a7b8788d2c · tduncan/Nvidia-Optimization-Stack-VLLM

docs/models/supported_models.md

+5 -5

@@ -328,7 +328,7 @@ th {

328 | `BambaForCausalLM` | Bamba | `ibm-ai-platform/Bamba-9B-fp8`, `ibm-ai-platform/Bamba-9B` | ✅︎ | ✅︎ | ✅︎ |

331 -| `ChatGLMModel`, `ChatGLMForConditionalGeneration` | ChatGLM | `THUDM/chatglm2-6b`, `THUDM/chatglm3-6b`, `ShieldLM-6B-chatglm3`, etc. | ✅︎ | ✅︎ | ✅︎ |

331 +| `ChatGLMModel`, `ChatGLMForConditionalGeneration` | ChatGLM | `zai-org/chatglm2-6b`, `zai-org/chatglm3-6b`, `ShieldLM-6B-chatglm3`, etc. | ✅︎ | ✅︎ | ✅︎ |

332 | `CohereForCausalLM`, `Cohere2ForCausalLM` | Command-R | `CohereForAI/c4ai-command-r-v01`, `CohereForAI/c4ai-command-r7b-12-2024`, etc. | ✅︎ | ✅︎ | ✅︎ |

333 | `DbrxForCausalLM` | DBRX | `databricks/dbrx-base`, `databricks/dbrx-instruct`, etc. | | ✅︎ | ✅︎ |

334 | `DeciLMForCausalLM` | DeciLM | `nvidia/Llama-3_3-Nemotron-Super-49B-v1`, etc. | ✅︎ | ✅︎ | ✅︎ |

@@ -348,8 +348,8 @@ th {

348 | `Gemma2ForCausalLM` | Gemma 2 | `google/gemma-2-9b`, `google/gemma-2-27b`, etc. | ✅︎ | ✅︎ | ✅︎ |

349 | `Gemma3ForCausalLM` | Gemma 3 | `google/gemma-3-1b-it`, etc. | ✅︎ | ✅︎ | ✅︎ |

351 -| `GlmForCausalLM` | GLM-4 | `THUDM/glm-4-9b-chat-hf`, etc. | ✅︎ | ✅︎ | ✅︎ |

352 -| `Glm4ForCausalLM` | GLM-4-0414 | `THUDM/GLM-4-32B-0414`, etc. | ✅︎ | ✅︎ | ✅︎ |

351 +| `GlmForCausalLM` | GLM-4 | `zai-org/glm-4-9b-chat-hf`, etc. | ✅︎ | ✅︎ | ✅︎ |

352 +| `Glm4ForCausalLM` | GLM-4-0414 | `zai-org/GLM-4-32B-0414`, etc. | ✅︎ | ✅︎ | ✅︎ |

353 | `GPT2LMHeadModel` | GPT-2 | `gpt2`, `gpt2-xl`, etc. | | ✅︎ | ✅︎ |

354 | `GPTBigCodeForCausalLM` | StarCoder, SantaCoder, WizardCoder | `bigcode/starcoder`, `bigcode/gpt_bigcode-santacoder`, `WizardLM/WizardCoder-15B-V1.0`, etc. | ✅︎ | ✅︎ | ✅︎ |

355 | `GPTJForCausalLM` | GPT-J | `EleutherAI/gpt-j-6b`, `nomic-ai/gpt4all-j`, etc. | | ✅︎ | ✅︎ |

@@ -589,8 +589,8 @@ See [this page](generative_models.md) for more information on how to use generat

590 | `FuyuForCausalLM` | Fuyu | T + I | `adept/fuyu-8b`, etc. | | ✅︎ | ✅︎ |

591 | `Gemma3ForConditionalGeneration` | Gemma 3 | T + I+ | `google/gemma-3-4b-it`, `google/gemma-3-27b-it`, etc. | ✅︎ | ✅︎ | ⚠️ |

592 -| `GLM4VForCausalLM`^ | GLM-4V | T + I | `THUDM/glm-4v-9b`, `THUDM/cogagent-9b-20241220`, etc. | ✅︎ | ✅︎ | ✅︎ |

