EngineX/xc-llm-ascend
3
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity
Files
460ea88276d743d6036efee2d30a8d7d02668878
xc-llm-ascend/vllm_ascend/quantization/methods/w4a8.py
Cao Yi a69ef10c3a [Refactor] Quantization Module Refactor (#5738) 
### Summary

This PR refactors the `vllm_ascend/quantization` module to improve code
organization, maintainability, and extensibility. The refactoring
introduces a clear separation of concerns with a registry-based scheme
discovery pattern, abstract base classes for quantization schemes, and
dedicated wrapper classes.

### Key Changes

#### 1. **Modular Directory Structure**

| Before | After |
|--------|-------|
| Flat file structure with mixed responsibilities | Organized into
`methods/` subpackage for schemes |
| Single `quant_config.py` (600+ lines) | Separate config files:
`modelslim_config.py`, `compressed_tensors_config.py` |
| `utils.py` with scheme lookup logic | `methods/registry.py` with
decorator-based registration |

#### 2. **Registry-Based Scheme Discovery**

Replaced hardcoded `ASCEND_QUANTIZATION_METHOD_MAP` dictionary with a
decorator-based registry pattern:

```python
# Before: Manual dictionary mapping
ASCEND_QUANTIZATION_METHOD_MAP = {
    "W8A8_DYNAMIC": {"linear": AscendW8A8DynamicLinearMethod, ...},
    ...
}

# After: Decorator-based registration
@register_scheme("W8A8_DYNAMIC", "linear")
class AscendW8A8DynamicLinearMethod(AscendLinearScheme):
    ...
```

#### 3. **Abstract Base Classes**

Introduced three abstract base classes in `methods/base.py`:
- `AscendLinearScheme` - Base for linear layer quantization
- `AscendMoEScheme` - Base for MoE layer quantization  
- `AscendAttentionScheme` - Base for attention layer quantization

#### 4. **Separated Config and Wrapper Classes**

- **Config classes** (`AscendModelSlimConfig`,
`AscendCompressedTensorsConfig`): Handle config parsing and scheme
selection
- **Wrapper classes** (`AscendLinearMethod`, `AscendFusedMoEMethod`,
etc.): Implement vLLM interfaces and delegate to schemes

#### 5. **Cleaner Public API**

```python
# New clean module interface
from vllm_ascend.quantization import (
    AscendModelSlimConfig,
    AscendCompressedTensorsConfig,
)
from vllm_ascend.quantization.methods import get_scheme_class
```

### Architecture Diagram

```mermaid
classDiagram
    direction TB
    
    class QuantizationConfig {
        <<vLLM Interface>>
        +get_quant_method()
    }
    
    class AscendModelSlimConfig {
        +quant_description
        +get_quant_method()
        -create_scheme_for_layer()
    }
    
    class AscendCompressedTensorsConfig {
        +target_scheme_map
        +get_quant_method()
        -_get_scheme_from_parts()
    }
    
    class AscendLinearMethod {
        <<Wrapper>>
        +quant_method: AscendLinearScheme
        +create_weights()
        +apply()
    }
    
    class AscendFusedMoEMethod {
        <<Wrapper>>
        +quant_method: AscendMoEScheme
        +create_weights()
        +apply()
    }
    
    class AscendLinearScheme {
        <<Abstract>>
        +get_weight()*
        +apply()*
        +get_pertensor_param()
        +get_perchannel_param()
    }
    
    class AscendMoEScheme {
        <<Abstract>>
        +get_weight()*
        +get_dynamic_quant_param()*
        +apply()*
    }
    
    class W8A8DynamicLinear {
        +get_weight()
        +apply()
    }
    
    class W8A8DynamicMoE {
        +get_weight()
        +apply()
    }
    
    QuantizationConfig <|-- AscendModelSlimConfig
    QuantizationConfig <|-- AscendCompressedTensorsConfig
    
    AscendModelSlimConfig ..> AscendLinearMethod : creates
    AscendModelSlimConfig ..> AscendFusedMoEMethod : creates
    AscendCompressedTensorsConfig ..> AscendLinearMethod : creates
    AscendCompressedTensorsConfig ..> AscendFusedMoEMethod : creates
    
    AscendLinearMethod o-- AscendLinearScheme : delegates to
    AscendFusedMoEMethod o-- AscendMoEScheme : delegates to
    
    AscendLinearScheme <|-- W8A8DynamicLinear
    AscendMoEScheme <|-- W8A8DynamicMoE
```

