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

[Lint]Style: Convert vllm-ascend/ to ruff format(Batch #10) (#6173)

Browse Source 
### What this PR does / why we need it?
**Scope of Changes**:
| File Path |
| :--- |
|`vllm_ascend/ops/layer_shard_linear.py`|
|`vllm_ascend/ops/linear.py`|
|`vllm_ascend/ops/linear_op.py`|
|`vllm_ascend/worker/worker.py`|
| ` vllm_ascend/patch/worker/patch_bert.py` |
| ` vllm_ascend/patch/worker/patch_deepseek.py` |
| ` vllm_ascend/patch/worker/patch_distributed.py` |
| ` vllm_ascend/patch/worker/patch_module.py` |
| ` vllm_ascend/patch/worker/patch_multimodal_merge.py` |
| ` vllm_ascend/patch/worker/patch_qwen3_next.py` |
| ` vllm_ascend/patch/worker/patch_qwen3_next_mtp.py` |
| ` vllm_ascend/patch/worker/patch_rejection_sampler.py` |
| ` vllm_ascend/patch/worker/patch_rope.py` |
| ` vllm_ascend/patch/worker/patch_triton.py` |
| ` vllm_ascend/patch/worker/patch_unquantized_gemm.py` |
| ` vllm_ascend/patch/worker/patch_v2_egale.py` |
|` vllm_ascend/worker/npu_input_batch.py`|
|` vllm_ascend/worker/v2/aclgraph_utils.py`|
|` vllm_ascend/worker/v2/attn_utils.py`|
|` vllm_ascend/worker/v2/model_runner.py`|
|` vllm_ascend/worker/v2/sample/gumbel.py`|
|` vllm_ascend/worker/v2/sample/penalties.py`|
|` vllm_ascend/worker/v2/sample/sampler.py`|
|` vllm_ascend/worker/v2/spec_decode/__init__.py`|
|` vllm_ascend/worker/v2/spec_decode/eagle.py`|
|` vllm_ascend/worker/v2/states.py`|
### Does this PR introduce _any_ user-facing change?

### How was this patch tested?

- vLLM version: v0.14.0
- vLLM main:
d68209402d

Signed-off-by: MrZ20 <2609716663@qq.com>
Signed-off-by: SILONG ZENG <2609716663@qq.com>
Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>
Co-authored-by: wangxiyuan <wangxiyuan1007@gmail.com>
This commit is contained in:
SILONG ZENG
2026-02-06 15:35:06 +08:00
committed by GitHub
parent 65b7f716e6
commit 19b5d44ea8
33 changed files with 938 additions and 1243 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
vllm_ascend/worker/v2/sample/penalties.py
View File

@@ -68,8 +68,7 @@ def _penalties_and_temperature_kernel(
if use_penalty:
req_state_idx = tl.load(idx_mapping_ptr + batch_idx)
output_bin_counts = tl.load(
output_bin_counts_ptr + req_state_idx * output_bin_counts_stride +
block,
output_bin_counts_ptr + req_state_idx * output_bin_counts_stride + block,
mask=mask,
)
# to use vector core, if use > 0 will use scalar to slow down performance
@@ -77,11 +76,9 @@ def _penalties_and_temperature_kernel(
# Apply repetition penalties.
if use_rep_penalty:
packed_block = block_idx * BLOCK_SIZE // 32 + tl.arange(
0, BLOCK_SIZE // 32)
packed_block = block_idx * BLOCK_SIZE // 32 + tl.arange(0, BLOCK_SIZE // 32)
packed_mask = tl.load(
prompt_bin_mask_ptr + req_state_idx * prompt_bin_mask_stride +
packed_block,
prompt_bin_mask_ptr + req_state_idx * prompt_bin_mask_stride + packed_block,
mask=packed_block < tl.cdiv(vocab_size, 32),
)
# the compiler itself does not optimize right-shift operations, so we change the same func
@@ -97,8 +94,7 @@ def _penalties_and_temperature_kernel(
prompt_bin_mask = prompt_bin_mask.reshape(BLOCK_SIZE)
# If token appears in prompt or output, apply, otherwise use 1.0 for no-op.
scale = tl.where(prompt_bin_mask | output_bin_mask, rep_penalty,
1.0)
scale = tl.where(prompt_bin_mask | output_bin_mask, rep_penalty, 1.0)
# If logits are positive, divide by penalty, otherwise multiply by penalty.
logits *= tl.where(logits > 0, 1.0 / scale, scale)