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

[CI] Remove compatibility maintenance for vllm v0.10.1 and v0.10.1.1 (#2840)

Browse Source 
### What this PR does / why we need it?
Remove compatibility maintenance for vllm v0.10.1 and v0.10.1.1

### Does this PR introduce _any_ user-facing change?
branch main of vllm-ascend will not be compatible with vllm v0.10.1 and
v0.10.1.1

### How was this patch tested?
CI passed with existing test.

- vLLM version: v0.10.1.1
- vLLM main:
6fb2788163

---------

Signed-off-by: MengqingCao <cmq0113@163.com>
This commit is contained in:
Mengqing Cao
2025-09-10 08:43:10 +08:00
committed by GitHub
parent 93e28e6862
commit edf1f600ad
22 changed files with 340 additions and 876 deletions
Download Patch File Download Diff File Expand all files Collapse all files

30
vllm_ascend/sample/sampler.py
View File

@@ -1,16 +1,12 @@
import torch
import torch_npu
from vllm.config import LogprobsMode
from vllm.v1.sample.ops.topk_topp_sampler import TopKTopPSampler, random_sample
from vllm.v1.sample.sampler import Sampler
from vllm_ascend.utils import is_310p, vllm_version_is
from vllm_ascend.utils import is_310p
if not (vllm_version_is("0.10.1.1") or vllm_version_is("0.10.1")):
from vllm.config import LogprobsMode
DEFAULT_LOGPROBS_MODE = LogprobsMode.RAW_LOGPROBS
else:
LogprobsMode = None
DEFAULT_LOGPROBS_MODE = "raw_logprobs"
DEFAULT_LOGPROBS_MODE = LogprobsMode.RAW_LOGPROBS
class AscendSampler(Sampler):
@@ -68,19 +64,11 @@ class AscendTopKTopPSampler(TopKTopPSampler):
def forward_native(self, logits, generators, k, p):
"""Override pytorch native implementation to torch_npu"""
logits = self._apply_top_k_top_p(logits, k, p)
if not (vllm_version_is("0.10.1.1") or vllm_version_is("0.10.1")):
logits_to_return = None
if self.logprobs_mode == LogprobsMode.PROCESSED_LOGITS:
logits_to_return = logits
elif self.logprobs_mode == LogprobsMode.PROCESSED_LOGPROBS:
logits_to_return = logits.log_softmax(dim=-1,
dtype=torch.float32)
logits_to_return = None
if self.logprobs_mode == LogprobsMode.PROCESSED_LOGITS:
logits_to_return = logits
elif self.logprobs_mode == LogprobsMode.PROCESSED_LOGPROBS:
logits_to_return = logits.log_softmax(dim=-1, dtype=torch.float32)
probs = logits.softmax(dim=-1, dtype=torch.float32)
output = None
if vllm_version_is("0.10.1.1") or vllm_version_is("0.10.1"):
output = random_sample(probs, generators)
else:
output = (random_sample(probs, generators), logits_to_return)
return output
return random_sample(probs, generators), logits_to_return