ef6abf3dc7
Adds ALL files needed for Dockerfile build:
- qwen3_6_scripts/ (baseline patches + our optimizations)
- vllm/ (full vllm package)
- paged_attention_v2_pytorch.py (V2 with single-bmm optimization)
- Dockerfile + computility-run.yaml
Our optimizations vs baseline:
1. paged_attn.py: pre-gathered context KV (eliminates 194 gather calls),
Triton try/fallback, V2 heuristic, threshold 32K→64K
2. paged_attention_v2_pytorch.py: fills NotImplementedError,
single-bmm Phase 1 (195 launches → 3)
3. patch_enable_triton.py: HAS_TRITON=True with safety fallback
4. patch_triton_tuning.py: BLOCK=64, NUM_WARPS=4 for BI-V100
5. computility-run.yaml: gpu-memory-utilization 0.9→0.95,
max-num-batched-tokens 8192→16384
This repo can now be submitted to dev.modelhub.org.cn as-is.
23 lines
820 B
Python
23 lines
820 B
Python
from typing import List
|
|
from typing import Sequence as GenericSequence
|
|
from typing import Union
|
|
|
|
from vllm.model_executor.layers.sampler import SamplerOutput
|
|
from vllm.sequence import PoolerOutput, SequenceGroupOutput
|
|
|
|
|
|
def create_output_by_sequence_group(
|
|
outputs: GenericSequence[Union[SamplerOutput, PoolerOutput]],
|
|
num_seq_groups: int) -> List[List[SequenceGroupOutput]]:
|
|
"""Helper method which transforms a 2d list organized by
|
|
[step][sequence group] into [sequence group][step].
|
|
"""
|
|
output_by_sequence_group: List[List[SequenceGroupOutput]] = [
|
|
[] for _ in range(num_seq_groups)
|
|
]
|
|
for step in outputs:
|
|
for i, sequence_group_output in enumerate(step):
|
|
output_by_sequence_group[i].append(sequence_group_output)
|
|
|
|
return output_by_sequence_group
|