129 lines
5.0 KiB
Python
129 lines
5.0 KiB
Python
#
|
|
# Copyright (c) 2025 Huawei Technologies Co., Ltd. All Rights Reserved.
|
|
#
|
|
# 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.
|
|
# This file is a part of the vllm-ascend project.
|
|
#
|
|
|
|
import torch
|
|
import vllm.envs as envs
|
|
|
|
from vllm_ascend.ascend_config import get_ascend_config
|
|
from vllm_ascend.sample.sampler import (
|
|
DEFAULT_LOGPROBS_MODE,
|
|
AscendSampler,
|
|
AscendTopKTopPSampler,
|
|
)
|
|
from vllm_ascend.utils import global_stream, npu_stream_switch
|
|
|
|
_CPU_GENERATOR_CACHE_310P: dict[int, tuple[torch.Generator, int]] = {}
|
|
|
|
|
|
def _get_cpu_generator_310p(i: int, generator: torch.Generator) -> torch.Generator:
|
|
cache_entry = _CPU_GENERATOR_CACHE_310P.get(i)
|
|
if cache_entry is None or cache_entry[1] != id(generator):
|
|
cpu_generator = torch.Generator(device="cpu")
|
|
try:
|
|
# Keep RNG stream consistent with the original generator.
|
|
cpu_generator.set_state(generator.get_state())
|
|
except Exception:
|
|
cpu_generator.manual_seed(generator.initial_seed())
|
|
cache_entry = (cpu_generator, id(generator))
|
|
_CPU_GENERATOR_CACHE_310P[i] = cache_entry
|
|
return cache_entry[0]
|
|
|
|
|
|
def _fill_cpu_exponential_310p(
|
|
q_cpu: torch.Tensor,
|
|
generators: dict[int, torch.Generator],
|
|
has_draft_mask: torch.Tensor | None = None,
|
|
) -> None:
|
|
"""Fill a CPU tensor with exponential values for 310P stability."""
|
|
if has_draft_mask is not None:
|
|
has_draft_mask = has_draft_mask.cpu()
|
|
# Prefill all rows so unmasked requests do not keep uninitialized values.
|
|
q_cpu.exponential_()
|
|
elif len(generators) != q_cpu.shape[0]:
|
|
q_cpu.exponential_()
|
|
if not generators:
|
|
return
|
|
for i, generator in generators.items():
|
|
cpu_gen = _get_cpu_generator_310p(i, generator)
|
|
if has_draft_mask is not None:
|
|
temp_q = torch.empty_like(q_cpu[i])
|
|
temp_q.exponential_(generator=cpu_gen)
|
|
q_cpu[i] = torch.where(has_draft_mask[i], temp_q, q_cpu[i])
|
|
else:
|
|
q_cpu[i].exponential_(generator=cpu_gen)
|
|
|
|
|
|
def fill_exponential_310p(
|
|
q: torch.Tensor,
|
|
generators: dict[int, torch.Generator],
|
|
has_draft_mask: torch.Tensor | None = None,
|
|
) -> None:
|
|
"""Fill ``q`` with exponential values using CPU RNG for 310P stability."""
|
|
q_cpu = q.cpu()
|
|
_fill_cpu_exponential_310p(q_cpu, generators, has_draft_mask)
|
|
q.copy_(q_cpu.to(q.device))
|
|
# Ensure H2D of q is visible before rejection/recover consumes it.
|
|
torch.npu.current_stream().synchronize()
|
|
|
|
|
|
def _random_sample_310p(
|
|
probs: torch.Tensor,
|
|
generators: dict[int, torch.Generator],
|
|
) -> torch.Tensor:
|
|
"""310P-specific random sampling with CPU exponential generation for q."""
|
|
with npu_stream_switch(global_stream()):
|
|
q = torch.empty_like(probs).cpu()
|
|
_fill_cpu_exponential_310p(q, generators)
|
|
q = q.npu()
|
|
torch.npu.current_stream().wait_stream(global_stream())
|
|
return probs.div_(q).argmax(dim=-1).view(-1)
|
|
|
|
|
|
class AscendTopKTopPSampler310(AscendTopKTopPSampler):
|
|
def forward_native(self, logits, generators, k, p):
|
|
if envs.VLLM_BATCH_INVARIANT:
|
|
return super().forward_native(logits, generators, k, p)
|
|
if get_ascend_config().enable_reduce_sample:
|
|
cand_logits, cand_idx = self.apply_top_k_top_p(logits, k, p, self.top_k)
|
|
logits_to_return = None
|
|
if self.logprobs_mode == "processed_logits":
|
|
logits_to_return = cand_logits
|
|
elif self.logprobs_mode == "processed_logprobs":
|
|
logits_to_return = cand_logits.log_softmax(dim=-1, dtype=torch.float32)
|
|
|
|
probs = torch.softmax(cand_logits, dim=-1)
|
|
pos = _random_sample_310p(probs, generators) # [B]
|
|
|
|
next_token = cand_idx.gather(dim=1, index=pos.unsqueeze(1)).squeeze(1) # [B]
|
|
return next_token, logits_to_return
|
|
else:
|
|
logits = self.apply_top_k_top_p(logits, k, p)
|
|
logits_to_return = None
|
|
if self.logprobs_mode == "processed_logits":
|
|
logits_to_return = logits
|
|
elif self.logprobs_mode == "processed_logprobs":
|
|
logits_to_return = logits.log_softmax(dim=-1, dtype=torch.float32)
|
|
|
|
probs = logits.softmax(dim=-1, dtype=torch.float32)
|
|
return _random_sample_310p(probs, generators), logits_to_return
|
|
|
|
|
|
class AscendSampler310(AscendSampler):
|
|
def __init__(self, logprobs_mode=DEFAULT_LOGPROBS_MODE):
|
|
super().__init__(logprobs_mode=logprobs_mode)
|
|
self.topk_topp_sampler = AscendTopKTopPSampler310(logprobs_mode=logprobs_mode)
|