### Scheme Registration Flow

```mermaid
sequenceDiagram
    participant Module as Scheme Module
    participant Registry as _SCHEME_REGISTRY
    participant Config as QuantConfig
    participant Wrapper as Wrapper Class
    
    Note over Module: At import time
    Module->>Registry: @register_scheme("W8A8_DYNAMIC", "linear")
    Registry->>Registry: Store (quant_type, layer_type) -> Class
    
    Note over Config: At runtime
    Config->>Config: Determine quant_type from description
    Config->>Registry: get_scheme_class(quant_type, layer_type)
    Registry-->>Config: Return scheme class
    Config->>Config: scheme = scheme_cls()
    Config->>Wrapper: Create wrapper with scheme
    Wrapper-->>Config: Return wrapper instance
```

### File Changes Summary

| Original Files | Refactored Files |
|----------------|------------------|
| `__init__.py` (empty) | `__init__.py` (exports public API) |
| `quant_config.py` | `modelslim_config.py` + `wrappers.py` |
| `compressed_tensors/` | `compressed_tensors_config.py` |
| `utils.py` | `methods/registry.py` |
| `w8a8_dynamic.py` | `methods/w8a8_dynamic.py` |
| `w8a8.py` | `methods/w8a8_static.py` |
| `w4a4_flatquant_dynamic.py` | `methods/w4a4_flatquant.py` |
| ... | `methods/base.py` (new) |

### Benefits

1. **Extensibility**: Adding new quantization schemes only requires
implementing the base class and adding `@register_scheme` decorator
2. **Maintainability**: Clear separation between config parsing, wrapper
logic, and scheme implementation
3. **Testability**: Abstract base classes enable easier unit testing and
mocking
4. **Discoverability**: Registry pattern makes it easy to list all
supported schemes
5. **Reduced Coupling**: Config classes no longer need to know about all
scheme implementations

___

- vLLM version: v0.13.0
- vLLM main:
2f4e6548ef

---------

Signed-off-by: SlightwindSec <slightwindsec@gmail.com>
2026-01-23 14:13:47 +08:00

