| import gradio as gr | |
| from transformers import AutoTokenizer, AutoModelForCausalLM | |
| import torch | |
| # Set model ID | |
| # comment/uncomment the model you want to use | |
| # GPT-2 (very small, general-purpose, mainly for testing or learning purposes) | |
| # model_id = "gpt2" | |
| # DeepSeek Coder 1.3B (base version, no instruction fine-tuning — better for raw code generation tasks) | |
| # model_id = "deepseek-ai/deepseek-coder-1.3b" | |
| # DeepSeek Coder 1.3B Base (same as above — explicit base naming, safe to use) | |
| # model_id = "deepseek-ai/deepseek-coder-1.3b-base" | |
| # CodeLlama 7B Instruct (powerful code generation model from Meta, instruction-tuned) | |
| # model_id = "codellama/CodeLlama-7b-Instruct-hf" | |
| # Meta-Llama 3.1 8B Instruct (very powerful general-purpose model, instruction-following, also decent for code & NLP) | |
| # model_id = "meta-llama/Llama-3.1-8B-Instruct" | |
| # DeepSeek-R1 + Qwen3 8B (highly capable multi-purpose model — great for reasoning, coding, general Q&A) | |
| # model_id = "deepseek-ai/DeepSeek-R1-0528-Qwen3-8B" | |
| # Qwen2.5-VL-7B Instruct (multimodal: can handle text + images, instruction-tuned — mostly for vision-language tasks) | |
| # model_id = "Qwen/Qwen2.5-VL-7B-Instruct" | |
| # DeepSeek Coder 1.3B Instruct (great for both natural language and coding tasks) | |
| model_id = "deepseek-ai/deepseek-coder-1.3b-instruct" | |
| # Load tokenizer and model | |
| tokenizer = AutoTokenizer.from_pretrained(model_id) | |
| model = AutoModelForCausalLM.from_pretrained( | |
| model_id, | |
| torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32 | |
| ) | |
| # Move model to GPU if available | |
| device = torch.device("cuda" if torch.cuda.is_available() else "cpu") | |
| model.to(device) | |
| def generate_code(prompt): | |
| if not prompt.strip(): | |
| return "⚠ Please enter a valid prompt." | |
| inputs = tokenizer(prompt, return_tensors="pt").to(device) | |
| with torch.no_grad(): | |
| outputs = model.generate( | |
| **inputs, | |
| max_new_tokens=200, | |
| temperature=0.7 | |
| ) | |
| output_text = tokenizer.decode(outputs[0], skip_special_tokens=True) | |
| # Strip the prompt if it appears at the start | |
| if output_text.startswith(prompt): | |
| output_text = output_text[len(prompt):].lstrip() | |
| return output_text | |
| demo = gr.Interface( | |
| fn=generate_code, | |
| inputs=gr.Textbox(lines=5, label="Ask me a question ? or tell me to generate-code ^_^ :"), | |
| outputs=gr.Textbox(label="Generated Output is:"), | |
| title="Code-NLP Fusion" | |
| ) | |
| demo.launch() | |