Update app.py

app.py

@@ -2,10 +2,18 @@ import os
 import shutil
 import subprocess
 import signal
+os.environ["GRADIO_ANALYTICS_ENABLED"] = "False"
 import gradio as gr
-from huggingface_hub import create_repo, HfApi, snapshot_download, whoami, ModelCard
+
+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")
@@ -32,7 +40,7 @@ def generate_importance_matrix(model_path, train_data_path):
         try:
             process.wait(timeout=5)  # grace period
         except subprocess.TimeoutExpired:
-            print("Imatrix proc still didn't term. Forcefully terminating process...")
+            print("Imatrix proc still didn't term. Forecfully terming process...")
             process.kill()
 
     os.chdir("..")
@@ -78,7 +86,7 @@ def split_upload_model(model_path, repo_id, oauth_token: gr.OAuthToken | None, s
     
     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.oauth.OAuthToken | None):
+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]
@@ -87,8 +95,11 @@ def process_model(model_id, q_method, use_imatrix, imatrix_q_method, private_rep
     try:
         api = HfApi(token=oauth_token.token)
 
-        dl_pattern = ["*.md", "*.json", "*.model"]
-        model_types = ["*.safetensors", "*.bin", "*.pt", "*.onnx", "*.h5", "*.tflite", "*.ckpt", "*.pb", "*.tar", "*.xml", "*.caffemodel"]
+        dl_pattern = [
+            "*.safetensors", "*.bin", "*.pt", "*.onnx", "*.h5", "*.tflite",
+            "*.ckpt", "*.pb", "*.tar", "*.xml", "*.caffemodel",
+            "*.md", "*.json", "*.model"
+        ]
 
         pattern = (
             "*.safetensors"
@@ -103,7 +114,6 @@ def process_model(model_id, q_method, use_imatrix, imatrix_q_method, private_rep
         )
 
         dl_pattern += pattern
-        dl_pattern += model_types
 
         api.snapshot_download(repo_id=model_id, local_dir=model_name, local_dir_use_symlinks=False, allow_patterns=dl_pattern)
         print("Model downloaded successfully!")
@@ -125,7 +135,7 @@ def process_model(model_id, q_method, use_imatrix, imatrix_q_method, private_rep
             if train_data_file:
                 train_data_path = train_data_file.name
             else:
-                train_data_path = "groups_merged.txt"  # fallback calibration dataset
+                train_data_path = "groups_merged.txt" # fallback calibration dataset
 
             print(f"Training data file path: {train_data_path}")
 
@@ -135,13 +145,16 @@ def process_model(model_id, q_method, use_imatrix, imatrix_q_method, private_rep
             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
+        
         if use_imatrix:
             quantise_ggml = f"./llama.cpp/llama-quantize --imatrix {imatrix_path} {fp16} {quantized_gguf_path} {imatrix_q_method}"
         else:
             quantise_ggml = f"./llama.cpp/llama-quantize {fp16} {quantized_gguf_path} {q_method}"
+        
         result = subprocess.run(quantise_ggml, shell=True, capture_output=True)
         if result.returncode != 0:
             raise Exception(f"Error quantizing: {result.stderr}")
@@ -196,77 +209,175 @@ def process_model(model_id, q_method, use_imatrix, imatrix_q_method, private_rep
             ```
             cd llama.cpp && LLAMA_CURL=1 make
             ```
-            Step 3: Fetch model weights from HF using curl command and run the models directly!
+            Step 3: Run inference through the main binary.
             ```
-            curl -L https://huggingface.co/{new_repo_id}/resolve/main/{quantized_gguf_name} -o ./models/{quantized_gguf_name}
-            ./llama -m ./models/{quantized_gguf_name} -p "Hello, world!"
+            ./llama-cli --hf-repo {new_repo_id} --hf-file {quantized_gguf_name} -p "The meaning to life and the universe is"
+            ```
+            or 
+            ```
+            ./llama-server --hf-repo {new_repo_id} --hf-file {quantized_gguf_name} -c 2048
             ```
-            
-            ## Additional Notes:
-            To gain higher performance, ensure that you have aligned on llama.cpp's threading tips by having your CPU fully utilized and setting threads dynamically using `OMP_NUM_THREADS`.
             """
         )
-        card.push_to_hub(repo_id=new_repo_id, token=oauth_token.token)
+        card.save(f"README.md")
+
+        if split_model:
+            split_upload_model(quantized_gguf_path, new_repo_id, oauth_token, split_max_tensors, split_max_size)
+        else:
+            try:
+                print(f"Uploading quantized model: {quantized_gguf_path}")
+                api.upload_file(
+                    path_or_fileobj=quantized_gguf_path,
+                    path_in_repo=quantized_gguf_name,
+                    repo_id=new_repo_id,
+                )
+            except Exception as e:
+                raise Exception(f"Error uploading quantized model: {e}")
+        
+        imatrix_path = "llama.cpp/imatrix.dat"
+        if os.path.isfile(imatrix_path):
+            try:
+                print(f"Uploading imatrix.dat: {imatrix_path}")
+                api.upload_file(
+                    path_or_fileobj=imatrix_path,
+                    path_in_repo="imatrix.dat",
+                    repo_id=new_repo_id,
+                )
+            except Exception as e:
+                raise Exception(f"Error uploading imatrix.dat: {e}")
+
         api.upload_file(
-            path_or_fileobj=quantized_gguf_path,
-            path_in_repo=quantized_gguf_name,
+            path_or_fileobj=f"README.md",
+            path_in_repo=f"README.md",
             repo_id=new_repo_id,
         )
-        if split_model:
-            split_upload_model(quantized_gguf_path, new_repo_id, oauth_token, split_max_tensors=split_max_tensors, split_max_size=split_max_size)
-        else:
-            print("Model split skipped by user.")
+        print(f"Uploaded successfully with {imatrix_q_method if use_imatrix else q_method} option!")
 