476 lines
21 KiB
Python
Raw Blame History

#
# Copyright (c) 2025 Huawei Technologies Co., Ltd. All Rights Reserved.
# This file is a part of the vllm-ascend project.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# 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 typing import Any, Callable, Dict, Optional
import numpy as np
import torch
import torch_npu
from vllm.config import get_current_vllm_config
from vllm.distributed import get_ep_group
from vllm.forward_context import get_forward_context
from vllm_ascend.ascend_config import get_ascend_config
from vllm_ascend.distributed.parallel_state import get_mc2_group
from vllm_ascend.ops.fused_moe.experts_selector import select_experts
from vllm_ascend.utils import maybe_trans_nz
from .base import AscendLinearScheme, AscendMoEScheme, QuantType
from .registry import register_scheme
@register_scheme("W4A8_DYNAMIC", "linear")
class AscendW4A8DynamicLinearMethod(AscendLinearScheme):
"""Linear method for Ascend W4A8_DYNAMIC."""
def __init__(self):
vllm_config = get_current_vllm_config()
self.group_size = vllm_config.quant_config.quant_description.get(
"group_size", 256)
quant_version = vllm_config.quant_config.quant_description.get(
"version", "0")
self.new_quant_version = quant_version == "1.0.0"
from vllm.distributed import get_tensor_model_parallel_world_size
self.tp_size = get_tensor_model_parallel_world_size()
def get_weight(self, input_size: int, output_size: int,
params_dtype: torch.dtype) -> Dict[str, Any]:
"""Create weight parameters.
For new quantization version (double int4 pack into int8), the output dimension
is compressed by factor 2 (e.g., [2048, 3072] -> [1024, 3072]). The returned
dict includes "_packed_dim" and "_packed_factor" for vLLM's weight loader.
"""
params_dict = {}
if self.new_quant_version:
# double int4 pack into int8: output dimension is compressed
pack_factor = 2
actual_output_size = output_size // pack_factor
params_dict["weight"] = torch.empty(actual_output_size,
input_size,
dtype=torch.int8)
# Add packing information for vLLM's weight_loader
params_dict["_packed_dim"] = 0
params_dict["_packed_factor"] = pack_factor
else:
params_dict["weight"] = torch.empty(output_size,
input_size,
dtype=torch.int8)
return params_dict
def get_pergroup_param(self,
input_size: int,
output_size: int,
params_dtype: torch.dtype,
layer_type: Optional[str] = None) -> Dict[str, Any]:
"""Create per-group quantization parameters."""
params_dict = {}
params_dict["weight_scale"] = torch.empty(output_size,
1,
dtype=params_dtype)
params_dict["weight_offset"] = torch.empty(output_size,
1,
dtype=params_dtype)
params_dict["weight_scale_second"] = torch.empty(output_size,
input_size //
self.group_size,
dtype=params_dtype)
params_dict["weight_offset_second"] = torch.empty(output_size,
input_size //
self.group_size,
dtype=params_dtype)
# NOTE: In w4a8 quantization implementation,
# for down_proj and o_proj(layer_type == "row") scale_bias shape is [output_size, 16],
# others are [output_size, 1]
if self.new_quant_version:
scale_bias_dim = 16 if layer_type == "row" else 1
params_dict["scale_bias"] = torch.empty(output_size,
scale_bias_dim,
dtype=torch.float32)
return params_dict
@staticmethod
def process_scale_second(weight: torch.Tensor,
scale: torch.Tensor,
per_group_scale: torch.Tensor,
is_new_quant: bool = False):
"""Process the scale for second-level quantization.
Args:
weight: weight tensor [k, n] (in new version, n is already compressed to n/2)
scale: first-level quantization scale [output_size]
per_group_scale: second-level per-group quantization scale [group_num, n_scale]
is_new_quant: whether it's the new quantization version (weight already compressed)
Returns:
(antiquant_scale, bias): dequantization scale and bias (bias=None for new version)
"""
k, n = weight.shape
group_num, n_scale = per_group_scale.shape
if is_new_quant:
# Restore logical dimension for compressed weight
n = n * 2
