From d26a28ab033697f55a1414b5b0435de7cd6045b6 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Juan=20P=C3=A9rez=20de=20Algaba?= <124347725+jperezdealgaba@users.noreply.github.com> Date: Fri, 15 May 2026 11:31:54 +0200 Subject: [PATCH] fix: propagate revision/code_revision pins to all artifact boundaries (#42616) Signed-off-by: jperezde Co-authored-by: Cyrus Leung --- tests/models/test_gguf_download.py | 7 +++++-- vllm/model_executor/model_loader/gguf_loader.py | 7 ++++++- vllm/model_executor/models/kimi_audio.py | 3 ++- vllm/model_executor/models/kimi_k25.py | 1 + vllm/model_executor/models/registry.py | 2 ++ vllm/model_executor/models/roberta.py | 2 +- 6 files changed, 17 insertions(+), 5 deletions(-) diff --git a/tests/models/test_gguf_download.py b/tests/models/test_gguf_download.py index e9ca35afd6..7cf8a7660c 100644 --- a/tests/models/test_gguf_download.py +++ b/tests/models/test_gguf_download.py @@ -122,14 +122,17 @@ class TestGGUFModelLoader: mock_hf_download.return_value = "/downloaded/model.gguf" - # Create a simple mock ModelConfig with only the model attribute model_config = MagicMock() model_config.model = "unsloth/Qwen3-0.6B-GGUF/model.gguf" + model_config.revision = "abc123" result = loader._prepare_weights(model_config) assert result == "/downloaded/model.gguf" mock_hf_download.assert_called_once_with( - repo_id="unsloth/Qwen3-0.6B-GGUF", filename="model.gguf" + repo_id="unsloth/Qwen3-0.6B-GGUF", + filename="model.gguf", + revision="abc123", + cache_dir=None, ) @patch("vllm.config.model.get_hf_image_processor_config", return_value=None) diff --git a/vllm/model_executor/model_loader/gguf_loader.py b/vllm/model_executor/model_loader/gguf_loader.py index fc6f88b49e..6148caa987 100644 --- a/vllm/model_executor/model_loader/gguf_loader.py +++ b/vllm/model_executor/model_loader/gguf_loader.py @@ -57,7 +57,12 @@ class GGUFModelLoader(BaseModelLoader): # repo id/filename.gguf if "/" in model_name_or_path and model_name_or_path.endswith(".gguf"): repo_id, filename = model_name_or_path.rsplit("/", 1) - return hf_hub_download(repo_id=repo_id, filename=filename) + return hf_hub_download( + repo_id=repo_id, + filename=filename, + revision=model_config.revision, + cache_dir=self.load_config.download_dir, + ) # repo_id:quant_type elif "/" in model_name_or_path and ":" in model_name_or_path: repo_id, quant_type = model_name_or_path.rsplit(":", 1) diff --git a/vllm/model_executor/models/kimi_audio.py b/vllm/model_executor/models/kimi_audio.py index 475661e0eb..6232c39a44 100644 --- a/vllm/model_executor/models/kimi_audio.py +++ b/vllm/model_executor/models/kimi_audio.py @@ -92,6 +92,7 @@ class KimiAudioWhisperEncoder(WhisperEncoder): whisper_config = HFWhisperConfig.from_pretrained( model_path, subfolder=KIMIA_WHISPER_SUBFOLDER, + revision=vllm_config.model_config.revision, ) super().__init__( @@ -426,7 +427,7 @@ class KimiAudioForConditionalGeneration( DefaultModelLoader.Source( model_or_path=vllm_config.model_config.model, subfolder="whisper-large-v3", - revision=None, + revision=vllm_config.model_config.revision, ) ] diff --git a/vllm/model_executor/models/kimi_k25.py b/vllm/model_executor/models/kimi_k25.py index eb7edfca64..297ebca29b 100644 --- a/vllm/model_executor/models/kimi_k25.py +++ b/vllm/model_executor/models/kimi_k25.py @@ -110,6 +110,7 @@ class KimiK25ProcessingInfo(BaseProcessingInfo): tokenizer = self.get_tokenizer() image_processor = cached_get_image_processor( self.ctx.model_config.model, + revision=self.ctx.model_config.revision, trust_remote_code=self.ctx.model_config.trust_remote_code, ) diff --git a/vllm/model_executor/models/registry.py b/vllm/model_executor/models/registry.py index 513185e127..d3d009ff81 100644 --- a/vllm/model_executor/models/registry.py +++ b/vllm/model_executor/models/registry.py @@ -1064,6 +1064,7 @@ class _ModelRegistry: module, model_config.model, revision=model_config.revision, + code_revision=model_config.code_revision, trust_remote_code=model_config.trust_remote_code, warn_on_fail=False, ) @@ -1077,6 +1078,7 @@ class _ModelRegistry: module, model_config.model, revision=model_config.revision, + code_revision=model_config.code_revision, trust_remote_code=model_config.trust_remote_code, warn_on_fail=True, ) diff --git a/vllm/model_executor/models/roberta.py b/vllm/model_executor/models/roberta.py index c7c292e709..43c37b01e4 100644 --- a/vllm/model_executor/models/roberta.py +++ b/vllm/model_executor/models/roberta.py @@ -203,7 +203,7 @@ class BgeM3EmbeddingModel(RobertaEmbeddingModel): self.secondary_weights = [ DefaultModelLoader.Source( model_or_path=vllm_config.model_config.model, - revision=None, + revision=vllm_config.model_config.revision, prefix=prefix, allow_patterns_overrides=[filename], )