593 -| `Glm4vForConditionalGeneration` | GLM-4.1V-Thinking | T + IE+ + VE+ | `THUDM/GLM-4.1V-9B-Thinking`, etc. | ✅︎ | ✅︎ | ✅︎ |

592 +| `GLM4VForCausalLM`^ | GLM-4V | T + I | `zai-org/glm-4v-9b`, `zai-org/cogagent-9b-20241220`, etc. | ✅︎ | ✅︎ | ✅︎ |

593 +| `Glm4vForConditionalGeneration` | GLM-4.1V-Thinking | T + IE+ + VE+ | `zai-org/GLM-4.1V-9B-Thinking`, etc. | ✅︎ | ✅︎ | ✅︎ |

594 | `Glm4MoeForCausalLM` | GLM-4.5 | T + IE+ + VE+ | `zai-org/GLM-4.5`, etc. | ✅︎ | ✅︎ | ✅︎ |

595 | `Glm4v_moeForConditionalGeneration` | GLM-4.5V | T + IE+ + VE+ | `zai-org/GLM-4.5V-Air`, etc. | ✅︎ | ✅︎ | ✅︎ |

596 | `GraniteSpeechForConditionalGeneration` | Granite Speech | T + A | `ibm-granite/granite-speech-3.3-8b` | ✅︎ | ✅︎ | ✅︎ |

examples/offline_inference/vision_language.py

+2 -2

@@ -221,7 +221,7 @@ def run_gemma3(questions: list[str], modality: str) -> ModelRequestData:

221 # GLM-4v

222 def run_glm4v(questions: list[str], modality: str) -> ModelRequestData:

223 assert modality == "image"

224 - model_name = "THUDM/glm-4v-9b"

224 + model_name = "zai-org/glm-4v-9b"

225

226 engine_args = EngineArgs(

227 model=model_name,

@@ -250,7 +250,7 @@ def run_glm4v(questions: list[str], modality: str) -> ModelRequestData:

250

251 # GLM-4.1V

252 def run_glm4_1v(questions: list[str], modality: str) -> ModelRequestData:

253 - model_name = "THUDM/GLM-4.1V-9B-Thinking"

253 + model_name = "zai-org/GLM-4.1V-9B-Thinking"

254

255 engine_args = EngineArgs(

256 model=model_name,

tests/distributed/test_pipeline_parallel.py

+2 -2

@@ -154,7 +154,7 @@ TEXT_GENERATION_MODELS = {

154 "baichuan-inc/Baichuan-7B": PPTestSettings.fast(),

155 "baichuan-inc/Baichuan2-13B-Chat": PPTestSettings.fast(),

156 "bigscience/bloomz-1b1": PPTestSettings.fast(),

157 - "THUDM/chatglm3-6b": PPTestSettings.fast(),

157 + "zai-org/chatglm3-6b": PPTestSettings.fast(),

158 "CohereForAI/c4ai-command-r-v01": PPTestSettings.fast(load_format="dummy"),

159 "databricks/dbrx-instruct": PPTestSettings.fast(load_format="dummy"),

160 "Deci/DeciLM-7B-instruct": PPTestSettings.fast(),

@@ -224,7 +224,7 @@ MULTIMODAL_MODELS = {

224 "Salesforce/blip2-opt-6.7b": PPTestSettings.fast(),

225 "facebook/chameleon-7b": PPTestSettings.fast(),

226 "adept/fuyu-8b": PPTestSettings.fast(),

227 - "THUDM/glm-4v-9b": PPTestSettings.fast(),

227 + "zai-org/glm-4v-9b": PPTestSettings.fast(),

228 "OpenGVLab/InternVL2-1B": PPTestSettings.fast(),

229 "llava-hf/llava-1.5-7b-hf": PPTestSettings.fast(),

230 "llava-hf/llava-v1.6-mistral-7b-hf": PPTestSettings.fast(),

tests/lora/test_add_lora.py

+1 -1

@@ -14,7 +14,7 @@ from vllm.lora.request import LoRARequest

14 from vllm.sampling_params import SamplingParams

15 from vllm.utils import merge_async_iterators

16

17 -MODEL_PATH = "THUDM/chatglm3-6b"

