Upload train_ministral_n8n.py with huggingface_hub

train_ministral_n8n.py

@@ -157,9 +157,31 @@ trainer = SFTTrainer(
     processing_class=tokenizer,
 )
 
-# Train
+# Train - resume from checkpoint if available on Hub
 print("Starting training...")
-trainer.train()
+# Try to resume from Hub checkpoint
+try:
+    from huggingface_hub import hf_hub_download, list_repo_files
+    files = list_repo_files(OUTPUT_MODEL)
+    if "last-checkpoint" in str(files) or "adapter_model.safetensors" in files:
+        print(f"Found existing checkpoint on Hub, downloading to resume...")
+        # Download checkpoint files
+        import os
+        os.makedirs("./resume-checkpoint", exist_ok=True)
+        for f in ["adapter_model.safetensors", "adapter_config.json", "trainer_state.json", "training_args.bin"]:
+            try:
+                hf_hub_download(OUTPUT_MODEL, f, local_dir="./resume-checkpoint")
+            except:
+                pass
+        if os.path.exists("./resume-checkpoint/trainer_state.json"):
+            trainer.train(resume_from_checkpoint="./resume-checkpoint")
+        else:
+            trainer.train()
+    else:
+        trainer.train()
+except Exception as e:
+    print(f"Could not resume from checkpoint: {e}, starting fresh...")
+    trainer.train()
 
 # Save final model
 print("Saving final model...")