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.
24 lines
678 B
Python
24 lines
678 B
Python
from typing import Any, Optional
|
|
|
|
import msgspec
|
|
|
|
|
|
class PoolingParams(
|
|
msgspec.Struct,
|
|
omit_defaults=True, # type: ignore[call-arg]
|
|
array_like=True): # type: ignore[call-arg]
|
|
"""Pooling parameters for pooling.
|
|
|
|
Attributes:
|
|
additional_data: Any additional data needed for pooling.
|
|
"""
|
|
additional_data: Optional[Any] = None
|
|
|
|
def clone(self) -> "PoolingParams":
|
|
"""Returns a deep copy of the PoolingParams instance."""
|
|
return PoolingParams(additional_data=self.additional_data, )
|
|
|
|
def __repr__(self) -> str:
|
|
return (f"PoolingParams("
|
|
f"additional_metadata={self.additional_data})")
|