Update app.py

app.py

@@ -1,22 +1,91 @@
 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()
+
+# Fine-Tuning GPT-2 on Hugging Face Spaces (Streaming 40GB Dataset, No Storage Issues)
+
+# Install required libraries
+# Install required libraries (Run this separately in a terminal or notebook cell)
+# !pip install transformers datasets peft accelerate bitsandbytes torch torchvision torchaudio gradio -q
+
 from transformers import AutoModelForCausalLM, AutoTokenizer, Trainer, TrainingArguments
 from datasets import load_dataset
 from peft import LoraConfig, get_peft_model
 import torch
 
-# Initialize Hugging Face Inference Client
-client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
+# Authenticate Hugging Face
+from huggingface_hub import notebook_login
+notebook_login()
 
-# GPT-2 Model Setup
+# Load GPT-2 model and tokenizer
 model_name = "gpt2"
 tokenizer = AutoTokenizer.from_pretrained(model_name)
 model = AutoModelForCausalLM.from_pretrained(model_name)
 
-# Add padding token
-tokenizer.pad_token = tokenizer.eos_token
+# Load the OpenWebText dataset using streaming (No download required)
 
-# Custom Dataset
+# Custom Dataset (Predefined Q&A Pairs for Project Expo)
 custom_data = [
     {"prompt": "Who are you?", "response": "I am Eva, a virtual voice assistant."},
     {"prompt": "What is your name?", "response": "I am Eva, how can I help you?"},
@@ -30,8 +99,8 @@ custom_data = [
 # Convert custom dataset to Hugging Face Dataset
 dataset_custom = load_dataset("json", data_files={"train": custom_data})
 
-# Load OpenWebText dataset with `trust_remote_code=True`
-dataset = load_dataset("Skylion007/openwebtext", split="train[:20%]", trust_remote_code=True)
+# Merge with OpenWebText dataset
+dataset = load_dataset("Skylion007/openwebtext", split="train[:20%]")  # Load 5% to avoid streaming issues
 
 # Tokenization function
 def tokenize_function(examples):
@@ -39,16 +108,18 @@ def tokenize_function(examples):
 
 tokenized_datasets = dataset.map(tokenize_function, batched=True)
 
-# LoRA Configuration
+# Apply LoRA for efficient fine-tuning
 lora_config = LoraConfig(
     r=8, lora_alpha=32, lora_dropout=0.05, bias="none",
-    target_modules=["c_attn", "c_proj"]
+    target_modules=["c_attn", "c_proj"]  # Apply LoRA to attention layers
 )
 
 model = get_peft_model(model, lora_config)
+
+# Enable gradient checkpointing to reduce memory usage
 model.gradient_checkpointing_enable()
 
-# Training Arguments
+# Training arguments
 training_args = TrainingArguments(
     output_dir="gpt2_finetuned",
     auto_find_batch_size=True,
@@ -61,14 +132,14 @@ training_args = TrainingArguments(
     push_to_hub=True
 )
 
-# Trainer Setup
+# Trainer setup
 trainer = Trainer(
     model=model,
     args=training_args,
     train_dataset=tokenized_datasets
 )
 
-# Fine-tuning
+# Start fine-tuning
 trainer.train()
 
 # Save and push the model to Hugging Face Hub
@@ -76,44 +147,11 @@ trainer.save_model("gpt2_finetuned")
 tokenizer.save_pretrained("gpt2_finetuned")
 trainer.push_to_hub()
 
-# Chatbot Response Function
-def respond(message, history, system_message, max_tokens, temperature, top_p):
-    messages = [{"role": "system", "content": system_message}]
-
-    if isinstance(history, list):
-        for entry in history:
-            if isinstance(entry, dict):
-                messages.append(entry)
-            elif isinstance(entry, tuple) and len(entry) == 2:
-                messages.append({"role": "user", "content": entry[0]})
-                messages.append({"role": "assistant", "content": entry[1]})
+# Deploy as Gradio Interface
+def generate_response(prompt):
+    inputs = tokenizer(prompt, return_tensors="pt")
+    outputs = model.generate(**inputs, max_length=100)
+    return tokenizer.decode(outputs[0], skip_special_tokens=True)
 
-    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
-
-# Gradio Chatbot Interface
-demo = gr.ChatInterface(
-    respond,
-    chatbot=gr.Chatbot(type="messages"),
-    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)"),
-    ],
-)
-
-# Launch the App
-if __name__ == "__main__":
-    demo.launch()
+demo = gr.Interface(fn=generate_response, inputs="text", outputs="text")
+demo.launch()