner.pin_lora(lora_id)
def execute_dummy_batch(self) -> None:
self.model_runner._dummy_run(
num_tokens=self.model_runner.decode_token_per_req,
uniform_decode=True)
self.model_runner._dummy_run(num_tokens=self.model_runner.decode_token_per_req, uniform_decode=True)
def _init_worker_distributed_environment(self) -> None:
"""Initialize the distributed environment."""
init_batch_invariance()
init_distributed_environment(self.parallel_config.world_size,
self.rank, self.distributed_init_method,
self.local_rank, "hccl")
init_distributed_environment(
self.parallel_config.world_size, self.rank, self.distributed_init_method, self.local_rank, "hccl"
)
ensure_model_parallel_initialized(
self.parallel_config.tensor_parallel_size,
self.parallel_config.pipeline_parallel_size,
self.parallel_config.prefill_context_parallel_size,
self.parallel_config.decode_context_parallel_size)
self.parallel_config.decode_context_parallel_size,
)
init_ascend_model_parallel(self.parallel_config)
ensure_kv_transfer_initialized(self.vllm_config)
ensure_ec_transfer_initialized(self.vllm_config)
@@ -553,12 +531,9 @@ class NPUWorker(WorkerBase):
profiler_config = self.vllm_config.profiler_config
if profiler_config.profiler == "torch" and profiler_config.torch_profiler_dir:
if envs_ascend.MSMONITOR_USE_DAEMON:
raise RuntimeError(
"MSMONITOR_USE_DAEMON and torch profiler cannot be both enabled at the same time."
)
raise RuntimeError("MSMONITOR_USE_DAEMON and torch profiler cannot be both enabled at the same time.")
torch_profiler_trace_dir = profiler_config.torch_profiler_dir
logger.info("Profiling enabled. Traces will be saved to: %s",
torch_profiler_trace_dir)
logger.info("Profiling enabled. Traces will be saved to: %s", torch_profiler_trace_dir)
experimental_config = torch_npu.profiler._ExperimentalConfig(
export_type=torch_npu.profiler.ExportType.Text,
@@ -583,8 +558,8 @@ class NPUWorker(WorkerBase):
# The with_stack option in torch_npu.profiler introduces significant time overhead.
with_modules=profiler_config.torch_profiler_with_stack,
experimental_config=experimental_config,
on_trace_ready=torch_npu.profiler.tensorboard_trace_handler(
torch_profiler_trace_dir))
on_trace_ready=torch_npu.profiler.tensorboard_trace_handler(torch_profiler_trace_dir),
)
else:
return None
@@ -594,5 +569,5 @@ class NPUWorker(WorkerBase):
def get_supported_tasks(self) -> "tuple[SupportedTask, ...]":
return self.model_runner.get_supported_tasks()
def take_draft_token_ids(self) -> Optional[DraftTokenIds]:
def take_draft_token_ids(self) -> DraftTokenIds | None:
return self.model_runner.take_draft_token_ids()

195
vllm_ascend/xlite/xlite.py
View File

@@ -14,49 +14,44 @@
# See the License for the specific language governing permissions and
# limitations under the License.
#
from typing import Any, Callable, Tuple
from collections.abc import Callable
from typing import Any
import torch
import torch.nn as nn
from vllm.config import VllmConfig
from vllm.distributed import (get_ep_group,
get_tensor_model_parallel_world_size,
get_world_group)
from vllm.distributed import get_ep_group, get_tensor_model_parallel_world_size, get_world_group
from vllm.forward_context import get_forward_context
from vllm.logger import logger
from vllm.sequence import IntermediateTensors
from xlite._C import (AttnMHA, Model, ModelAttnMeta, ModelConfig, Runtime, # type: ignore[attr-defined]
ScoringFuncSoftmax)
from xlite._C import ( # type: ignore[attr-defined]
AttnMHA,
Model,
ModelAttnMeta,
ModelConfig,
Runtime,
ScoringFuncSoftmax,
)
import vllm_ascend.envs as envs_ascend
from vllm_ascend.ascend_config import get_ascend_config
from vllm_ascend.attention.attention_v1 import (AscendAttentionState,
AscendMetadata)
from vllm_ascend.attention.attention_v1 import AscendAttentionState, AscendMetadata
class XliteModel:
def initialize(
self, runnable: nn.Module,
vllm_config: VllmConfig) -> Tuple[Model, int, int, torch.dtype]:
raise NotImplementedError(
"Xlite Model initialize function not implemented.")
