Upload convert_vision_model.py with huggingface_hub

convert_vision_model.py

@@ -35,11 +35,10 @@ import torch
 from transformers import AutoModel, AutoTokenizer, AutoProcessor
 from huggingface_hub import HfApi, hf_hub_download, login
 
-# Login with HF_TOKEN for private repo access
-hf_token = os.environ.get("HF_TOKEN")
-if hf_token:
-    login(token=hf_token)
-    print("Logged in to Hugging Face Hub")
+# Get HF_TOKEN for private repo access
+HF_TOKEN = os.environ.get("HF_TOKEN")
+if HF_TOKEN:
+    print(f"HF_TOKEN found (length: {len(HF_TOKEN)})")
 import subprocess
 import shutil
 import glob
@@ -86,6 +85,7 @@ if IS_LORA:
             model_name=MODEL_PATH,
             dtype=torch.float16,
             load_in_4bit=False,
+            token=HF_TOKEN,
         )
         print("  Loaded with unsloth FastModel")
 
@@ -103,8 +103,8 @@ if IS_LORA:
         from safetensors.torch import load_file
 
         # Download adapter weights
-        adapter_weights_path = hf_hub_download(MODEL_PATH, "adapter_model.safetensors")
-        adapter_config_path = hf_hub_download(MODEL_PATH, "adapter_config.json")
+        adapter_weights_path = hf_hub_download(MODEL_PATH, "adapter_model.safetensors", token=HF_TOKEN)
+        adapter_config_path = hf_hub_download(MODEL_PATH, "adapter_config.json", token=HF_TOKEN)
 
         with open(adapter_config_path) as f:
             adapter_config = json.load(f)
@@ -129,6 +129,7 @@ if IS_LORA:
                     torch_dtype=torch.float16,
                     device_map="cpu",  # Load on CPU first
                     trust_remote_code=True,
+                    token=HF_TOKEN,
                 )
                 print(f"  Base model loaded with {class_name}")
                 break
@@ -175,7 +176,7 @@ if IS_LORA:
     processor_saved = False
     for source in [MODEL_PATH, BASE_MODEL]:
         try:
-            processor = AutoProcessor.from_pretrained(source, trust_remote_code=True)
+            processor = AutoProcessor.from_pretrained(source, trust_remote_code=True, token=HF_TOKEN)
             processor.save_pretrained(merged_dir)
             print(f"  Processor saved from {source}")
             processor_saved = True
@@ -186,7 +187,7 @@ if IS_LORA:
     if not processor_saved:
         for source in [MODEL_PATH, BASE_MODEL]:
             try:
-                tokenizer = AutoTokenizer.from_pretrained(source, trust_remote_code=True)
+                tokenizer = AutoTokenizer.from_pretrained(source, trust_remote_code=True, token=HF_TOKEN)
                 tokenizer.save_pretrained(merged_dir)
                 print(f"  Tokenizer saved from {source}")
                 break
@@ -195,7 +196,7 @@ if IS_LORA:
 
     # Copy chat template if exists in adapter
     try:
-        chat_template_path = hf_hub_download(MODEL_PATH, "chat_template.jinja")
+        chat_template_path = hf_hub_download(MODEL_PATH, "chat_template.jinja", token=HF_TOKEN)
         shutil.copy(chat_template_path, f"{merged_dir}/chat_template.jinja")
         print("  Copied chat_template.jinja from adapter")
     except:
@@ -208,6 +209,7 @@ else:
         repo_id=MODEL_PATH,
         local_dir=merged_dir,
         local_dir_use_symlinks=False,
+        token=HF_TOKEN,
     )
     print(f"  Model downloaded to {merged_dir}")
 
@@ -318,7 +320,7 @@ for quant_type, desc in quant_formats:
 
 # Step 6: Upload to Hub
 print("\n[6/7] Uploading to Hugging Face Hub...")
-api = HfApi()
+api = HfApi(token=HF_TOKEN)
 
 # Upload all GGUF files
 for f in os.listdir(gguf_output_dir):