Update app.py

app.py

@@ -4,53 +4,70 @@ import subprocess
 import signal
 os.environ["GRADIO_ANALYTICS_ENABLED"] = "False"
 import gradio as gr
+
 from huggingface_hub import create_repo, HfApi
 from huggingface_hub import snapshot_download
 from huggingface_hub import whoami
 from huggingface_hub import ModelCard
+
 from gradio_huggingfacehub_search import HuggingfaceHubSearch
+
 from apscheduler.schedulers.background import BackgroundScheduler
+
 from textwrap import dedent
 
 HF_TOKEN = os.environ.get("HF_TOKEN")
 
 def generate_importance_matrix(model_path, train_data_path):
     imatrix_command = f"./llama-imatrix -m ../{model_path} -f {train_data_path} -ngl 99 --output-frequency 10"
+
     os.chdir("llama.cpp")
+
     print(f"Current working directory: {os.getcwd()}")
     print(f"Files in the current directory: {os.listdir('.')}")
+
     if not os.path.isfile(f"../{model_path}"):
         raise Exception(f"Model file not found: {model_path}")
+
     print("Running imatrix command...")
     process = subprocess.Popen(imatrix_command, shell=True)
+
     try:
-        process.wait(timeout=60)
+        process.wait(timeout=60)  # added wait
     except subprocess.TimeoutExpired:
         print("Imatrix computation timed out. Sending SIGINT to allow graceful termination...")
         process.send_signal(signal.SIGINT)
         try:
-            process.wait(timeout=5)
+            process.wait(timeout=5)  # grace period
         except subprocess.TimeoutExpired:
             print("Imatrix proc still didn't term. Forecfully terming process...")
             process.kill()
+
     os.chdir("..")
+
     print("Importance matrix generation completed.")
 
 def split_upload_model(model_path, repo_id, oauth_token: gr.OAuthToken | None, split_max_tensors=256, split_max_size=None):
     if oauth_token.token is None:
         raise ValueError("You have to be logged in.")
+    
     split_cmd = f"llama.cpp/llama-gguf-split --split --split-max-tensors {split_max_tensors}"
     if split_max_size:
         split_cmd += f" --split-max-size {split_max_size}"
-    split_cmd += f" {model_path} {os.path.splitext(model_path)[0]}"
-    print(f"Split command: {split_cmd}")
+    split_cmd += f" {model_path} {model_path.split('.')[0]}"
+    
+    print(f"Split command: {split_cmd}") 
+    
     result = subprocess.run(split_cmd, shell=True, capture_output=True, text=True)
-    print(f"Split command stdout: {result.stdout}")
-    print(f"Split command stderr: {result.stderr}")
+    print(f"Split command stdout: {result.stdout}") 
+    print(f"Split command stderr: {result.stderr}") 
+    
     if result.returncode != 0:
         raise Exception(f"Error splitting the model: {result.stderr}")
     print("Model split successfully!")
-    sharded_model_files = [f for f in os.listdir('.') if f.startswith(os.path.splitext(model_path)[0])]
+     
+    
+    sharded_model_files = [f for f in os.listdir('.') if f.startswith(model_path.split('.')[0])]
     if sharded_model_files:
         print(f"Sharded model files: {sharded_model_files}")
         api = HfApi(token=oauth_token.token)
@@ -67,45 +84,47 @@ def split_upload_model(model_path, repo_id, oauth_token: gr.OAuthToken | None, s
                 raise Exception(f"Error uploading file {file_path}: {e}")
     else:
         raise Exception("No sharded files found.")
+    
     print("Sharded model has been uploaded successfully!")
 
 def process_model(model_id, q_method, use_imatrix, imatrix_q_method, private_repo, train_data_file, split_model, split_max_tensors, split_max_size, oauth_token: gr.OAuthToken | None):
     if oauth_token.token is None:
         raise ValueError("You must be logged in to use GGUF-my-repo")
     model_name = model_id.split('/')[-1]
-    model_file = None
+    fp16 = f"{model_name}.fp16.gguf"
 
     try:
         api = HfApi(token=oauth_token.token)
 
