Upload merge_and_push.py

merge_and_push.py

@@ -3,6 +3,9 @@ Merge LoRA adapters into the base model and push to Hugging Face Hub.
 Run this after train.py completes successfully.
 """
 
+import os
+import sys
+
 import torch
 from transformers import AutoModelForCausalLM, AutoTokenizer
 from peft import PeftModel
@@ -19,10 +22,16 @@ HUB_MODEL_ID = "nraptisss/qwen2.5-7b-telecom-intent-merged"
 
 def merge_and_push():
     """Load adapters, merge, save locally, and push to hub."""
+    adapter_path = os.path.abspath(ADAPTER_PATH)
+    if not os.path.isdir(adapter_path):
+        print(f"ERROR: Adapter path not found: {adapter_path}")
+        print("Run train.py first to generate adapters.")
+        sys.exit(1)
+
     print(f"Loading base model: {BASE_MODEL}")
     base_model = AutoModelForCausalLM.from_pretrained(
         BASE_MODEL,
-        torch_dtype=torch.float16,
+        dtype=torch.float16,
         device_map="auto",
         trust_remote_code=True,
     )
@@ -34,8 +43,8 @@ def merge_and_push():
     if tokenizer.pad_token is None:
         tokenizer.pad_token = tokenizer.eos_token
 
-    print(f"Loading adapters from: {ADAPTER_PATH}")
-    model = PeftModel.from_pretrained(base_model, ADAPTER_PATH)
+    print(f"Loading adapters from: {adapter_path}")
+    model = PeftModel.from_pretrained(base_model, adapter_path)
 
     print("Merging adapters into base model...")
     model = model.merge_and_unload()