def initialize(self, runnable: nn.Module, vllm_config: VllmConfig) -> tuple[Model, int, int, torch.dtype]:
raise NotImplementedError("Xlite Model initialize function not implemented.")
class LlamaXliteModel(XliteModel):
def initialize(
self, runnable: nn.Module,
vllm_config: VllmConfig) -> Tuple[Model, int, int, torch.dtype]:
def initialize(self, runnable: nn.Module, vllm_config: VllmConfig) -> tuple[Model, int, int, torch.dtype]:
dtype = vllm_config.model_config.dtype
config = self._build_model_config(vllm_config)
xlite_model = self._build_model(runnable, vllm_config, config)
rank = torch.distributed.get_rank()
xlite_model.init(config, rank)
freq_cis = self._precompute_freqs_cis(config.head_dim,
config.max_seq_len, dtype,
config.rope_theta)
freq_cis = self._precompute_freqs_cis(config.head_dim, config.max_seq_len, dtype, config.rope_theta)
return (xlite_model, freq_cis, config.hidden_size, dtype)
@@ -96,8 +91,7 @@ class LlamaXliteModel(XliteModel):
config.block_size = vllm_config.cache_config.block_size
return config
def _build_model(self, runnable: nn.Module, vllm_config: VllmConfig,
config: ModelConfig) -> Model:
def _build_model(self, runnable: nn.Module, vllm_config: VllmConfig, config: ModelConfig) -> Model:
params_dict = dict(runnable.named_parameters())
if hasattr(runnable, "language_model"):
@@ -108,48 +102,33 @@ class LlamaXliteModel(XliteModel):
model_prefix = ""
xlite_model = Model()
xlite_model.embed = params_dict.get(model_prefix +
"model.embed_tokens.weight")
xlite_model.embed = params_dict.get(model_prefix + "model.embed_tokens.weight")
xlite_model.norm = params_dict.get(model_prefix + "model.norm.weight")
if vllm_config.model_config.hf_text_config.tie_word_embeddings:
xlite_model.head = xlite_model.embed
else:
xlite_model.head = params_dict.get(model_prefix + "lm_head.weight")
xlite_model.attn_norm = [
layer.input_layernorm.weight for layer in layers
]
xlite_model.attn_out = [
layer.self_attn.o_proj.weight for layer in layers
]
xlite_model.mha_qkv = [
layer.self_attn.qkv_proj.weight for layer in layers
]
xlite_model.mlp_norm = [
layer.post_attention_layernorm.weight for layer in layers
]
xlite_model.attn_norm = [layer.input_layernorm.weight for layer in layers]
xlite_model.attn_out = [layer.self_attn.o_proj.weight for layer in layers]
xlite_model.mha_qkv = [layer.self_attn.qkv_proj.weight for layer in layers]
xlite_model.mlp_norm = [layer.post_attention_layernorm.weight for layer in layers]
xlite_model.mlp_up_gate = [
layer.mlp.gate_up_proj.weight for layer in layers
if hasattr(layer.mlp, "gate_up_proj")
and layer.mlp.gate_up_proj.weight is not None
layer.mlp.gate_up_proj.weight
for layer in layers
