EngineX-Ascend/enginex-ascend-910-vllm
11
0
Fork 1
Code Issues Pull Requests Actions Projects Releases Wiki Activity

init v0.23.0

Browse Source 
Signed-off-by: Sun Ruoxi <sunruoxi@4paradigm.com>
This commit is contained in:
Sun Ruoxi
2026-08-27 15:11:51 +08:00
parent b582a8e7d1
commit 7f8a1b1f7a
2849 changed files with 712887 additions and 22001 deletions
Download Patch File Download Diff File Expand all files Collapse all files

34
vllm_ascend/patch/worker/patch_v2/patch_triton.py Normal file
View File

@@ -0,0 +1,34 @@
from vllm.v1.worker.gpu import input_batch, model_runner, structured_outputs
from vllm.v1.worker.gpu.sample import bad_words, gumbel, logprob, penalties, prompt_logprob, sampler, states
from vllm.v1.worker.gpu.spec_decode import rejection_sampler, rejection_sampler_utils
from vllm.v1.worker.gpu.spec_decode.eagle import speculator
from vllm_ascend.worker.v2.input_batch import post_update
from vllm_ascend.worker.v2.sample.bad_words import apply_bad_words
from vllm_ascend.worker.v2.sample.gumbel import apply_temperature, gumbel_sample
from vllm_ascend.worker.v2.sample.logprob import compute_token_logprobs, compute_topk_logprobs
from vllm_ascend.worker.v2.sample.min_p import apply_min_p
from vllm_ascend.worker.v2.sample.penalties import apply_penalties, bincount
from vllm_ascend.worker.v2.spec_decode.rejection_sampler_utils import (
rejection_sample as npu_rejection_sample,
)
from vllm_ascend.worker.v2.structured_outputs import _apply_grammar_bitmask_kernel
penalties.apply_penalties = apply_penalties
# because sampler.py and speculator.py are imported before this patch, they must be overridden
sampler.gumbel_sample = gumbel_sample
input_batch.post_update = post_update
prompt_logprob.compute_topk_logprobs = compute_topk_logprobs
sampler.compute_topk_logprobs = compute_topk_logprobs
rejection_sampler.compute_topk_logprobs = compute_topk_logprobs
states.apply_min_p = apply_min_p
penalties.bincount = bincount
speculator.gumbel_sample = gumbel_sample
model_runner.post_update = post_update
bad_words.apply_bad_words = apply_bad_words
gumbel.apply_temperature = apply_temperature
states.apply_temperature = apply_temperature
logprob.compute_token_logprobs = compute_token_logprobs
structured_outputs._apply_grammar_bitmask_kernel = _apply_grammar_bitmask_kernel
rejection_sampler_utils.rejection_sample = npu_rejection_sample
rejection_sampler.rejection_sample = npu_rejection_sample