bias = None
if not is_new_quant:
weight_high = weight.to(torch.float32).reshape(
group_num, -1, n) * per_group_scale.reshape(group_num, 1, n)
weight_high = weight_high.reshape(k, n)
bias = 8 * (weight_high.to(torch.float32) * scale).sum(dim=0)
# NOTE: scale_bias is not used currently
# because in msmodelslim w4a8 uses symmetric quantization
# TODO: support potential future asymmetric quantization
antiquant_scale = (scale * per_group_scale).reshape(group_num, n)
return antiquant_scale.npu(), bias
def apply(
self,
layer: torch.nn.Module,
x: torch.Tensor,
bias: Optional[torch.Tensor] = None,
tp_rank: Optional[int] = None,
) -> torch.Tensor:
return torch_npu.npu_weight_quant_batchmatmul(
x,
layer.weight,
antiquant_scale=layer.weight_scale_second.to(x.dtype),
antiquant_group_size=self.group_size,
)
def process_weights_after_loading(self, layer: torch.nn.Module):
layer.weight.data = layer.weight.data.transpose(0, 1).contiguous()
layer.weight.data = maybe_trans_nz(layer.weight.data)
layer.weight_scale.data = layer.weight_scale.data.flatten().to(
torch.float32)
layer.weight_offset.data = layer.weight_offset.data.flatten()
layer.weight_scale_second.data, scale_bias = self.process_scale_second(
layer.weight.data,
layer.weight_scale.data,
layer.weight_scale_second.data.transpose(0, 1).contiguous(),
is_new_quant=self.new_quant_version,
)
if self.new_quant_version:
# Process the loaded data based on layer type
if hasattr(layer, "scale_bias"):
if layer.scale_bias.data.shape[1] == 1:
layer.scale_bias.data = layer.scale_bias.data.flatten()
else:
layer.scale_bias.data = layer.scale_bias.data.contiguous()
else:
if scale_bias is not None:
param = torch.nn.Parameter(scale_bias, requires_grad=False)
layer.register_parameter("weight_scale_bias", param)
# Convert to NPU-specific int4pack format
if self.new_quant_version:
# weights on disk are already in packed int4 format
# pack 4 int8(int4*2) to int32
assert layer.weight.data.shape[-1] % 4 == 0, \
f"the last dim of weight needs to be divided by 4, got shape {layer.weight.data.shape}"
layer.weight.data = layer.weight.data.view(
torch.int32).contiguous()
else:
# weights are not compressed
# need to be packed via npu_convert_weight_to_int4pack
layer.weight.data = torch_npu.npu_convert_weight_to_int4pack(
layer.weight.data.to(torch.int32))
@register_scheme("W4A8_DYNAMIC", "moe")
class AscendW4A8DynamicFusedMoEMethod(AscendMoEScheme):
"""FusedMoE method for Ascend W4A8_DYNAMIC."""
# Declare the quantization type for this scheme
quant_type: QuantType = QuantType.W4A8
def __init__(self):
self.ep_group = get_ep_group()
vllm_config = get_current_vllm_config()
self.group_size = vllm_config.quant_config.quant_description.get(
"group_size", 256)
# NOTE: the weights are quantized from bf16 to int4 through a per-channel quantization process
self.is_per_channel_weight = self.group_size == 0
quant_version = vllm_config.quant_config.quant_description.get(
"version", "0")
# NOTE: new quantize weights: 2 int4 pack into int8
self.new_quant_version = quant_version == "1.0.0"
self.tp_size = 1 if vllm_config.parallel_config.enable_expert_parallel else self.ep_group.world_size
self.dynamic_eplb = get_ascend_config().eplb_config.dynamic_eplb
if self.new_quant_version and self.tp_size > 16:
raise ValueError(
"The current weight does not support moe part tp>16.")
try:
device_group = get_mc2_group().device_group
# TODO: Try local_rank = ep_group.rank_in_group
local_rank = torch.distributed.get_rank(group=device_group)
backend = device_group._get_backend(torch.device("npu"))
self.moe_all_to_all_group_name = backend.get_hccl_comm_name(
local_rank)
except AttributeError:
self.moe_all_to_all_group_name = ""
def get_weight(self, num_experts: int,
intermediate_size_per_partition: int, hidden_sizes: int,
params_dtype: torch.dtype) -> Dict[str, Any]:
param_dict = {}
if self.new_quant_version:
w13_output_size = intermediate_size_per_partition