if hasattr(layer.mlp, "gate_up_proj") and layer.mlp.gate_up_proj.weight is not None
]
xlite_model.mlp_down = [
layer.mlp.down_proj.weight for layer in layers
if hasattr(layer.mlp, "down_proj")
and layer.mlp.down_proj.weight is not None
layer.mlp.down_proj.weight
for layer in layers
if hasattr(layer.mlp, "down_proj") and layer.mlp.down_proj.weight is not None
]
mha_qkv_bias = [
layer.self_attn.qkv_proj.bias for layer in layers
if hasattr(layer.self_attn.qkv_proj, "bias")
and layer.self_attn.qkv_proj.bias is not None
]
q_norm = [
layer.self_attn.q_norm.weight for layer in layers
if hasattr(layer.self_attn, "q_norm")
]
k_norm = [
layer.self_attn.k_norm.weight for layer in layers
if hasattr(layer.self_attn, "k_norm")
layer.self_attn.qkv_proj.bias
for layer in layers
if hasattr(layer.self_attn.qkv_proj, "bias") and layer.self_attn.qkv_proj.bias is not None
]
q_norm = [layer.self_attn.q_norm.weight for layer in layers if hasattr(layer.self_attn, "q_norm")]
k_norm = [layer.self_attn.k_norm.weight for layer in layers if hasattr(layer.self_attn, "k_norm")]
if len(mha_qkv_bias) != config.n_layers:
config.qkv_bias = False
@@ -157,7 +136,7 @@ class LlamaXliteModel(XliteModel):
config.qkv_bias = True
xlite_model.mha_qkv_bias = mha_qkv_bias
if (len(q_norm) != config.n_layers or len(k_norm) != config.n_layers):
if len(q_norm) != config.n_layers or len(k_norm) != config.n_layers:
config.qk_norm = False
else:
config.qk_norm = True
@@ -166,39 +145,28 @@ class LlamaXliteModel(XliteModel):
return xlite_model
def _precompute_freqs_cis(self,
dim: int,
end: int,
dtype: torch.dtype,
theta: float = 10000.0):
freqs = 1.0 / (theta**(torch.arange(
0, dim, 2, dtype=torch.float32, device='cpu')[:(dim // 2)] / dim))
def _precompute_freqs_cis(self, dim: int, end: int, dtype: torch.dtype, theta: float = 10000.0):
freqs = 1.0 / (theta ** (torch.arange(0, dim, 2, dtype=torch.float32, device="cpu")[: (dim // 2)] / dim))
t = torch.arange(end, device=freqs.device) # type: ignore
freqs = torch.outer(t, freqs).float() # type: ignore
cos_cache = freqs.cos().to(dtype)
sin_cache = freqs.sin().to(dtype)
freq_cis = torch.cat((cos_cache, sin_cache), dim=-1)
return freq_cis.to(device='npu')
return freq_cis.to(device="npu")
class QwenMoeXliteModel(LlamaXliteModel):
def initialize(
self, runnable: nn.Module,
vllm_config: VllmConfig) -> Tuple[Model, int, int, torch.dtype]:
def initialize(self, runnable: nn.Module, vllm_config: VllmConfig) -> tuple[Model, int, int, torch.dtype]:
if envs_ascend.VLLM_ASCEND_ENABLE_NZ == 2:
architecture = vllm_config.model_config.architectures[0]
raise ValueError(
f"{architecture} not support VLLM_ASCEND_ENABLE_NZ = 2!")
raise ValueError(f"{architecture} not support VLLM_ASCEND_ENABLE_NZ = 2!")