-        # Download only necessary files based on model format
-        dl_pattern = ["*.md", "*.json", "*.safetensors", "*.bin", "*.pkl", "*.ckp", "*.pth", "*.gguf"]
+        dl_pattern = ["*.md", "*.json", "*.model"]
+
+        pattern = (
+            "*.safetensors"
+            if any(
+                file.path.endswith(".safetensors")
+                for file in api.list_repo_tree(
+                    repo_id=model_id,
+                    recursive=True,
+                )
+            )
+            else "*.bin"
+        )
+
+        dl_pattern += pattern
+
         api.snapshot_download(repo_id=model_id, local_dir=model_name, local_dir_use_symlinks=False, allow_patterns=dl_pattern)
         print("Model downloaded successfully!")
         print(f"Current working directory: {os.getcwd()}")
         print(f"Model directory contents: {os.listdir(model_name)}")
 
-        # Find downloaded model file
-        supported_extensions = (".bin", ".safetensors", ".pkl", ".ckp", ".pth", ".gguf")
-        for filename in os.listdir(model_name):
-            if any(filename.endswith(ext) for ext in supported_extensions):
-                model_file = os.path.join(model_name, filename)
-                break
-
-        if model_file is None:
-            raise FileNotFoundError(f"No supported model file found in the downloaded files. Supported formats: {', '.join(supported_extensions)}")
-
-        # If the model is not already in GGUF format, convert it
-        if not model_file.endswith(".gguf"):
-            gguf_model_file = f"{os.path.splitext(model_file)[0]}.gguf"
-            conversion_command = f"python llama.cpp/convert_hf_to_gguf.py {model_file} --outfile {gguf_model_file}"
-            result = subprocess.run(conversion_command, shell=True, capture_output=True)
-            if result.returncode != 0:
-                raise Exception(f"Error converting to GGUF: {result.stderr}")
-            print("Model converted to GGUF successfully!")
-            print(f"Converted model path: {gguf_model_file}")
-        else:
-            gguf_model_file = model_file  # If already GGUF, use the original file
+        conversion_script = "convert_hf_to_gguf.py"
+        fp16_conversion = f"python llama.cpp/{conversion_script} {model_name} --outtype f16 --outfile {fp16}"
+        result = subprocess.run(fp16_conversion, 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!")
+        print(f"Converted model path: {fp16}")
 
         imatrix_path = "llama.cpp/imatrix.dat"
 
@@ -120,27 +139,23 @@ def process_model(model_id, q_method, use_imatrix, imatrix_q_method, private_rep
             if not os.path.isfile(train_data_path):
                 raise Exception(f"Training data file not found: {train_data_path}")
 
-            generate_importance_matrix(gguf_model_file, train_data_path)
+            generate_importance_matrix(fp16, train_data_path)
         else:
             print("Not using imatrix quantization.")
-
         username = whoami(oauth_token.token)["name"]
         quantized_gguf_name = f"{model_name.lower()}-{imatrix_q_method.lower()}-imat.gguf" if use_imatrix else f"{model_name.lower()}-{q_method.lower()}.gguf"
         quantized_gguf_path = quantized_gguf_name
-
-        os.chdir("llama.cpp")
         if use_imatrix:
-            quantise_ggml = f"./llama-quantize --imatrix {imatrix_path} ../{gguf_model_file} ../{quantized_gguf_path} {imatrix_q_method}"
+            quantise_ggml = f"./llama.cpp/llama-quantize --imatrix {imatrix_path} {fp16} {quantized_gguf_path} {imatrix_q_method}"
         else:
-            quantise_ggml = f"./llama-quantize ../{gguf_model_file} ../{quantized_gguf_path} {q_method}"
+            quantise_ggml = f"./llama.cpp/llama-quantize {fp16} {quantized_gguf_path} {q_method}"
         result = subprocess.run(quantise_ggml, shell=True, capture_output=True)
-        os.chdir("..")
-
         if result.returncode != 0:
             raise Exception(f"Error quantizing: {result.stderr}")
         print(f"Quantized successfully with {imatrix_q_method if use_imatrix else q_method} option!")
         print(f"Quantized model path: {quantized_gguf_path}")
 
