Upload merge_ministral.py with huggingface_hub

merge_ministral.py

@@ -1,54 +1,69 @@
 # /// script
-# dependencies = ["torch", "transformers>=5.0.0rc0", "peft", "accelerate", "huggingface_hub", "safetensors", "mistral-common>=1.8.6"]
+# dependencies = [
+#     "torch",
+#     "transformers>=5.0.0rc0",
+#     "peft",
+#     "accelerate",
+#     "huggingface_hub",
+#     "safetensors",
+#     "mistral-common>=1.8.6",
+# ]
 # ///
 
+"""
+Merge Ministral 14B LoRA adapter with the official Mistral base model.
+Run on HuggingFace Jobs with: hf jobs uv run --flavor a10g-large --timeout 2h --secrets HF_TOKEN merge_ministral_official.py
+"""
+
 import torch
 import os
 from peft import PeftModel
-from transformers import AutoModel, AutoTokenizer
+from transformers import Mistral3ForConditionalGeneration, AutoTokenizer
 from huggingface_hub import HfApi
 
-print("=== Ministral 14B LoRA Merge ===")
-print("Base: mistralai/Ministral-3-14B-Base-2512")
-print("LoRA: RoleModel/ministral-14b-reasoning-merged")
-print("Output: RoleModel/ministral-14b-merged-official")
+BASE_MODEL = "mistralai/Ministral-3-14B-Base-2512"
+LORA_ADAPTER = "RoleModel/ministral-14b-reasoning-merged"
+OUTPUT_REPO = os.environ.get("OUTPUT_REPO", "RoleModel/ministral-14b-merged-official")
+
+print(f"Base model: {BASE_MODEL}")
+print(f"LoRA adapter: {LORA_ADAPTER}")
+print(f"Output repo: {OUTPUT_REPO}")
 
-print("[1/5] Loading base model...")
-# Use AutoModel since this is a vision-language model (Pixtral/Mistral3)
-base = AutoModel.from_pretrained(
-    "mistralai/Ministral-3-14B-Base-2512",
+print("\n=== Loading base model ===")
+base = Mistral3ForConditionalGeneration.from_pretrained(
+    BASE_MODEL,
     torch_dtype=torch.bfloat16,
     device_map="auto",
     trust_remote_code=True
 )
-print(f"Base model class: {base.__class__.__name__}")
+print(f"Base model loaded: {base.__class__.__name__}")
 
-print("[2/5] Loading LoRA adapter...")
-model = PeftModel.from_pretrained(base, "RoleModel/ministral-14b-reasoning-merged")
+print("\n=== Loading LoRA adapter ===")
+model = PeftModel.from_pretrained(base, LORA_ADAPTER)
 print("LoRA adapter loaded")
 
-print("[3/5] Merging weights...")
+print("\n=== Merging weights ===")
 merged = model.merge_and_unload()
 print("Merge complete")
 
-print("[4/5] Saving merged model...")
+print("\n=== Saving merged model ===")
 merged.save_pretrained("./merged-model", safe_serialization=True)
 print("Model saved locally")
 
-print("[4b/5] Saving tokenizer...")
-tok = AutoTokenizer.from_pretrained("mistralai/Ministral-3-14B-Base-2512", trust_remote_code=True)
+print("\n=== Saving tokenizer ===")
+tok = AutoTokenizer.from_pretrained(BASE_MODEL, trust_remote_code=True)
 tok.save_pretrained("./merged-model")
 print("Tokenizer saved")
 
-print("[5/5] Pushing to Hub...")
+print(f"\n=== Pushing to Hub: {OUTPUT_REPO} ===")
 api = HfApi()
-api.create_repo("RoleModel/ministral-14b-merged-official", exist_ok=True, private=True, token=os.environ.get("HF_TOKEN"))
+api.create_repo(OUTPUT_REPO, exist_ok=True, private=True, token=os.environ.get("HF_TOKEN"))
 api.upload_folder(
     folder_path="./merged-model",
-    repo_id="RoleModel/ministral-14b-merged-official",
+    repo_id=OUTPUT_REPO,
     repo_type="model",
     token=os.environ.get("HF_TOKEN")
 )
 
-print("=== DONE ===")
-print("Merged model: https://huggingface.co/RoleModel/ministral-14b-merged-official")
+print(f"\n=== DONE ===")
+print(f"Merged model available at: https://huggingface.co/{OUTPUT_REPO}")