dtype = vllm_config.model_config.dtype
config = self._build_model_config(vllm_config)
xlite_model = self._build_model(runnable, vllm_config, config)
rank = torch.distributed.get_rank()
xlite_model.init(config, rank)
freq_cis = super()._precompute_freqs_cis(config.head_dim,
config.max_seq_len, dtype,
config.rope_theta)
freq_cis = super()._precompute_freqs_cis(config.head_dim, config.max_seq_len, dtype, config.rope_theta)
return (xlite_model, freq_cis, config.hidden_size, dtype)
@@ -214,32 +182,27 @@ class QwenMoeXliteModel(LlamaXliteModel):
config.def_dp_size = vllm_config.parallel_config.data_parallel_size
config.moe_ep_size = ep_group.world_size if vllm_config.parallel_config.enable_expert_parallel else 1
config.moe_tp_size = 1 if vllm_config.parallel_config.enable_expert_parallel else ep_group.world_size
config.experts_weight_transpose = True # type: ignore
config.experts_weight_transpose = True # type: ignore
config.moe_intermediate_size = hf_config.moe_intermediate_size
config.norm_topk_prob = hf_config.norm_topk_prob # type: ignore
config.scoring_func = ScoringFuncSoftmax # type: ignore
config.norm_topk_prob = hf_config.norm_topk_prob # type: ignore
config.scoring_func = ScoringFuncSoftmax # type: ignore
return config
def _build_model(self, runnable: nn.Module, vllm_config: VllmConfig,
config: ModelConfig) -> Model:
def _build_model(self, runnable: nn.Module, vllm_config: VllmConfig, config: ModelConfig) -> Model:
xlite_model = super()._build_model(runnable, vllm_config, config)
layers = runnable.model.layers
xlite_model.gate = [layer.mlp.gate.weight for layer in layers]
xlite_model.re_up_gate = [
layer.mlp.experts.w13_weight[i] for layer in layers
for i in range(layer.mlp.experts.local_num_experts)
layer.mlp.experts.w13_weight[i] for layer in layers for i in range(layer.mlp.experts.local_num_experts)
]
xlite_model.re_down = [
layer.mlp.experts.w2_weight[i] for layer in layers
for i in range(layer.mlp.experts.local_num_experts)
layer.mlp.experts.w2_weight[i] for layer in layers for i in range(layer.mlp.experts.local_num_experts)
]
return xlite_model
def xlite_model_init(
runnable: nn.Module,
vllm_config: VllmConfig) -> Tuple[Model, int, int, torch.dtype]:
def xlite_model_init(runnable: nn.Module, vllm_config: VllmConfig) -> tuple[Model, int, int, torch.dtype]:
strategy_map = {
"LlamaForCausalLM": LlamaXliteModel,
"Qwen2ForCausalLM": LlamaXliteModel,
@@ -266,33 +229,26 @@ class XliteWrapper:
rank = torch.distributed.get_rank()
local_rank = get_world_group().local_rank
self.xlite_rt = Runtime(local_rank, 0, rank,
get_tensor_model_parallel_world_size(),
vllm_config.parallel_config.data_parallel_size)
self.xlite_rt = Runtime(
local_rank, 0, rank, get_tensor_model_parallel_world_size(), vllm_config.parallel_config.data_parallel_size
)
(self.xlite_model, self.freq_cis, hidden_size,
dtype) = xlite_model_init(runnable, vllm_config)
(self.xlite_model, self.freq_cis, hidden_size, dtype) = xlite_model_init(runnable, vllm_config)
rt_pool_size = self.xlite_model.get_tensor_pool_size()
if rank == 0:
logger.info(f"xlite runtime pool size: {rt_pool_size} MB")
if self.xlite_rt.init_tensor_pool(rt_pool_size) != 0:
raise ValueError(
f"xlite wrapper init failed! runtime pool size: {rt_pool_size} MB"
)
raise ValueError(f"xlite wrapper init failed! runtime pool size: {rt_pool_size} MB")
max_num_tokens = vllm_config.scheduler_config.max_num_batched_tokens
self.hidden_states = torch.empty(max_num_tokens,
hidden_size,
device=f"npu:{local_rank}",
dtype=dtype)
self.hidden_states = torch.empty(max_num_tokens, hidden_size, device=f"npu:{local_rank}", dtype=dtype)
def __getattr__(self, key: str):
# allow accessing the attributes of the runnable.
if hasattr(self.runnable, key):
return getattr(self.runnable, key)
raise AttributeError(f"Attribute {key} not exists in the runnable of "
f"xlite wrapper: {self.runnable}")
raise AttributeError(f"Attribute {key} not exists in the runnable of xlite wrapper: {self.runnable}")
def unwrap(self) -> Callable:
