A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 14, 04:15 UTC
[chore] Remove models weight on S3 logic (#27725)
Signed-off-by: kevin <kevin@anyscale.com>
Kevin H. Luu committed
Oct 29, 2025 at 03:29 UTC
3481e4074322cebf0c22aacbffc2fac4163220da
2 files changed
-139
vllm/engine/arg_utils.py
-10
@@ -75,7 +75,6 @@ from vllm.platforms import CpuArchEnum, current_platform
75
from vllm.plugins import load_general_plugins
76
from vllm.ray.lazy_utils import is_in_ray_actor, is_ray_initialized
77
from vllm.reasoning import ReasoningParserManager
78
-from vllm.test_utils import MODEL_WEIGHTS_S3_BUCKET, MODELS_ON_S3
78
from vllm.transformers_utils.config import (
79
get_model_path,
80
is_interleaved,
@@ -1126,15 +1125,6 @@ class EngineArgs:
1125
if check_gguf_file(self.model):
1126
self.quantization = self.load_format = "gguf"
1127
1129
- # NOTE: This is to allow model loading from S3 in CI
1130
- if (
1131
- not isinstance(self, AsyncEngineArgs)
1132
- and envs.VLLM_CI_USE_S3
1133
- and self.model in MODELS_ON_S3
1134
- and self.load_format == "auto"
1135
- ):
1136
- self.model = f"{MODEL_WEIGHTS_S3_BUCKET}/{self.model}"
1137
-
1128
if self.disable_mm_preprocessor_cache:
1129
logger.warning(
1130
"`--disable-mm-preprocessor-cache` is deprecated "
vllm/test_utils.py
deleted
-129
@@ -1,129 +0,0 @@
1
-# SPDX-License-Identifier: Apache-2.0
2
-# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
3
-MODELS_ON_S3 = [
4
- "adept/fuyu-8b",
5
- "ai21labs/AI21-Jamba-1.5-Mini",
6
- "ai21labs/Jamba-tiny-random",
7
- "ai21labs/Jamba-tiny-reward-dev",
8
- "allenai/Molmo-7B-D-0924",
9
- "allenai/OLMo-1B-hf",
10
- "allenai/OLMoE-1B-7B-0924-Instruct",
11
- "amd/Llama-3.1-8B-Instruct-FP8-KV-Quark-test",
12
- "AMead10/Llama-3.2-1B-Instruct-AWQ",
13
- "hmellor/Ilama-3.2-1B",
14
- "BAAI/bge-base-en-v1.5",
15
- "BAAI/bge-multilingual-gemma2",
16
- "BAAI/bge-reranker-v2-m3",
17
- "bigcode/starcoder2-3b",
18
- "cross-encoder/ms-marco-MiniLM-L-6-v2",
19
- "cross-encoder/quora-roberta-base",
20
- "deepseek-ai/deepseek-vl2-tiny",
21
- "distilbert/distilgpt2",
22
- "facebook/bart-base",
23
- "facebook/bart-large-cnn",
24
- # "fixie-ai/ultravox-v0_5-llama-3_2-1b",
25
- "google/gemma-1.1-2b-it",
26
- "google/gemma-2-2b-it",
27
- "google/paligemma-3b-pt-224",
28
- "h2oai/h2ovl-mississippi-800m",
29
- "HuggingFaceM4/Idefics3-8B-Llama3",
30
- "internlm/internlm2-1_8b-reward",
31
- "intfloat/e5-mistral-7b-instruct",
32
- "intfloat/multilingual-e5-small",
33
- "jason9693/Qwen2.5-1.5B-apeach",