17 +MODEL_PATH = "zai-org/chatglm3-6b"

18 LORA_RANK = 64

19 DEFAULT_MAX_LORAS = 4 * 3

20

tests/lora/test_chatglm3_tp.py

+1 -1

@@ -6,7 +6,7 @@ from vllm.lora.request import LoRARequest

6

7 from ..utils import create_new_process_for_each_test, multi_gpu_test

8

9 -MODEL_PATH = "THUDM/chatglm3-6b"

9 +MODEL_PATH = "zai-org/chatglm3-6b"

10

11 PROMPT_TEMPLATE = """I want you to act as a SQL terminal in front of an example database, you need only to return the sql command to me.Below is an instruction that describes a task, Write a response that appropriately completes the request.\n"\n##Instruction:\nconcert_singer contains tables such as stadium, singer, concert, singer_in_concert. Table stadium has columns such as Stadium_ID, Location, Name, Capacity, Highest, Lowest, Average. Stadium_ID is the primary key.\nTable singer has columns such as Singer_ID, Name, Country, Song_Name, Song_release_year, Age, Is_male. Singer_ID is the primary key.\nTable concert has columns such as concert_ID, concert_Name, Theme, Stadium_ID, Year. concert_ID is the primary key.\nTable singer_in_concert has columns such as concert_ID, Singer_ID. concert_ID is the primary key.\nThe Stadium_ID of concert is the foreign key of Stadium_ID of stadium.\nThe Singer_ID of singer_in_concert is the foreign key of Singer_ID of singer.\nThe concert_ID of singer_in_concert is the foreign key of concert_ID of concert.\n\n###Input:\n{query}\n\n###Response:""" # noqa: E501

12

tests/models/language/generation/test_common.py

+1 -1

@@ -53,7 +53,7 @@ AITER_MODEL_LIST = [

53 marks=[pytest.mark.core_model, pytest.mark.cpu_model],

54 ),

55 pytest.param(

56 - "THUDM/chatglm3-6b", # chatglm (text-only)

56 + "zai-org/chatglm3-6b", # chatglm (text-only)

57 ),

58 pytest.param(

59 "meta-llama/Llama-3.2-1B-Instruct", # llama

tests/models/multimodal/generation/test_common.py

+3 -3

@@ -355,7 +355,7 @@ VLM_TEST_SETTINGS = {

355 num_logprobs=10,

356 ),

357 "glm4v": VLMTestInfo(

358 - models=["THUDM/glm-4v-9b"],

358 + models=["zai-org/glm-4v-9b"],

359 test_type=VLMTestType.IMAGE,

360 prompt_formatter=lambda img_prompt: f"<|user|>\n{img_prompt}<|assistant|>", # noqa: E501

361 single_image_prompts=IMAGE_ASSETS.prompts({

@@ -374,7 +374,7 @@ VLM_TEST_SETTINGS = {

374 marks=[large_gpu_mark(min_gb=32)],

375 ),

376 "glm4_1v": VLMTestInfo(

377 - models=["THUDM/GLM-4.1V-9B-Thinking"],

377 + models=["zai-org/GLM-4.1V-9B-Thinking"],

378 test_type=(VLMTestType.IMAGE, VLMTestType.MULTI_IMAGE),

379 prompt_formatter=lambda img_prompt: f"<|user|>\n{img_prompt}<|assistant|>", # noqa: E501

