Update space

README.md

@@ -10,4 +10,19 @@ pinned: false
 short_description: chat with gpt2-nepali
 ---
 
-An example chatbot using [Gradio](https://gradio.app), [`huggingface_hub`](https://huggingface.co/docs/huggingface_hub/v0.22.2/en/index), and the [Hugging Face Inference API](https://huggingface.co/docs/api-inference/index).
+An example chatbot using [Gradio](https://gradio.app), [`huggingface_hub`](https://huggingface.co/docs/huggingface_hub/v0.22.2/en/index), and the [Hugging Face Inference API](https://huggingface.co/docs/api-inference/index).
+
+# Gradio instructions
+
+```
+# When prompted for a password, use an access token with write permissions.
+
+# Generate one from your settings: https://huggingface.co/settings/tokens
+
+git clone https://huggingface.co/spaces/Aananda-giri/gpt2-nepalis
+
+# modify these files locally, then commit and push
+
+git commit -am "Update space"
+git push
+```

app.py

@@ -1,64 +1,82 @@
+# app.py (second app by claude)
 import gradio as gr
-from huggingface_hub import InferenceClient
+import torch
+from model import GPTModel
+from transformers import PreTrainedTokenizerFast
+from gpt_model_code import load_model_n_tokenizer, generate
+# Load model and tokenizer once at startup
+model, tokenizer = load_model_n_tokenizer()
+model.eval()
 
-"""
-For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
-"""
-client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
+def generate_text(prompt, max_new_tokens, top_k, top_p, temperature, repetition_penalty, penalize_len_below):
+    device = next(model.parameters()).device
+    
+    # Convert top_k to None if using top_p
+    if top_p > 0:
+        top_k = None
+    else:
+        top_p = None
+    
+    with torch.no_grad():
+        output_text = generate(  # function uses `with torch.no_grad()` internally already
+            model=model,
+            prompt=prompt,
+            tokenizer=tokenizer,
+            max_new_tokens=max_new_tokens,
+            top_p=top_p,# top p sampling is prefered over top k if top_p != None
+            top_k=top_k,
+            temperature=0.7,
+            repetition_penalty=repetition_penalty,  # New parameter: Repetition penalty factor
+            penalize_len_below=penalize_len_below  # New parameter: Minimum content length for penalizing EOT token.
+        )
+        
+        return output_text
 
+# Create Gradio interface
+with gr.Blocks(title="Nepali GPT-2 Text Generator") as interface:
+    gr.Markdown("# Nepali GPT-2 Text Generator")
+    gr.Markdown("Enter Nepali text to generate content using the custom GPT-2 model.")
+    
+    with gr.Row():
+        with gr.Column():
+            prompt = gr.Textbox(label="Prompt", placeholder="Enter Nepali text here...")
+            max_tokens = gr.Slider(minimum=1, maximum=512, value=50, step=1, label="Max New Tokens")
+            
+            with gr.Row():
+                with gr.Column():
+                    top_k = gr.Slider(minimum=0, maximum=100, value=50, step=1, label="Top K (set to 0 to use Top P)")
+                    temperature = gr.Slider(minimum=0.1, maximum=2.0, value=0.7, step=0.1, label="Temperature")
+                with gr.Column():
+                    top_p = gr.Slider(minimum=0, maximum=1.0, value=0, step=0.05, label="Top P (set above 0 to use instead of Top K)")
+                    repetition_penalty = gr.Slider(minimum=1.0, maximum=2.0, value=1.2, step=0.1, label="Repetition Penalty")
+            
+            min_length = gr.Slider(minimum=1, maximum=200, value=50, step=1, label="Minimum Length Penalty")
+            generate_btn = gr.Button("Generate Text")
+        
+        with gr.Column():
+            output = gr.Textbox(label="Generated Text", lines=10)
+    
+    # Add examples if you have any
+    gr.Examples(
+        examples=[
+            ["रामले भात खायो", 50, 50, 0, 0.7, 1.2, 50],
+            ["नेपाल एउटा", 100, 0, 0.9, 0.8, 1.2, 100],
+        ],
+        inputs=[prompt, max_tokens, top_k, top_p, temperature, repetition_penalty, min_length],
+        outputs=output,
+        fn=generate_text,
+        cache_examples=True,
+    )
+    
+    generate_btn.click(
+        fn=generate_text,
+        inputs=[prompt, max_tokens, top_p, top_k, temperature, repetition_penalty, min_length],
+        outputs=output
+    )
 
-def respond(
-    message,
-    history: list[tuple[str, str]],
-    system_message,
-    max_tokens,
-    temperature,
-    top_p,
-):
-    messages = [{"role": "system", "content": system_message}]
 