34
- "llava-hf/llava-1.5-7b-hf",
35
- "llava-hf/llava-onevision-qwen2-0.5b-ov-hf",
36
- "llava-hf/llava-v1.6-mistral-7b-hf",
37
- "llava-hf/LLaVA-NeXT-Video-7B-hf",
38
- # "meta-llama/Llama-2-7b-hf",
39
- "meta-llama/Llama-3.2-1B",
40
- "meta-llama/Llama-3.2-1B-Instruct",
41
- "meta-llama/Meta-Llama-3-8B",
42
- "microsoft/phi-2",
43
- "microsoft/Phi-3-mini-4k-instruct",
44
- "microsoft/Phi-3-small-8k-instruct",
45
- "microsoft/Phi-3-vision-128k-instruct",
46
- "microsoft/Phi-3.5-MoE-instruct",
47
- "microsoft/Phi-3.5-vision-instruct",
48
- # "mistralai/Mistral-7B-Instruct-v0.1",
49
- "mistralai/Mixtral-8x7B-Instruct-v0.1",
50
- "mistralai/Pixtral-12B-2409",
51
- "mistral-community/Mixtral-8x22B-v0.1-AWQ",
52
- "ModelCloud/Qwen1.5-1.8B-Chat-GPTQ-4bits-dynamic-cfg-with-lm_head",
53
- "ModelCloud/Qwen1.5-1.8B-Chat-GPTQ-4bits-dynamic-cfg-with-lm_head-symFalse",
54
- "ModelCloud/Qwen1.5-1.8B-Chat-GPTQ-4bits-dynamic-cfg-with-lm_head-symTrue",
55
- "ModelCloud/TinyLlama-1.1B-Chat-v1.0-GPTQ-4bit-10-25-2024",
56
- "neuralmagic/Llama-3.2-1B-quantized.w8a8",
57
- "neuralmagic/Meta-Llama-3-8B-Instruct-FP8",
58
- "neuralmagic/Meta-Llama-3-8B-Instruct-FP8-KV",
59
- "nm-testing/asym-w8w8-int8-static-per-tensor-tiny-llama",
60
- "nm-testing/llama2.c-stories42M-pruned2.4-compressed",
61
- "nm-testing/llama7b-one-shot-2_4-w4a16-marlin24-t",
62
- "nm-testing/Meta-Llama-3-8B-FP8-compressed-tensors-test",
63
- "nm-testing/Meta-Llama-3-8B-Instruct-FP8-Dynamic-2of4-testing",
64
- "nm-testing/Meta-Llama-3-8B-Instruct-FP8-Dynamic-IA-Per-Tensor-Weight-testing",
65
- "nm-testing/Meta-Llama-3-8B-Instruct-FP8-Static-Per-Tensor-testing",
66
- "nm-testing/Meta-Llama-3-8B-Instruct-FP8-Static-testing",
67
- "nm-testing/Meta-Llama-3-8B-Instruct-W8A8-Dynamic-Asym",
68
- "nm-testing/Meta-Llama-3-8B-Instruct-W8A8-Static-Per-Tensor-Asym",
69
- "nm-testing/Meta-Llama-3-8B-Instruct-W8A8-Static-Per-Tensor-Sym",
70
- "nm-testing/Phi-3-mini-128k-instruct-FP8",
71
- "nm-testing/Qwen2-0.5B-Instruct-FP8-SkipQKV",
72
- "nm-testing/Qwen2-1.5B-Instruct-FP8-K-V",
73
- "nm-testing/tinyllama-oneshot-w4a16-channel-v2",
74
- "nm-testing/tinyllama-oneshot-w4a16-group128-v2",
75
- "nm-testing/tinyllama-oneshot-w8-channel-a8-tensor",
76
- "nm-testing/tinyllama-oneshot-w8a16-per-channel",
77
- "nm-testing/tinyllama-oneshot-w8a8-channel-dynamic-token-v2",
78
- "nm-testing/tinyllama-oneshot-w8a8-channel-dynamic-token-v2-asym",
79
- "nm-testing/tinyllama-oneshot-w8a8-dynamic-token-v2",
80