@@ -388,7 +388,7 @@ VLM_TEST_SETTINGS = {

388 marks=[large_gpu_mark(min_gb=32)],

389 ),

390 "glm4_1v-video": VLMTestInfo(

391 - models=["THUDM/GLM-4.1V-9B-Thinking"],

391 + models=["zai-org/GLM-4.1V-9B-Thinking"],

392 # GLM4.1V require include video metadata for input

393 test_type=VLMTestType.CUSTOM_INPUTS,

394 max_model_len=4096,

tests/models/multimodal/processing/test_common.py

+2 -2

@@ -271,8 +271,8 @@ def _test_processing_correctness_one(

271 "microsoft/Florence-2-base",

272 "adept/fuyu-8b",

273 "google/gemma-3-4b-it",

274 - "THUDM/glm-4v-9b",

275 - "THUDM/GLM-4.1V-9B-Thinking",

274 + "zai-org/glm-4v-9b",

275 + "zai-org/GLM-4.1V-9B-Thinking",

276 "ibm-granite/granite-speech-3.3-2b",

277 "h2oai/h2ovl-mississippi-800m",

278 "internlm/Intern-S1",

tests/models/multimodal/processing/test_glm4_1v.py

+1 -1

@@ -9,7 +9,7 @@ from vllm.multimodal import MULTIMODAL_REGISTRY

9 from ...utils import build_model_context

10

11

12 -@pytest.mark.parametrize("model_id", ["THUDM/GLM-4.1V-9B-Thinking"])

12 +@pytest.mark.parametrize("model_id", ["zai-org/GLM-4.1V-9B-Thinking"])

13 @pytest.mark.parametrize("expected_toks_per_frame", [299])

14 @pytest.mark.parametrize("num_frames", [32, 128])

15 @pytest.mark.parametrize("fps, expected_grid_t", [(1, 5), (2, 10)])

tests/models/registry.py

+5 -5

@@ -153,7 +153,7 @@ _TEXT_GENERATION_EXAMPLE_MODELS = {

153 extras={"tiny": "hmellor/tiny-random-BambaForCausalLM"}), # noqa: E501

154 "BloomForCausalLM": _HfExamplesInfo("bigscience/bloom-560m",

155 {"1b": "bigscience/bloomz-1b1"}),

156 - "ChatGLMModel": _HfExamplesInfo("THUDM/chatglm3-6b",

156 + "ChatGLMModel": _HfExamplesInfo("zai-org/chatglm3-6b",

157 trust_remote_code=True,

158 max_transformers_version="4.48"),

159 "ChatGLMForConditionalGeneration": _HfExamplesInfo("thu-coai/ShieldLM-6B-chatglm3", # noqa: E501

@@ -187,8 +187,8 @@ _TEXT_GENERATION_EXAMPLE_MODELS = {

187 "Gemma3ForCausalLM": _HfExamplesInfo("google/gemma-3-1b-it"),

188 "Gemma3nForConditionalGeneration": _HfExamplesInfo("google/gemma-3n-E2B-it", # noqa: E501

189 min_transformers_version="4.53"),

190 - "GlmForCausalLM": _HfExamplesInfo("THUDM/glm-4-9b-chat-hf"),

191 - "Glm4ForCausalLM": _HfExamplesInfo("THUDM/GLM-4-9B-0414"),

190 + "GlmForCausalLM": _HfExamplesInfo("zai-org/glm-4-9b-chat-hf"),

191 + "Glm4ForCausalLM": _HfExamplesInfo("zai-org/GLM-4-9B-0414"),

192 "Glm4MoeForCausalLM": _HfExamplesInfo("zai-org/GLM-4.5",

193 min_transformers_version="4.54"), # noqa: E501

194 "GPT2LMHeadModel": _HfExamplesInfo("openai-community/gpt2",

@@ -380,10 +380,10 @@ _MULTIMODAL_EXAMPLE_MODELS = {

380 "FuyuForCausalLM": _HfExamplesInfo("adept/fuyu-8b"),

381 "Gemma3ForConditionalGeneration": _HfExamplesInfo("google/gemma-3-4b-it"),

382 "GraniteSpeechForConditionalGeneration": _HfExamplesInfo("ibm-granite/granite-speech-3.3-2b"), # noqa: E501

383 - "GLM4VForCausalLM": _HfExamplesInfo("THUDM/glm-4v-9b",

383 + "GLM4VForCausalLM": _HfExamplesInfo("zai-org/glm-4v-9b",

384 trust_remote_code=True,

385 hf_overrides={"architectures": ["GLM4VForCausalLM"]}), # noqa: E501

386 - "Glm4vForConditionalGeneration": _HfExamplesInfo("THUDM/GLM-4.1V-9B-Thinking"), # noqa: E501

386 + "Glm4vForConditionalGeneration": _HfExamplesInfo("zai-org/GLM-4.1V-9B-Thinking"), # noqa: E501

