Update app.py

app.py

@@ -1,45 +1,68 @@
 import gradio as gr
-from huggingface_hub import InferenceClient
 from transformers import pipeline
 
-"""
-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
-"""
+# Initialize the text generation pipeline
+generator = pipeline("text-generation", model="gpt2", tokenizer="gpt2")
 
-# Initialize the inference client with the model you're using
-client = InferenceClient(model="isitcoding/gpt2_120_finetuned")
+# Define the response function with additional options for customization
+def text_generation(
+    prompt: str,
+    details: bool = False,
+    stream: bool = False,
+    model: str = None,
+    best_of: int = None,
+    decoder_input_details: bool = None,
+    do_sample: bool = False,
+    frequency_penalty: float = None,
+    grammar: None = None,
+    max_new_tokens: int = None,
+    repetition_penalty: float = None
+):
+    # Setup the configuration for the model generation
+    gen_params = {
+        "max_length": 50,  # Default, you can tweak it or set from parameters
+        "num_return_sequences": 1,
+        "do_sample": do_sample,
+        "temperature": 0.7,  # Controls randomness
+        "top_k": 50,  # You can adjust for more control over sampling
+        "top_p": 0.9,  # Same as above, for sampling
+    }
 
-# Initialize a text generation pipeline using Hugging Face's transformer
-generator = pipeline('text-generation', model=client)
+    if max_new_tokens:
+        gen_params["max_length"] = max_new_tokens + len(prompt.split())
 
-def respond(message, history: list[tuple[str, str]]):
-    """
-    Respond function to generate text based on the user's message and conversation history.
-    The `history` parameter keeps track of the conversation context.
-    """
-    # Add the new message to the conversation history
-    history.append(("User", message))
-    
-    # Use the generator model to get a response from the model
-    input_text = " ".join([h[1] for h in history])  # Combine the conversation history into one string
-    output = generator(input_text, max_length=500, num_return_sequences=1)
+    if frequency_penalty:
+        gen_params["frequency_penalty"] = frequency_penalty
 
-    # Extract the response from the output
-    response = output[0]['generated_text'].strip()
+    if repetition_penalty:
+        gen_params["repetition_penalty"] = repetition_penalty
 
-    # Add the model's response to the history
-    history.append(("Bot", response))
+    # Generate the text based on the input prompt and parameters
+    generated_text = generator(prompt, **gen_params)[0]["generated_text"]
 
-    return response, history
+    if details:
+        # Return additional details for debugging if needed
+        return {
+            "generated_text": generated_text,
+            "params_used": gen_params
+        }
+    else:
+        return generated_text
 
-# Create a Gradio interface for interaction
+# Create Gradio interface
 iface = gr.Interface(
-    fn=respond, 
-    inputs=[gr.Textbox(label="Enter your message", placeholder="Type here..."), gr.State()],
-    outputs=[gr.Textbox(label="Response"), gr.State()],
-    live=True
+    fn=text_generation,  # The function we defined
+    inputs=[
+        gr.Textbox(label="Input Prompt"),  # User input prompt
+        gr.Checkbox(label="Show Details", default=False),  # Option for additional details
+        gr.Checkbox(label="Stream Mode", default=False),  # Streaming checkbox (not used in this example)
+        gr.Textbox(label="Model (optional)", default=None),  # Optional model name
+        gr.Slider(minimum=1, maximum=5, label="Best of (Optional)", default=None),
+        gr.Slider(minimum=0.0, maximum=2.0, label="Frequency Penalty (Optional)", default=None),
+        gr.Slider(minimum=0.0, maximum=2.0, label="Repetition Penalty (Optional)", default=None),
+    ],
+    outputs="text"  # Output is plain text
 )
 
-# Launch the Gradio interface
+# Launch the interface
 iface.launch()