-    for val in history:
-        if val[0]:
-            messages.append({"role": "user", "content": val[0]})
-        if val[1]:
-            messages.append({"role": "assistant", "content": val[1]})
+'''
 
-    messages.append({"role": "user", "content": message})
+'''
 
-    response = ""
-
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        token = message.choices[0].delta.content
-
-        response += token
-        yield response
-
-
-"""
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
-"""
-demo = gr.ChatInterface(
-    respond,
-    additional_inputs=[
-        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
-        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
-        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
-        gr.Slider(
-            minimum=0.1,
-            maximum=1.0,
-            value=0.95,
-            step=0.05,
-            label="Top-p (nucleus sampling)",
-        ),
-    ],
-)
-
-
-if __name__ == "__main__":
-    demo.launch()
+interface.launch()

app_autogenerated_code.py

@@ -0,0 +1,64 @@
+import gradio as gr
+from huggingface_hub import InferenceClient
+
+"""
+For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
+"""
+client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
+
+
+def respond(
+    message,
+    history: list[tuple[str, str]],
+    system_message,
+    max_tokens,
+    temperature,
+    top_p,
+):
+    messages = [{"role": "system", "content": system_message}]
+
+    for val in history:
+        if val[0]:
+            messages.append({"role": "user", "content": val[0]})
+        if val[1]:
+            messages.append({"role": "assistant", "content": val[1]})
+
+    messages.append({"role": "user", "content": message})
+
+    response = ""
+
+    for message in client.chat_completion(
+        messages,
+        max_tokens=max_tokens,
+        stream=True,
+        temperature=temperature,
+        top_p=top_p,
+    ):
+        token = message.choices[0].delta.content
+
+        response += token
+        yield response
+
+
+"""
+For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
+"""
+demo = gr.ChatInterface(
+    respond,
+    additional_inputs=[
+        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
+        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
+        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
+        gr.Slider(
+            minimum=0.1,
+            maximum=1.0,
+            value=0.95,
+            step=0.05,
+            label="Top-p (nucleus sampling)",
+        ),
+    ],
+)
+
+
+if __name__ == "__main__":
+    demo.launch()

app_by_claude.py

@@ -0,0 +1,54 @@
+# app.py
+import gradio as gr
+import torch
+from model import GPTModel  # Import your specific GPT model class
+from transformers import PreTrainedTokenizerFast
+
+# Load model and tokenizer once at startup
+def load_model_n_tokenizer():
+    model = GPTModel.from_pretrained("Aananda-giri/GPT2-Nepali")
+    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+    model.to(device)
+    tokenizer = PreTrainedTokenizerFast.from_pretrained("Aananda-giri/NepaliBPE")
+    return model, tokenizer
+
+# Initialize at startup
+model, tokenizer = load_model_n_tokenizer()
+model.eval()
+
+def generate(prompt, max_new_tokens, top_k, temperature, repetition_penalty, penalize_len_below):
+    device = next(model.parameters()).device
+    
+    with torch.no_grad():
+        input_ids = tokenizer(prompt, return_tensors="pt").input_ids.to(device)
+        
+        outputs = model.generate(
+            input_ids,
+            max_new_tokens=max_new_tokens,
+            top_k=top_k,
+            temperature=temperature,
+            repetition_penalty=repetition_penalty,
+            min_length=penalize_len_below,
+            pad_token_id=tokenizer.pad_token_id,
+            eos_token_id=tokenizer.eos_token_id,
+        )
+        
+        return tokenizer.decode(outputs[0], skip_special_tokens=True)
+
+# Create Gradio interface
+interface = gr.Interface(
+    fn=generate,
+    inputs=[
+        gr.Textbox(label="Prompt", placeholder="Enter Nepali text here..."),
+        gr.Slider(minimum=1, maximum=512, value=50, step=1, label="Max New Tokens"),
+        gr.Slider(minimum=1, maximum=100, value=3, step=1, label="Top K"),
+        gr.Slider(minimum=0.1, maximum=2.0, value=0.7, step=0.1, label="Temperature"),
+        gr.Slider(minimum=1.0, maximum=2.0, value=1.2, step=0.1, label="Repetition Penalty"),
+        gr.Slider(minimum=1, maximum=200, value=50, step=1, label="Minimum Length Penalty"),
+    ],
+    outputs=gr.Textbox(label="Generated Text"),
+    title="Nepali GPT-2 Text Generator",
+    description="Enter Nepali text to generate content using the custom GPT-2 model."
+)
+
+interface.launch()