Update app.py

app.py

@@ -59,6 +59,81 @@ demo = gr.ChatInterface(
     ],
 )
 
-
 if __name__ == "__main__":
     demo.launch()
+
+# Fine-Tuning GPT-2 on Hugging Face Spaces (Streaming 40GB Dataset, No Storage Issues)
+
+# Install required libraries
+!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
+
+# Authenticate Hugging Face
+from huggingface_hub import notebook_login
+notebook_login()
+
+# Load GPT-2 model and tokenizer
+model_name = "gpt2"
+tokenizer = AutoTokenizer.from_pretrained(model_name)
+model = AutoModelForCausalLM.from_pretrained(model_name)
+
+# Load the OpenWebText dataset using streaming (No download required)
+dataset = load_dataset("Skylion007/openwebtext", split="train[:5%]")  # Load 5% to avoid streaming issues
+
+# Tokenization function
+def tokenize_function(examples):
+    return tokenizer(examples["text"], truncation=True, padding="max_length", max_length=512)
+
+tokenized_datasets = dataset.map(tokenize_function, batched=True)
+
+# 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"]  # 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_args = TrainingArguments(
+    output_dir="gpt2_finetuned",
+    auto_find_batch_size=True,
+    gradient_accumulation_steps=4,
+    learning_rate=5e-5,
+    num_train_epochs=3,
+    save_strategy="epoch",
+    logging_dir="logs",
+    bf16=True,
+    push_to_hub=True
+)
+
+# Trainer setup
+trainer = Trainer(
+    model=model,
+    args=training_args,
+    train_dataset=tokenized_datasets
+)
+
+# Start fine-tuning
+trainer.train()
+
+# Save and push the model to Hugging Face Hub
+trainer.save_model("gpt2_finetuned")
+tokenizer.save_pretrained("gpt2_finetuned")
+trainer.push_to_hub()
+
+# 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)
+
+demo = gr.Interface(fn=generate_response, inputs="text", outputs="text")
+demo.launch()