387 "Glm4v_moeForConditionalGeneration": _HfExamplesInfo("zai-org/GLM-4.5V-Air",

388 is_available_online=False), # noqa: E501

389 "H2OVLChatModel": _HfExamplesInfo("h2oai/h2ovl-mississippi-800m",

tests/tokenization/test_cached_tokenizer.py

+1 -1

@@ -10,7 +10,7 @@ from vllm.transformers_utils.tokenizer import (AnyTokenizer,

10 get_cached_tokenizer)

11

12

13 -@pytest.mark.parametrize("model_id", ["gpt2", "THUDM/chatglm3-6b"])

13 +@pytest.mark.parametrize("model_id", ["gpt2", "zai-org/chatglm3-6b"])

14 def test_cached_tokenizer(model_id: str):

15 reference_tokenizer = AutoTokenizer.from_pretrained(model_id,

16 trust_remote_code=True)

vllm/model_executor/models/chatglm.py

+3 -3

@@ -1,7 +1,7 @@

1 # SPDX-License-Identifier: Apache-2.0

2 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project

3 # Adapted from

4 -# https://github.com/THUDM/ChatGLM2-6B

4 +# https://github.com/zai-org/ChatGLM2-6B

5 """Inference-only ChatGLM model compatible with THUDM weights."""

6 import json

7 from collections.abc import Iterable

@@ -86,10 +86,10 @@ class GLMAttention(nn.Module):

86 prefix=f"{prefix}.dense",

87 )

88

89 - # https://huggingface.co/THUDM/chatglm3-6b-32k/blob/e210410255278dd9d74463cf396ba559c0ef801c/modeling_chatglm.py#L141

89 + # https://huggingface.co/zai-org/chatglm3-6b-32k/blob/e210410255278dd9d74463cf396ba559c0ef801c/modeling_chatglm.py#L141

90 rope_ratio = getattr(config, "rope_ratio", 1.0)

91 max_positions = getattr(config, "seq_length", 8192)

92 - # NOTE: THUDM/cogagent-9b-20241220 uses original_rope=False,

92 + # NOTE: zai-org/cogagent-9b-20241220 uses original_rope=False,

93 # which is equivalent to is_neox_style=True

94 is_neox_style = not config.original_rope

95 self.rotary_emb = get_rope(

vllm/model_executor/models/glm4v.py

+1 -1

@@ -2,7 +2,7 @@

2 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project

3

4 # Adapted from

5 -# https://github.com/THUDM/CogAgent

5 +# https://github.com/zai-org/CogAgent

6 """Inference-only CogAgent model compatible with THUDM weights."""

7 from argparse import Namespace

8 from collections.abc import Mapping, Sequence

vllm/test_utils.py

+1 -1

@@ -118,7 +118,7 @@ MODELS_ON_S3 = [

118 "stabilityai/stablelm-zephyr-3b",

119 "state-spaces/mamba-130m-hf",

120 "TheBloke/TinyLlama-1.1B-Chat-v1.0-GPTQ",

121 - "THUDM/glm-4v-9b",

121 + "zai-org/glm-4v-9b",

122 "TIGER-Lab/Mantis-8B-siglip-llama3",

123 "TIGER-Lab/VLM2Vec-Full",

124 "tiiuae/falcon-40b",

vllm/transformers_utils/configs/chatglm.py

+1 -1

@@ -2,7 +2,7 @@

2 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project

3

4 # Adapted from

5 -# https://github.com/THUDM/ChatGLM2-6B

5 +# https://github.com/zai-org/ChatGLM2-6B

6 from transformers import PretrainedConfig

7

8

vllm/transformers_utils/tokenizer.py

+1 -1

@@ -271,7 +271,7 @@ def get_tokenizer(

271 }

272 tokenizer.add_special_tokens(special_tokens_map)

273

274 - # NOTE: We can remove this after https://github.com/THUDM/ChatGLM3/issues/1324

274 + # NOTE: We can remove this after https://github.com/zai-org/ChatGLM3/issues/1324

275 if type(tokenizer).__name__ in ("ChatGLMTokenizer",

276 "ChatGLM4Tokenizer"):

277 assert isinstance(tokenizer, PreTrainedTokenizer)

[Misc] Modify the organization of GLM series (#22171)