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.
41 lines
1.7 KiB
Python
41 lines
1.7 KiB
Python
from vllm.transformers_utils.configs.chatglm import ChatGLMConfig
|
|
from vllm.transformers_utils.configs.dbrx import DbrxConfig
|
|
from vllm.transformers_utils.configs.eagle import EAGLEConfig
|
|
from vllm.transformers_utils.configs.exaone import ExaoneConfig
|
|
# RWConfig is for the original tiiuae/falcon-40b(-instruct) and
|
|
# tiiuae/falcon-7b(-instruct) models. Newer Falcon models will use the
|
|
# `FalconConfig` class from the official HuggingFace transformers library.
|
|
from vllm.transformers_utils.configs.falcon import RWConfig
|
|
from vllm.transformers_utils.configs.internvl import InternVLChatConfig
|
|
from vllm.transformers_utils.configs.jais import JAISConfig
|
|
from vllm.transformers_utils.configs.medusa import MedusaConfig
|
|
from vllm.transformers_utils.configs.mllama import MllamaConfig
|
|
from vllm.transformers_utils.configs.mlp_speculator import MLPSpeculatorConfig
|
|
from vllm.transformers_utils.configs.mpt import MPTConfig
|
|
from vllm.transformers_utils.configs.nemotron import NemotronConfig
|
|
from vllm.transformers_utils.configs.nvlm_d import NVLM_D_Config
|
|
from vllm.transformers_utils.configs.qwen2vl import (Qwen2VLConfig,
|
|
Qwen2VLVisionConfig)
|
|
from vllm.transformers_utils.configs.solar import SolarConfig
|
|
from vllm.transformers_utils.configs.ultravox import UltravoxConfig
|
|
|
|
__all__ = [
|
|
"ChatGLMConfig",
|
|
"DbrxConfig",
|
|
"MPTConfig",
|
|
"RWConfig",
|
|
"InternVLChatConfig",
|
|
"JAISConfig",
|
|
"MedusaConfig",
|
|
"EAGLEConfig",
|
|
"ExaoneConfig",
|
|
"MllamaConfig",
|
|
"MLPSpeculatorConfig",
|
|
"NemotronConfig",
|
|
"NVLM_D_Config",
|
|
"SolarConfig",
|
|
"UltravoxConfig",
|
|
"Qwen2VLConfig",
|
|
"Qwen2VLVisionConfig",
|
|
]
|