Upload merge_ministral.py with huggingface_hub

merge_ministral.py

@@ -0,0 +1,60 @@
+# /// script
+# dependencies = ["torch", "transformers>=5.0.0rc0", "peft", "accelerate", "huggingface_hub", "safetensors", "mistral-common>=1.8.6"]
+# ///
+
+import torch
+import os
+from peft import PeftModel
+from transformers import AutoModelForCausalLM, 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")
+
+print("
+[1/5] Loading base model...")
+base = AutoModelForCausalLM.from_pretrained(
+    "mistralai/Ministral-3-14B-Base-2512",
+    torch_dtype=torch.bfloat16,
+    device_map="auto",
+    trust_remote_code=True
+)
+print(f"Base model class: {base.__class__.__name__}")
+
+print("
+[2/5] Loading LoRA adapter...")
+model = PeftModel.from_pretrained(base, "RoleModel/ministral-14b-reasoning-merged")
+print("LoRA adapter loaded")
+
+print("
+[3/5] Merging weights...")
+merged = model.merge_and_unload()
+print("Merge complete")
+
+print("
+[4/5] 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)
+tok.save_pretrained("./merged-model")
+print("Tokenizer saved")
+
+print("
+[5/5] Pushing to Hub...")
+api = HfApi()
+api.create_repo("RoleModel/ministral-14b-merged-official", 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_type="model",
+    token=os.environ.get("HF_TOKEN")
+)
+
+print("
+=== DONE ===")
+print("Merged model: https://huggingface.co/RoleModel/ministral-14b-merged-official")