fix(ml): drop resume_download, removed from from_pretrained in transformers v5 (#15054)

autobot-backend/ai_hardware_accelerator.py

@@ -622,13 +622,10 @@ def _initialize_clip_model(self, device: Any) -> None:
         torch = _get_torch()
 
         # HuggingFace model loaded by name; revision pinning managed operationally.
-        self.clip_processor = CLIPProcessor.from_pretrained(  # nosec B615
-            "openai/clip-vit-base-patch32", resume_download=True
-        )
+        self.clip_processor = CLIPProcessor.from_pretrained("openai/clip-vit-base-patch32")  # nosec B615
         self.clip_model = CLIPModel.from_pretrained(  # nosec B615
             "openai/clip-vit-base-patch32",
             torch_dtype=(torch.float16 if torch.cuda.is_available() else torch.float32),
-            resume_download=True,
         ).to(device)
         self.clip_model.eval()
 
@@ -641,13 +638,10 @@ def _initialize_wav2vec_model(self, device: Any) -> None:
         torch = _get_torch()
 
         # HuggingFace model loaded by name; revision pinning managed operationally.
-        self.wav2vec_processor = Wav2Vec2Processor.from_pretrained(  # nosec B615
-            "facebook/wav2vec2-base-960h", resume_download=True
-        )
+        self.wav2vec_processor = Wav2Vec2Processor.from_pretrained("facebook/wav2vec2-base-960h")  # nosec B615
         self.wav2vec_model = Wav2Vec2Model.from_pretrained(  # nosec B615
             "facebook/wav2vec2-base-960h",
             torch_dtype=(torch.float16 if torch.cuda.is_available() else torch.float32),
-            resume_download=True,
         ).to(device)
         self.wav2vec_model.eval()
 

autobot-backend/code_embedding_generator.py

@@ -121,9 +121,9 @@ def _load_sync():
             from transformers import AutoModel, AutoTokenizer
 
             # HuggingFace model loaded by name; revision pinning managed operationally.
-            self.tokenizer = AutoTokenizer.from_pretrained(self.model_name, resume_download=True)  # nosec B615
+            self.tokenizer = AutoTokenizer.from_pretrained(self.model_name)  # nosec B615
             # HuggingFace model loaded by name; revision pinning managed operationally.
-            self.model = AutoModel.from_pretrained(self.model_name, resume_download=True)  # nosec B615
+            self.model = AutoModel.from_pretrained(self.model_name)  # nosec B615
 
             if self.npu_available:
                 self._convert_to_openvino()

autobot-backend/llm_shared/optimization/layer_inference.py

@@ -195,7 +195,7 @@ def load_model_config(self, model_name: str) -> Dict[str, Any]:
 
         logger.debug("Loading model config for %s", model_name)
         # HuggingFace model loaded by name; revision pinning managed operationally.
-        auto_cfg = AutoConfig.from_pretrained(model_name, resume_download=True, **kwargs)  # nosec B615
+        auto_cfg = AutoConfig.from_pretrained(model_name, **kwargs)  # nosec B615
         cfg_dict: Dict[str, Any] = auto_cfg.to_dict()
         logger.info(
             "Loaded model config for %s: arch=%s",
@@ -562,7 +562,7 @@ def _load_tokeniser(self, AutoTokenizer: Any) -> Any:
         if self._config.cache_dir:
             kwargs["cache_dir"] = self._config.cache_dir
         # HuggingFace model loaded by name; revision pinning managed operationally.
-        return AutoTokenizer.from_pretrained(self._config.model_name, resume_download=True, **kwargs)  # nosec B615
+        return AutoTokenizer.from_pretrained(self._config.model_name, **kwargs)  # nosec B615
 
     def _resolve_checkpoint_path(self) -> str:
         """Return the checkpoint path for the configured model.