w2_output_size = hidden_sizes // 2
else:
w13_output_size = 2 * intermediate_size_per_partition
w2_output_size = hidden_sizes
param_dict["w13_weight"] = torch.empty(num_experts,
w13_output_size,
hidden_sizes,
dtype=torch.int8)
param_dict["w2_weight"] = torch.empty(num_experts,
w2_output_size,
intermediate_size_per_partition,
dtype=torch.int8)
return param_dict
def get_dynamic_quant_param(self, num_experts: int,
intermediate_size_per_partition: int,
hidden_sizes: int,
params_dtype: torch.dtype) -> Dict[str, Any]:
param_dict = {}
param_dict["w13_weight_scale"] = torch.empty(
num_experts,
2 * intermediate_size_per_partition,
1,
dtype=torch.float32)
param_dict["w13_weight_offset"] = torch.empty(
num_experts,
2 * intermediate_size_per_partition,
1,
dtype=torch.float32)
param_dict["w2_weight_scale"] = torch.empty(num_experts,
hidden_sizes,
1,
dtype=torch.float32)
param_dict["w2_weight_offset"] = torch.empty(num_experts,
hidden_sizes,
1,
dtype=torch.float32)
if not self.is_per_channel_weight:
param_dict["w13_weight_scale_second"] = torch.empty(
num_experts,
2 * intermediate_size_per_partition,
hidden_sizes // self.group_size,
dtype=torch.float32)
param_dict["w13_weight_offset_second"] = torch.empty(
num_experts,
2 * intermediate_size_per_partition,
hidden_sizes // self.group_size,
dtype=torch.float32)
param_dict["w2_weight_scale_second"] = torch.empty(
num_experts,
hidden_sizes,
intermediate_size_per_partition // self.group_size,
dtype=torch.float32)
param_dict["w2_weight_offset_second"] = torch.empty(
num_experts,
hidden_sizes,
intermediate_size_per_partition // self.group_size,
dtype=torch.float32)
if self.new_quant_version:
param_dict["w13_scale_bias"] = torch.empty(
num_experts,
2 * intermediate_size_per_partition,
1,
dtype=torch.float32)
param_dict["w2_scale_bias"] = torch.empty(num_experts,
hidden_sizes,
16 // self.tp_size,
dtype=torch.float32)
return param_dict
def apply(
self,
layer: torch.nn.Module,
x: torch.Tensor,
router_logits: torch.Tensor,
top_k: int,
renormalize: bool,
use_grouped_topk: bool = False,
global_num_experts: int = -1,
expert_map: Optional[torch.Tensor] = None,
topk_group: Optional[int] = None,
num_expert_group: Optional[int] = None,
custom_routing_function: Optional[Callable] = None,
scoring_func: str = "softmax",
routed_scaling_factor: float = 1.0,
e_score_correction_bias: Optional[torch.Tensor] = None,
is_prefill: bool = True,
enable_force_load_balance: bool = False,
log2phy: Optional[torch.Tensor] = None,
global_redundant_expert_num: int = 0,
**kwargs,
) -> torch.Tensor:
assert router_logits.shape[
1] == global_num_experts - global_redundant_expert_num, "Number of global experts mismatch (excluding redundancy)"
# NOTE: now npu_moe_gating_top_k can only support `group_count=256` pattern
topk_weights, topk_ids = select_experts(
hidden_states=x,
router_logits=router_logits,
top_k=top_k,
use_grouped_topk=use_grouped_topk,
renormalize=renormalize,
topk_group=topk_group,
num_expert_group=num_expert_group,
custom_routing_function=custom_routing_function,
scoring_func=scoring_func,
e_score_correction_bias=e_score_correction_bias,
global_num_experts=global_num_experts)
# this is a naive implementation for experts load balance so as
# to avoid accumulating too much tokens on a single rank.
# currently it is only activated when doing profile runs.
if enable_force_load_balance:
random_matrix = torch.rand(topk_ids.size(0),
global_num_experts -
global_redundant_expert_num,
device=topk_ids.device)
topk_ids = torch.argsort(
random_matrix, dim=1)[:, :topk_ids.size(1)].to(topk_ids.dtype)
topk_weights = topk_weights.to(x.dtype)
moe_comm_method = get_forward_context().moe_comm_method
return moe_comm_method.fused_experts(
hidden_states=x,
w1=[layer.w13_weight],
w2=[layer.w2_weight],
w1_scale=[layer.w13_weight_scale],
w2_scale=[layer.w2_weight_scale],
w1_scale_bias=layer.w13_scale_bias,