# in case we need to access the original runnable.
@@ -307,22 +263,19 @@ class XliteWrapper:
positions: torch.Tensor,
intermediate_tensors: IntermediateTensors | None = None,
inputs_embeds: torch.Tensor | None = None,
) -> torch.Tensor | IntermediateTensors | tuple[torch.Tensor,
list[torch.Tensor]]:
) -> torch.Tensor | IntermediateTensors | tuple[torch.Tensor, list[torch.Tensor]]:
forward_context = get_forward_context()
attn_metadata: Any = forward_context.attn_metadata
if attn_metadata is None:
return self.runnable(input_ids, positions, intermediate_tensors,
inputs_embeds)
return self.runnable(input_ids, positions, intermediate_tensors, inputs_embeds)
attn_metadata = next(iter(attn_metadata.values()), None)
if attn_metadata is None or not isinstance(attn_metadata,
AscendMetadata):
return self.runnable(input_ids, positions, intermediate_tensors,
inputs_embeds)
if attn_metadata is None or not isinstance(attn_metadata, AscendMetadata):
return self.runnable(input_ids, positions, intermediate_tensors, inputs_embeds)
with_prefill = attn_metadata.attn_state not in [
AscendAttentionState.DecodeOnly, AscendAttentionState.SpecDecoding
AscendAttentionState.DecodeOnly,
AscendAttentionState.SpecDecoding,
]
if not with_prefill or self.full_mode:
@@ -335,11 +288,7 @@ class XliteWrapper:
num_prefills = attn_metadata.num_prefills
batch = num_prefills + num_decodes
seq_lens = attn_metadata.seq_lens[:batch]
seq_tensor = torch.cat([
torch.tensor([0]),
torch.tensor(attn_metadata.actual_seq_lengths_q)
],
dim=0)
seq_tensor = torch.cat([torch.tensor([0]), torch.tensor(attn_metadata.actual_seq_lengths_q)], dim=0)
query_lens = seq_tensor[1:] - seq_tensor[:-1]
query_lens = query_lens[:batch]
cached_lens = seq_lens - query_lens
@@ -347,23 +296,19 @@ class XliteWrapper:
xlite_attn_metadata = ModelAttnMeta()
xlite_attn_metadata.lens = query_lens.tolist()
xlite_attn_metadata.cached_lens = cached_lens.tolist()
xlite_attn_metadata.is_prefills = [False] * num_decodes + [
True
] * num_prefills
xlite_attn_metadata.block_tables = attn_metadata.block_tables.cpu(
).tolist()
xlite_attn_metadata.is_prefills = [False] * num_decodes + [True] * num_prefills
xlite_attn_metadata.block_tables = attn_metadata.block_tables.cpu().tolist()
h = self.hidden_states[:attn_metadata.num_actual_tokens]
h = self.hidden_states[: attn_metadata.num_actual_tokens]
stream = torch.npu.current_stream().npu_stream
if inputs_embeds is None:
self.xlite_model.forward(self.xlite_rt, input_ids,
xlite_attn_metadata, self.kv_caches,
self.freq_cis, h, stream)
self.xlite_model.forward(
self.xlite_rt, input_ids, xlite_attn_metadata, self.kv_caches, self.freq_cis, h, stream
)
else:
self.xlite_model.forward_with_inputs_embeds(
self.xlite_rt, inputs_embeds, xlite_attn_metadata,
self.kv_caches, self.freq_cis, h, stream)
self.xlite_rt, inputs_embeds, xlite_attn_metadata, self.kv_caches, self.freq_cis, h, stream
)
return h
else:
return self.runnable(input_ids, positions, intermediate_tensors,
inputs_embeds)
return self.runnable(input_ids, positions, intermediate_tensors, inputs_embeds)

2
vllm_ascend/xlite/xlite_model_runner.py
View File

@@ -22,13 +22,13 @@ from vllm_ascend.worker.model_runner_v1 import NPUModelRunner
class XliteModelRunner(NPUModelRunner):
def get_model(self) -> nn.Module:
return self.model.unwrap()
def load_model(self) -> None:
super().load_model()
from vllm_ascend.xlite.xlite import XliteWrapper
self.model = XliteWrapper(self.model, self.vllm_config)
def initialize_kv_cache(self, kv_cache_config: KVCacheConfig) -> None: