Update app.py

app.py

@@ -21,6 +21,7 @@ token = os.environ.get("HF_TOKEN")
 library_username = os.environ.get("OLLAMA_USERNAME").lower()
 ollama_pubkey = None
 ollama_model_name = None
+download_gguf_link = None
 # model.num_parameters()
 def regenerate_pubkey(pubkey, oauth_token: gr.OAuthToken | None):
     if oauth_token.token is None:
@@ -30,7 +31,7 @@ def regenerate_pubkey(pubkey, oauth_token: gr.OAuthToken | None):
         generate_ollama_host_file = f"echo $(ss -natp | grep (cat ollama.pid) | awk '{{print $4}}') > ollama.host"
         generate_ollama_pid_file = f"echo $! > ollama.pid"
         ollama_pubkey = f"cat {HOME}/{hash_oauth}/.ollama/id_ed25519.pub"  
-        ollama_start = f"HOME={HOME}/{hash_oauth} ollama serve & {ollama_pid_file} & sleep 5 & {ollama_host_file}"    
+        ollama_start = f"HOME={HOME}/{hash_oauth} ollama serve & {generate_ollama_pid_file} & sleep 5 & {generate_ollama_host_file}"    
         ollama_stop = f"kill -9 'cat ${HOME}/{hash_oauth}/ollama.pid'"
         delete_home = f"rm -Rf {HOME}/{hash_oauth}/.ollama"
             
@@ -63,9 +64,10 @@ def regenerate_pubkey(pubkey, oauth_token: gr.OAuthToken | None):
         return (f"Error: {e}", "error.png")
     finally:
         # shutil.rmtree(model_name, ignore_errors=True)
-        print("Ollama Pubkey Generated! Copy to your user profile in teh Ollama Library.")
+        print("Ollama Pubkey Generated! Copy to your user profile in the Ollama Library.")
 
 
+# def ollamafy_model(login, model_id, ollama_library_username , ollama_q_method, latest, download_gguf_link, maintainer, oauth_token: gr.OAuthToken | None, ollama_model_name):
 def ollamafy_model(login, model_id, ollama_library_username , ollama_q_method, latest, maintainer, oauth_token: gr.OAuthToken | None, ollama_model_name):
     ollama_library_username: library_username | None 
     
@@ -78,6 +80,7 @@ def ollamafy_model(login, model_id, ollama_library_username , ollama_q_method, l
     fp16 = f"{model_name}-fp16.gguf"
     ollama_stop = f"kill -9 'cat ${HOME}/{hash_oauth}/ollama.pid'"
     delete_home = f"rm -Rf {HOME}/{hash_oauth}/.ollama"
+    # download_gguf = f"wget download_gguf_link"
 
     try:
         api = HfApi(token=oauth_token.token)   
@@ -95,6 +98,7 @@ def ollamafy_model(login, model_id, ollama_library_username , ollama_q_method, l
         )
         dl_pattern += pattern
         
+        # if not os.path.isfile(fp16) and if download_gguf_link is None:         
         if not os.path.isfile(fp16):
             api.snapshot_download(repo_id=model_id, local_dir=model_name, local_dir_use_symlinks=False, allow_patterns=dl_pattern)
             print("Model downloaded successfully!")
@@ -111,22 +115,31 @@ def ollamafy_model(login, model_id, ollama_library_username , ollama_q_method, l
             print(f"Converted model path: {fp16}")
             
             HfApi().delete_repo(repo_id=model_id)
+        else:
+            if urlparse.urlparse(download_gguf_link).scheme:
+                result = subprocess.run(download_gguf, shell=True, capture_output=True)
+                print(result)
+                if result.returncode != 0:
+                    raise Exception(f"Error converting to fp16: {result.stderr}")
+                print("Model converted to fp16 successfully!")
+            else:
+                print("Invalid GGUF Download Link")
+                
 
         ### Ollamafy ###
         model_maintainer = model_id.split('/')[-2]
         if ollama_model_name is None:
             ollama_model_name = model_maintainer.lower() + '_' + model_name.lower() 
         ollama_modelfile_name = model_name + '_modelfile'
-        # model_path = f"{HOME}/.cache/huggingface/hub/{model_id}"
+        model_path = f"{HOME}/.cache/huggingface/hub/{model_id}"
       
         ollama_modelfile = open(ollama_modelfile_name, "w")
-        # ollama_modelfile_path = quantized_gguf_path
+        ollama_modelfile_path = quantized_gguf_path
         ollama_modelfile.write(quantized_gguf_path)
         ollama_modelfile.close()
         
         print(quantized_gguf_path)
 
-        # for ollama_q_method in ollama_q_methods:
         if ollama_q_method == "FP16":
             ollama_conversion = f"ollama create -f {model_file} {library_username}/{ollama_model_name}:{ollama_q_method.lower()}"
         else:
@@ -230,7 +243,12 @@ with gr.Blocks(css=css) as demo:
         placeholder="Search for model id on Huggingface",
         search_type="model",
     )
-
+    
+    download_gguf_link = gr.Textbox(
+        label="Download Link",
+        info="If you have access to the GGUF, you can insert the downlaod link here.",
+    )
+    
     ollama_q_method = gr.Dropdown(
         ["FP16", "Q3_K_S", "Q3_K_M", "Q3_K_L", "Q4_0", "Q4_1", "Q4_K_S", "Q4_K_M", "Q5_0", "Q5_1", "Q5_K_S", "Q5_K_M", "Q6_K", "Q8_0"],
         label="Ollama Quantization Method",
@@ -282,6 +300,7 @@ with gr.Blocks(css=css) as demo:
             generate_pubkey,
             model_id,
             ollama_model_name,
+            download_gguf_link,
             ollama_library_username,
             ollama_q_method,
             latest,