+        # Create empty repo
         new_repo_url = api.create_repo(repo_id=f"{username}/{model_name}-{imatrix_q_method if use_imatrix else q_method}-GGUF", exist_ok=True, private=private_repo)
         new_repo_id = new_repo_url.repo_id
         print("Repo created successfully!", new_repo_url)
@@ -159,26 +174,26 @@ def process_model(model_id, q_method, use_imatrix, imatrix_q_method, private_rep
             # {new_repo_id}
             This model was converted to GGUF format from [`{model_id}`](https://huggingface.co/{model_id}) using llama.cpp via the ggml.ai's [GGUF-my-repo](https://huggingface.co/spaces/ggml-org/gguf-my-repo) space.
             Refer to the [original model card](https://huggingface.co/{model_id}) for more details on the model.
-
+            
             ## Use with llama.cpp
             Install llama.cpp through brew (works on Mac and Linux)
-
+            
             ```bash
             brew install llama.cpp
-
+            
             ```
             Invoke the llama.cpp server or the CLI.
-
+            
             ### CLI:
             ```bash
             llama-cli --hf-repo {new_repo_id} --hf-file {quantized_gguf_name} -p "The meaning to life and the universe is"
             ```
-
+            
             ### Server:
             ```bash
             llama-server --hf-repo {new_repo_id} --hf-file {quantized_gguf_name} -c 2048
             ```
-
+            
             Note: You can also use this checkpoint directly through the [usage steps](https://github.com/ggerganov/llama.cpp?tab=readme-ov-file#usage) listed in the Llama.cpp repo as well.
             Step 1: Clone llama.cpp from GitHub.
             ```
@@ -192,7 +207,7 @@ def process_model(model_id, q_method, use_imatrix, imatrix_q_method, private_rep
             ```
             ./llama-cli --hf-repo {new_repo_id} --hf-file {quantized_gguf_name} -p "The meaning to life and the universe is"
             ```
-            or
+            or 
             ```
             ./llama-server --hf-repo {new_repo_id} --hf-file {quantized_gguf_name} -c 2048
             ```
@@ -212,8 +227,10 @@ def process_model(model_id, q_method, use_imatrix, imatrix_q_method, private_rep
                 )
             except Exception as e:
                 raise Exception(f"Error uploading quantized model: {e}")
-
-        if use_imatrix and os.path.isfile(imatrix_path):
+        
+        
+        imatrix_path = "llama.cpp/imatrix.dat"
+        if os.path.isfile(imatrix_path):
             try:
                 print(f"Uploading imatrix.dat: {imatrix_path}")
                 api.upload_file(
@@ -244,7 +261,8 @@ def process_model(model_id, q_method, use_imatrix, imatrix_q_method, private_rep
 css="""/* Custom CSS to allow scrolling */
 .gradio-container {overflow-y: auto;}
 """
-with gr.Blocks(css=css) as demo:
+# Create Gradio interface
+with gr.Blocks(css=css) as demo: 
     gr.Markdown("You must be logged in to use GGUF-my-repo.")
     gr.LoginButton(min_width=250)
 
@@ -267,7 +285,7 @@ with gr.Blocks(css=css) as demo:
         ["IQ3_M", "IQ3_XXS", "Q4_K_M", "Q4_K_S", "IQ4_NL", "IQ4_XS", "Q5_K_M", "Q5_K_S"],
         label="Imatrix Quantization Method",
         info="GGML imatrix quants type",
-        value="IQ4_NL",
+        value="IQ4_NL", 
         filterable=False,
         visible=False
     )
@@ -311,7 +329,7 @@ with gr.Blocks(css=css) as demo:
 
     def update_visibility(use_imatrix):
         return gr.update(visible=not use_imatrix), gr.update(visible=use_imatrix), gr.update(visible=use_imatrix)
-
+    
     use_imatrix.change(
         fn=update_visibility,
         inputs=use_imatrix,
@@ -335,7 +353,7 @@ with gr.Blocks(css=css) as demo:
             gr.Markdown(label="output"),
             gr.Image(show_label=False),
         ],
-        title="Create your own GGUF Quants, blazingly fast !",
+        title="Create your own GGUF Quants, blazingly fast ⚡!",
         description="The space takes an HF repo as an input, quantizes it and creates a Public repo containing the selected quant under your HF user namespace.",
         api_name=False
     )
@@ -350,10 +368,11 @@ with gr.Blocks(css=css) as demo:
     )
 
 def restart_space():
-    HfApi().restart_space(repo_id="Ffftdtd5dtft/gguf-my-repo", token=HF_TOKEN, factory_reboot=True)
+    HfApi().restart_space(repo_id="ggml-org/gguf-my-repo", token=HF_TOKEN, factory_reboot=True)
 
 scheduler = BackgroundScheduler()
 scheduler.add_job(restart_space, "interval", seconds=21600)
 scheduler.start()
 
-demo.queue(default_concurrency_limit=2, max_size=5).launch(debug=True, show_api=False)
+# Launch the interface
+demo.queue(default_concurrency_limit=999, max_size=5).launch(debug=True, show_api=False)