- "nm-testing/tinyllama-oneshot-w8a8-dynamic-token-v2-asym",
81
- "nm-testing/tinyllama-oneshot-w8w8-test-static-shape-change",
82
- "nm-testing/TinyLlama-1.1B-Chat-v1.0-2of4-Sparse-Dense-Compressor",
83
- "nm-testing/TinyLlama-1.1B-Chat-v1.0-gsm8k-pruned.2of4-chnl_wts_per_tok_dyn_act_fp8-BitM",
84
- "nm-testing/TinyLlama-1.1B-Chat-v1.0-gsm8k-pruned.2of4-chnl_wts_per_tok_dyn_act_int8-BitM",
85
- "nm-testing/TinyLlama-1.1B-Chat-v1.0-gsm8k-pruned.2of4-chnl_wts_tensor_act_fp8-BitM",
86
- "nm-testing/TinyLlama-1.1B-Chat-v1.0-gsm8k-pruned.2of4-chnl_wts_tensor_act_int8-BitM",
87
- "nm-testing/TinyLlama-1.1B-Chat-v1.0-gsm8k-pruned.2of4-tensor_wts_per_tok_dyn_act_fp8-BitM",
88
- "nm-testing/TinyLlama-1.1B-Chat-v1.0-gsm8k-pruned.2of4-tensor_wts_per_tok_dyn_act_int8-BitM",
89
- "nm-testing/TinyLlama-1.1B-Chat-v1.0-gsm8k-pruned.2of4-tensor_wts_tensor_act_fp8-BitM",
90
- "nm-testing/TinyLlama-1.1B-Chat-v1.0-gsm8k-pruned.2of4-tensor_wts_tensor_act_int8-BitM",
91
- "nm-testing/TinyLlama-1.1B-Chat-v1.0-INT8-Dynamic-IA-Per-Channel-Weight-testing",
92
- "nm-testing/TinyLlama-1.1B-Chat-v1.0-INT8-Dynamic-IA-Per-Tensor-Weight-testing",
93
- "nm-testing/TinyLlama-1.1B-Chat-v1.0-INT8-Static-testing",
94
- "nm-testing/TinyLlama-1.1B-compressed-tensors-kv-cache-scheme",
95
- "nvidia/NVLM-D-72B",
96
- "openai-community/gpt2",
97
- # "openai/whisper-large-v3",
98
- "openbmb/MiniCPM-o-2_6",
99
- "openbmb/MiniCPM-V-2_6",
100
- "OpenGVLab/InternVL2-1B",
101
- "parasail-ai/GritLM-7B-vllm",
102
- "Qwen/Qwen1.5-MoE-A2.7B-Chat",
103
- "Qwen/Qwen2-7B-Instruct",
104
- "Qwen/Qwen2-Audio-7B-Instruct",
105
- "Qwen/Qwen2-VL-2B-Instruct",
106
- "Qwen/Qwen2.5-1.5B-Instruct",
107
- "Qwen/Qwen2.5-Math-PRM-7B",
108
- "Qwen/Qwen2.5-Math-RM-72B",
109
- "Qwen/Qwen2.5-VL-3B-Instruct",
110
- "royokong/e5-v",
111
- "sentence-transformers/all-roberta-large-v1",
112
- "sentence-transformers/stsb-roberta-base-v2",
113
- "allenai/OLMo-2-0425-1B",
114
- "shuyuej/Llama-3.2-1B-Instruct-GPTQ",
115
- "ssmits/Qwen2-7B-Instruct-embed-base",
116
- "stabilityai/stablelm-3b-4e1t",
117
- "stabilityai/stablelm-zephyr-3b",
118
- "state-spaces/mamba-130m-hf",
119
- "TheBloke/TinyLlama-1.1B-Chat-v1.0-GPTQ",
120
- "zai-org/glm-4v-9b",
121
- "TIGER-Lab/Mantis-8B-siglip-llama3",
122
- "TIGER-Lab/VLM2Vec-Full",
123
- "tiiuae/falcon-40b",
124
- "tiiuae/falcon-mamba-7b-instruct",
125
- "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
126
- "upstage/solar-pro-preview-instruct",
127
-]
128
-
129
-MODEL_WEIGHTS_S3_BUCKET = "s3://vllm-ci-model-weights"