clean 0.15.0 support (#6852)
Clean up vllm 0.15.0 related code
- vLLM version: v0.16.0
- vLLM main:
15d76f74e2
Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>
This commit is contained in:
@@ -19,7 +19,6 @@ from vllm_ascend.ops.triton.reject_sample import (
|
||||
sample_recovered_tokens_kernel,
|
||||
)
|
||||
from vllm_ascend.sample.sampler import apply_top_k_top_p
|
||||
from vllm_ascend.utils import vllm_version_is
|
||||
|
||||
|
||||
def apply_sampling_constraints(
|
||||
@@ -167,10 +166,7 @@ def rejection_sample(
|
||||
return output_token_ids
|
||||
|
||||
# Compute probability distribution from target logits.
|
||||
if vllm_version_is("0.15.0"):
|
||||
target_probs = target_logits
|
||||
else:
|
||||
target_probs = target_logits.softmax(dim=-1, dtype=torch.float32)
|
||||
target_probs = target_logits.softmax(dim=-1, dtype=torch.float32)
|
||||
assert target_probs.is_contiguous()
|
||||
|
||||
# Generate uniform probabilities for rejection sampling.
|
||||
|
||||
Reference in New Issue
Block a user