update

.gradio/certificate.pem

@@ -0,0 +1,31 @@
+-----BEGIN CERTIFICATE-----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=
+-----END CERTIFICATE-----

app.py

@@ -7,7 +7,6 @@ from huggingface_hub import list_models
 from gradio_huggingfacehub_search import HuggingfaceHubSearch
 from packaging import version
 import os
-import spaces
 
 
 def hello(profile: gr.OAuthProfile | None, oauth_token: gr.OAuthToken | None) -> str:
@@ -62,10 +61,7 @@ model = AutoModel.from_pretrained("{model_name}")"""
     return model_card
 
 def load_model(model_name, quantization_config, auth_token) : 
-    return AutoModel.from_pretrained(model_name, quantization_config=quantization_config, device_map="cuda", use_auth_token=auth_token.token)
-
-def load_model_cpu(model_name, quantization_config, auth_token) : 
-    return AutoModel.from_pretrained(model_name, quantization_config=quantization_config, use_auth_token=auth_token.token)
+    return AutoModel.from_pretrained(model_name, quantization_config=quantization_config, device_map="cpu", use_auth_token=auth_token.token)
 
 def quantize_model(model_name, quantization_type, threshold, quant_type_4, double_quant_4, auth_token=None, username=None):
     print(f"Quantizing model: {quantization_type}")
@@ -89,14 +85,12 @@ def save_model(model, model_name, quantization_type, threshold, quant_type_4, do
     with tempfile.TemporaryDirectory() as tmpdirname:
 
 
-        model.save_pretrained(tmpdirname, safe_serialization=False, use_auth_token=auth_token.token)
+        model.save_pretrained(tmpdirname, safe_serialization=True, use_auth_token=auth_token.token)
         if quantized_model_name : 
             repo_name = f"{username}/{quantized_model_name}"
-        else : 
-            if quantization_type == "int4_weight_only" : 
-                repo_name = f"{username}/{model_name.split('/')[-1]}-BNB-{quantization_type}"
-            else : 
-                repo_name = f"{username}/{model_name.split('/')[-1]}-BNB-{quantization_type}"
+        else :   
+            repo_name = f"{username}/{model_name.split('/')[-1]}-BNB-{quantization_type}"
+            
 
         model_card = create_model_card(repo_name, quantization_type, threshold, quant_type_4, double_quant_4)
         with open(os.path.join(tmpdirname, "README.md"), "w") as f:
@@ -132,17 +126,18 @@ def quantize_and_save(profile: gr.OAuthProfile | None, oauth_token: gr.OAuthToke
     
     threshold = float(threshold)
 
-    try:
-        quantized_model = quantize_model(model_name, quantization_type, threshold, quant_type_4, double_quant_4, oauth_token, profile.username)
-        return save_model(quantized_model, model_name, quantization_type, threshold, quant_type_4, double_quant_4, profile.username, oauth_token, quantized_model_name)
-    except Exception as e : 
-        return f"An error occurred: {str(e)}"
+    # try:
+    quantized_model = quantize_model(model_name, quantization_type, threshold, quant_type_4, double_quant_4, oauth_token, profile.username)
+    return save_model(quantized_model, model_name, quantization_type, threshold, quant_type_4, double_quant_4, profile.username, oauth_token, quantized_model_name)
+    # except Exception as e : 
+    #     print(e)
+    #     return f"An error occurred: {str(e)}"
 
 
 css="""/* Custom CSS to allow scrolling */
 .gradio-container {overflow-y: auto;}
 """
-with gr.Blocks(theme=gr.themes.Ocean(), css=css) as app:
+with gr.Blocks(theme=gr.themes.Ocean(), css=css) as demo:
     gr.Markdown(
         """
         # 🤗 LLM Model BitsAndBytes Quantization App
@@ -154,10 +149,10 @@ with gr.Blocks(theme=gr.themes.Ocean(), css=css) as app:
     gr.LoginButton(elem_id="login-button", elem_classes="center-button", min_width=250)
 
     m1 = gr.Markdown()
-    app.load(hello, inputs=None, outputs=m1)
+    demo.load(hello, inputs=None, outputs=m1)
 
 
-    radio = gr.Radio(["show", "hide"], label="Show Instructions")
+    # radio = gr.Radio(["show", "hide"], label="Show Instructions")
     instructions = gr.Markdown(
         """
         ## Instructions
@@ -174,13 +169,24 @@ with gr.Blocks(theme=gr.themes.Ocean(), css=css) as app:
         """,
         visible=False
     )
-    def update_visibility(radio):  # Accept the event argument, even if not used
-        value = radio  # Get the selected value from the radio button
-        if value == "show":
-            return gr.Textbox(visible=True) #make it visible
-        else:
-            return gr.Textbox(visible=False)
-    radio.change(update_visibility, radio, instructions)
+    
+    instructions_visible = gr.State(False)
+    toggle_button = gr.Button("▼ Show Instructions", elem_id="toggle-button", elem_classes="toggle-button")
+    
+    def toggle_instructions(instructions_visible):
+        new_visibility = not instructions_visible  # Toggle the state
+        new_label = "▲ Hide Instructions" if new_visibility else "▼ Show Instructions"  # Change label based on visibility
+        return gr.update(visible=new_visibility), new_visibility, gr.update(value=new_label)  # Toggle visibility and return new state
+    
+    toggle_button.click(toggle_instructions, instructions_visible, [instructions, instructions_visible, toggle_button])
+    
+    # def update_visibility(radio):  # Accept the event argument, even if not used
+    #     value = radio  # Get the selected value from the radio button
+    #     if value == "show":
+    #         return gr.Textbox(visible=True) #make it visible
+    #     else:
+    #         return gr.Textbox(visible=False)
+    # radio.change(update_visibility, radio, instructions)
 
     with gr.Row():
         with gr.Column():
@@ -228,18 +234,18 @@ with gr.Blocks(theme=gr.themes.Ocean(), css=css) as app:
                     )
         with gr.Column():
             quantize_button = gr.Button("Quantize and Save Model", variant="primary")
-            output_link = gr.Markdown(label="Quantized Model Link", container=True, min_height=40)
+            output_link = gr.Markdown(label="Quantized Model Link", container=True, min_height=80)
     
     
     # Adding CSS styles for the username box
-    app.css = """
+    demo.css = """
     #username-box {
         background-color: #f0f8ff; /* Light color */
         border-radius: 8px;
         padding: 10px;
     }
     """
-    app.css = """
+    demo.css = """
     .center-button {
         display: flex;
         justify-content: center;
@@ -254,6 +260,7 @@ with gr.Blocks(theme=gr.themes.Ocean(), css=css) as app:
         outputs=[output_link]
     )
 
-
+if __name__ == "__main__":
+    demo.launch(share=True)
 # Launch the app
-app.launch()
+# demo.launch(share=True, debug=True)

requirements.txt

@@ -1,6 +1,5 @@
-git+https://github.com/huggingface/transformers.git@main#egg=transformers
+transformers
 accelerate
-torchao
 huggingface-hub
 gradio-huggingfacehub-search
-bitsandbytes
+https://github.com/bitsandbytes-foundation/bitsandbytes/releases/download/continuous-release_multi-backend-refactor/bitsandbytes-0.44.1.dev0-py3-none-manylinux_2_24_x86_64.whl