-        print("Model has been uploaded successfully!")
+        return (
+            f'Find your repo <a href=\'{new_repo_url}\' target="_blank" style="text-decoration:underline">here</a>',
+            "llama.png",
+        )
     except Exception as e:
-        print(f"An error occurred: {str(e)}")
-        return False, str(e)
+        return (f"Error: {e}", "error.png")
     finally:
-        if os.path.exists(fp16):
-            os.remove(fp16)
-        if os.path.exists(quantized_gguf_path):
-            os.remove(quantized_gguf_path)
-        shutil.rmtree(model_name)
-        print(f"Removed temporary files for model {model_name}")
-
-    return True, None
-
-def app_interface():
-    with gr.Blocks() as demo:
-        gr.Markdown("## GGUF Model Processing")
-        
-        with gr.Row():
-            with gr.Column():
-                repo_id = gr.Textbox(label="HuggingFace Repo ID")
-                model_id = gr.Textbox(label="Model ID")
-                q_method = gr.Dropdown(["q4_0", "q4_1", "q5_0", "q5_1", "q8_0"], label="Quantization Method")
-                imatrix_q_method = gr.Dropdown(["q4_0", "q4_1", "q5_0", "q5_1", "q8_0"], label="Imatrix Quantization Method")
-                use_imatrix = gr.Checkbox(label="Use Importance Matrix")
-                private_repo = gr.Checkbox(label="Private Repo")
-                train_data_file = gr.File(label="Training Data File (Optional)")
-                split_model = gr.Checkbox(label="Split Model")
-                split_max_tensors = gr.Number(label="Max Tensors per Shard", value=256)
-                split_max_size = gr.Number(label="Max Shard Size (MB)", value=None)
-            with gr.Column():
-                oauth_token = gr.oauth.HuggingFace(
-                    "Gradio OAuth Authentication",
-                    token=HF_TOKEN,
-                )
-        
-        process_btn = gr.Button("Process Model")
-        process_btn.click(
-            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],
-            outputs=["status_text"]
-        )
-        
-    return demo
+        shutil.rmtree(model_name, ignore_errors=True)
+        print("Folder cleaned up successfully!")
+
+css="""/* Custom CSS to allow scrolling */
+.gradio-container {overflow-y: auto;}
+"""
+# 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)
+
+    model_id = HuggingfaceHubSearch(
+        label="Hub Model ID",
+        placeholder="Search for model id on Huggingface",
+        search_type="model",
+    )
+
+    q_method = gr.Dropdown(
+        ["Q2_K", "Q3_K_S", "Q3_K_M", "Q3_K_L", "Q4_0", "Q4_K_S", "Q4_K_M", "Q5_0", "Q5_K_S", "Q5_K_M", "Q6_K", "Q8_0"],
+        label="Quantization Method",
+        info="GGML quantization type",
+        value="Q4_K_M",
+        filterable=False,
+        visible=True
+    )
+
+    imatrix_q_method = gr.Dropdown(
+        ["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", 
+        filterable=False,
+        visible=False
+    )
+
+    use_imatrix = gr.Checkbox(
+        value=False,
+        label="Use Imatrix Quantization",
+        info="Use importance matrix for quantization."
+    )
+
+    private_repo = gr.Checkbox(
+        value=False,
+        label="Private Repo",
+        info="Create a private repo under your username."
+    )
+
+    train_data_file = gr.File(
+        label="Training Data File",
+        file_types=["txt"],
+        visible=False
+    )
+
+    split_model = gr.Checkbox(
+        value=False,
+        label="Split Model",
+        info="Shard the model using gguf-split."
+    )
+
+    split_max_tensors = gr.Number(
+        value=256,
+        label="Max Tensors per File",
+        info="Maximum number of tensors per file when splitting model.",
+        visible=False
+    )
+
+    split_max_size = gr.Textbox(
+        label="Max File Size",
+        info="Maximum file size when splitting model (--split-max-size). May leave empty to use the default.",
+        visible=False
+    )
+
+    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,
+        outputs=[q_method, imatrix_q_method, train_data_file]
+    )
+
+    iface = gr.Interface(
+        fn=process_model,
+        inputs=[
+            model_id,
+            q_method,
+            use_imatrix,
+            imatrix_q_method,
+            private_repo,
+            train_data_file,
+            split_model,
+            split_max_tensors,
+            split_max_size,
+        ],
+        outputs=[
+            gr.Markdown(label="output"),
+            gr.Image(show_label=False),
+        ],
+        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
+    )
+
+    def update_split_visibility(split_model):
+        return gr.update(visible=split_model), gr.update(visible=split_model)
+
+    split_model.change(
+        fn=update_split_visibility,
+        inputs=split_model,
+        outputs=[split_max_tensors, split_max_size]
+    )
 
-if __name__ == "__main__":
-    scheduler = BackgroundScheduler(daemon=True)
-    scheduler.start()
+def restart_space():
+    HfApi().restart_space(repo_id="ggml-org/gguf-my-repo", token=HF_TOKEN, factory_reboot=True)
 
-    demo = app_interface()
-    demo.launch()
+scheduler = BackgroundScheduler()
+scheduler.add_job(restart_space, "interval", seconds=21600)
+scheduler.start()
 
-    signal.signal(signal.SIGINT, signal.SIG_DFL)
+# Launch the interface
+demo.queue(default_concurrency_limit=1, max_size=5).launch(debug=True, show_api=False)