fix

app.py

@@ -15,7 +15,6 @@ from apscheduler.schedulers.background import BackgroundScheduler
 from textwrap import dedent
 
 HF_TOKEN = os.environ.get("HF_TOKEN")
-items_to_keep = ['build', 'llama.cpp' 'venv', 'app.py', 'start.sh', 'Dockerfile', 'README.md']
 
 
 class bcolors:
@@ -101,6 +100,7 @@ def process_lora(model_id, lora_id, merged_name, methods, private_repo, oauth_to
     lora_name = lora_id.split('/')[-1]
     model_fp16 = f"{model_name}-f16.gguf"
     lora_fp16 = f"{lora_name}-fp16.gguf"
+    items_to_remove = []
 
     try:
         api = HfApi(token=oauth_token)
@@ -127,12 +127,14 @@ def process_lora(model_id, lora_id, merged_name, methods, private_repo, oauth_to
 
         # Download Raw Model from HF
         api.snapshot_download(repo_id=model_id, local_dir=model_name, allow_patterns=dl_pattern)
+        items_to_remove.append(model_name)
         print("Model downloaded successfully!")
         print(f"{bcolors.OKGREEN}Files in current working directory: {os.listdir(current_directory)} {bcolors.ENDC}")
         print(f"{bcolors.OKBLUE}Model directory contents: {os.listdir(model_name)} {bcolors.ENDC}")
 
         # Download LoRA adapter from HF
         api.snapshot_download(repo_id=lora_id, local_dir=lora_name, allow_patterns=dl_pattern)
+        items_to_remove.append(lora_name)
         print("LoRA downloaded successfully!")
         print(f"{bcolors.OKGREEN}Files in current working directory: {os.listdir(current_directory)} {bcolors.ENDC}")
         print(f"{bcolors.OKBLUE}LoRA directory contents: {os.listdir(lora_name)} {bcolors.ENDC}")
@@ -141,6 +143,7 @@ def process_lora(model_id, lora_id, merged_name, methods, private_repo, oauth_to
         lora_conversion_script = "convert_lora_to_gguf.py"
         lora_fp16_conversion = f"python llama.cpp/{lora_conversion_script} --base {model_name} {lora_name} --outtype f16 --outfile {lora_fp16}"
         lora_result = subprocess.run(lora_fp16_conversion, shell=True, capture_output=True)
+        items_to_remove.append(lora_fp16)
         print(lora_result)
         if lora_result.returncode != 0:
             raise Exception(f"Error converting to fp16: {lora_result.stderr}")
@@ -162,6 +165,7 @@ def process_lora(model_id, lora_id, merged_name, methods, private_repo, oauth_to
         base_conversion_script = "convert_hf_to_gguf.py"
         base_fp16_conversion = f"python llama.cpp/{base_conversion_script} {model_name} --outtype f16 --outfile {model_fp16}"
         base_result = subprocess.run(base_fp16_conversion, shell=True, capture_output=True)
+        items_to_remove.append(model_fp16)
         print(base_result)
         if base_result.returncode != 0:
             raise Exception(f"Error converting to fp16: {base_result.stderr}")
@@ -172,31 +176,39 @@ def process_lora(model_id, lora_id, merged_name, methods, private_repo, oauth_to
         # Clean storage: hf-model & hf-lora
         shutil.rmtree(model_name, ignore_errors=True)
         shutil.rmtree(lora_name, ignore_errors=True)
+        items_to_remove.remove(model_name)
+        items_to_remove.remove(lora_name)
         print("HF model & LoRA cleaned up successfully!")
         print(f"{bcolors.OKGREEN}Files in current working directory: {os.listdir(current_directory)} {bcolors.ENDC}")
 
         # Merge LoRA with Raw Model to GGUF-fp16
         print(f"Merging LoRA with Model => fp16")
         merged_fp16 = export_lora_to_gguf(model_fp16, lora_fp16, merged_name)
+        items_to_remove.append(f"{merged_fp16}-fp16.gguf")
         print(f"{bcolors.OKGREEN}Files in current working directory: {os.listdir(current_directory)} {bcolors.ENDC}")
         # Upload Merged GGUF-fp16 to HF
         upload_file_to_hf(f"{merged_name}-fp16.gguf", new_repo_id, api)
         # Remove LoRA-GGUF & Model-GGUF
         os.remove(model_fp16)
         os.remove(lora_fp16)
+        items_to_remove.remove(model_fp16)
+        items_to_remove.remove(lora_fp16)
 
         if methods is not None:
             # Quantize GGUF-fp16 one by one
             for method in methods:
                 print(f"Quantizing merged fp16-gguf to {method}")
                 quantized_name = quantize_merged_gguf(merged_fp16, method)
+                items_to_remove.append(quantized_name)
                 print(f"{bcolors.OKGREEN}Files in current working directory: {os.listdir(current_directory)} {bcolors.ENDC}")
                 upload_file_to_hf(quantized_name, new_repo_id, api)
                 os.remove(quantized_name)
+                items_to_remove.remove(quantized_name)
                 print("Removed the uploaded model.")
                 print(f"{bcolors.OKGREEN}Files in current working directory: {os.listdir(current_directory)} {bcolors.ENDC}")
 
         os.remove(f"{merged_fp16}-fp16.gguf")
+        items_to_remove.remove(f"{merged_fp16}-fp16.gguf")
         print("Remove the fp16 GGUF file.")
         print(f"{bcolors.OKGREEN}Files in current working directory: {os.listdir(current_directory)} {bcolors.ENDC}")
 
@@ -210,10 +222,10 @@ def process_lora(model_id, lora_id, merged_name, methods, private_repo, oauth_to
         all_items = os.listdir(current_directory)
         for item in all_items:
             item_path = os.path.join(current_directory, item)
-            if os.path.isfile(item_path) and item not in items_to_keep:
+            if os.path.isfile(item_path) and item in items_to_remove:
                 os.remove(item_path)
                 print(f"Delete file: {item_path}")
-            elif os.path.isdir(item_path) and item not in items_to_keep:
+            elif os.path.isdir(item_path) and item in items_to_remove:
                 shutil.rmtree(item_path, ignore_errors=True)
                 print(f"Delete folder: {item_path}")
         print("Folder cleaned up successfully!")