w2_scale_bias=layer.w2_scale_bias,
topk_weights=topk_weights,
topk_ids=topk_ids,
use_int4_w4a8=True,
expert_map=expert_map,
log2phy=log2phy,
dynamic_eplb=self.dynamic_eplb,
mc2_mask=kwargs.get("mc2_mask", None))
def process_scale(self, weight: torch.Tensor, scale, per_group_scale):
scale = scale.transpose(1, 2).contiguous()
if self.is_per_channel_weight:
scale_np = scale.cpu().numpy()
scale_np.dtype = np.uint32
scale_uint64_tensor = torch.from_numpy(scale_np.astype(
np.int64)).npu()
return scale_uint64_tensor, None
per_group_scale = per_group_scale.transpose(1, 2).contiguous()
group_num, k, n = weight.shape
# the weight of the new version is reduced by half by pack n, so it needs to be restored
if self.new_quant_version:
n = n * 2
per_group_scale = per_group_scale.reshape(group_num, -1, n)
group_num, quantgroup_num, n = per_group_scale.shape
bias = None
if not self.new_quant_version:
weight_high = weight.to(torch.float32).reshape([group_num, quantgroup_num, -1, n]) * \
per_group_scale.reshape([group_num, quantgroup_num, 1, n])
weight_high = weight_high.reshape([group_num, k, n])
bias = 8 * (weight_high.to(torch.float32) * scale).sum(axis=1)
scale_fp32 = (scale * per_group_scale).to(torch.float16).to(
torch.float32)
scale_fp32_np = scale_fp32.cpu().numpy()
scale_fp32_np.dtype = np.uint32
sscale_uint64 = np.zeros((group_num, quantgroup_num, n * 2),
dtype=np.uint32)
sscale_uint64[..., ::2] = scale_fp32_np
sscale_uint64_buffer = np.frombuffer(sscale_uint64.tobytes(),
dtype=np.int64).copy()
sscale_uint64_tensor = torch.from_numpy(sscale_uint64_buffer).reshape(
group_num, quantgroup_num, n)
sscale_uint64_tensor = sscale_uint64_tensor.npu()
return sscale_uint64_tensor, bias
def update_bias(self, layer, w13_bias, w2_bias):
if self.new_quant_version:
layer.w13_scale_bias.data = layer.w13_scale_bias.data.transpose(
1, 2).contiguous().sum(axis=1)
layer.w2_scale_bias.data = layer.w2_scale_bias.data.transpose(
1, 2).contiguous().sum(axis=1)
else:
w13_scale_bias = torch.nn.Parameter(w13_bias, requires_grad=False)
layer.register_parameter("w13_scale_bias", w13_scale_bias)
w2_scale_bias = torch.nn.Parameter(w2_bias, requires_grad=False)
layer.register_parameter("w2_scale_bias", w2_scale_bias)
def pack_to_int32(self, weight: torch.Tensor):
if self.new_quant_version:
# pack 4 int8(int4*2) to int32, because in pytorch, we need to use int32 to represent int4
assert weight.shape[
-1] % 4 == 0, "the last dim of weight needs to be divided by 4"
return weight.view(torch.int32).contiguous()
else:
return torch_npu.npu_quantize(weight.to(torch.float32),
torch.tensor([1.]).npu(), None,
torch.quint4x2, -1, False)
def process_weights_after_loading(self, layer):
layer.w13_weight.data = layer.w13_weight.data.transpose(
1, 2).contiguous()
layer.w2_weight.data = layer.w2_weight.data.transpose(1,
2).contiguous()
w13_weight_scale_second = layer.w13_weight_scale_second.data if hasattr(
layer, "w13_weight_scale_second") else None
w2_weight_scale_second = layer.w2_weight_scale_second.data if hasattr(
layer, "w2_weight_scale_second") else None
layer.w13_weight_scale.data, w13_bias = self.process_scale(
layer.w13_weight, layer.w13_weight_scale.data,
w13_weight_scale_second)
layer.w2_weight_scale.data, w2_bias = self.process_scale(
layer.w2_weight, layer.w2_weight_scale.data,
w2_weight_scale_second)
if hasattr(layer, "w13_weight_scale_second"):
# scale_second is no longer used, release this part of the memory
del layer.w13_weight_scale_second
del layer.w2_weight_scale_second
del layer.w13_weight_offset_second
del layer.w2_weight_offset_second
self.update_bias(layer, w13_bias, w2_bias)
layer.w13_weight.data = maybe_trans_nz(layer.w13_weight.data)
layer.w2_weight.data = maybe_trans_nz(layer.w2_weight.data)
layer.w13_weight.data = self.pack_to_int32(layer.w13_weight.data)
layer.w2_weight.data = self.pack_to_int32(layer.w2_weight.data)
Reference in New Issue View Git Blame Copy Permalink