autobot-backend/llm_shared/optimization/model_inspector.py

@@ -264,7 +264,7 @@ def _inspect_via_config(model_name: str) -> ModelInfo | None:
 
     try:
         # HuggingFace model loaded by name; revision pinning managed operationally.
-        cfg = transformers.AutoConfig.from_pretrained(model_name, resume_download=True)  # nosec B615
+        cfg = transformers.AutoConfig.from_pretrained(model_name)  # nosec B615
         param_count = _count_params_via_skeleton(cfg, transformers, accelerate)
         if param_count is None:
             logger.debug("model_inspector: using formula fallback for %s", model_name)

autobot-backend/multimodal_processor/processors/vision.py

@@ -100,19 +100,19 @@ def _load_models(self):
             # Load CLIP model for image embeddings and classification
             self.logger.info("Loading CLIP model...")
             self.clip_model = CLIPModel.from_pretrained(
-                "openai/clip-vit-base-patch32", resume_download=True
+                "openai/clip-vit-base-patch32"
             ).to(  # nosec B615  # HuggingFace model loaded by name; revision pinning managed operationally
                 self.device
             )
             self.clip_processor = CLIPProcessor.from_pretrained(  # nosec B615
-                "openai/clip-vit-base-patch32", use_fast=True, resume_download=True
+                "openai/clip-vit-base-patch32", use_fast=True
             )
 
             # Load BLIP-2 model for image captioning and VQA
             # Using smaller model for memory efficiency
             self.logger.info("Loading BLIP-2 model...")
             self.blip_processor = Blip2Processor.from_pretrained(  # nosec B615
-                "Salesforce/blip2-opt-2.7b", use_fast=True, resume_download=True
+                "Salesforce/blip2-opt-2.7b", use_fast=True
             )
 
             # Check if accelerate is available for device_map
@@ -128,13 +128,11 @@ def _load_models(self):
                     "Salesforce/blip2-opt-2.7b",
                     torch_dtype=torch.float16,
                     device_map="auto",
-                    resume_download=True,
                 )
             else:
                 self.blip_model = Blip2ForConditionalGeneration.from_pretrained(  # nosec B615
                     "Salesforce/blip2-opt-2.7b",
                     torch_dtype=(torch.float16 if torch.cuda.is_available() else torch.float32),
-                    resume_download=True,
                 ).to(self.device)
 
             # Set models to evaluation mode

autobot-backend/multimodal_processor/processors/voice.py

@@ -116,24 +116,20 @@ def _load_models(self):
             # Load Whisper model for speech recognition
             self.logger.info("Loading Whisper model...")
             # HuggingFace model loaded by name; revision pinning managed operationally.
-            self.whisper_processor = WhisperProcessor.from_pretrained(  # nosec B615
-                "openai/whisper-base", resume_download=True
-            )
+            self.whisper_processor = WhisperProcessor.from_pretrained("openai/whisper-base")  # nosec B615
             self.whisper_model = WhisperForConditionalGeneration.from_pretrained(  # nosec B615
                 "openai/whisper-base",
                 torch_dtype=(torch.float16 if torch.cuda.is_available() else torch.float32),
-                resume_download=True,
             ).to(self.device)
 
             # Load Wav2Vec2 model for audio embeddings and feature extraction
             self.logger.info("Loading Wav2Vec2 model...")
             self.wav2vec_processor = Wav2Vec2Processor.from_pretrained(  # nosec B615
-                "facebook/wav2vec2-base-960h", use_fast=True, resume_download=True
+                "facebook/wav2vec2-base-960h", use_fast=True
             )
             self.wav2vec_model = Wav2Vec2ForCTC.from_pretrained(  # nosec B615
                 "facebook/wav2vec2-base-960h",
                 torch_dtype=(torch.float16 if torch.cuda.is_available() else torch.float32),
-                resume_download=True,
             ).to(self.device)
 
             # Set models to evaluation mode

repo_tests/python_file_size_ratchet_baseline.py

@@ -51,7 +51,7 @@
     "autobot-backend/agents/system_command_agent.py": 903,
     "autobot-backend/agents/system_knowledge_manager.py": 843,
     "autobot-backend/agents/web_researcher.py": 1435,
-    "autobot-backend/ai_hardware_accelerator.py": 1048,
+    "autobot-backend/ai_hardware_accelerator.py": 1042,
     "autobot-backend/api/a2a.py": 650,
     "autobot-backend/api/advanced_control.py": 626,
     "autobot-backend/api/agent.py": 1380,