| import gradio as gr | |
| from transformers import AutoModelForCausalLM, AutoTokenizer | |
| import torch | |
| model_name = "AddieFoote0/language-100M-MaxEnt-distilled-relearned" | |
| model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype=torch.bfloat16) | |
| tokenizer = AutoTokenizer.from_pretrained(model_name) | |
| if hasattr(torch, "compile"): | |
| model = torch.compile(model) | |
| print("compiled model") | |
| else: | |
| print("no compile") | |
| def generate_response(prompt): | |
| inputs = tokenizer(prompt, return_tensors="pt") | |
| outputs = model.generate(**inputs, | |
| max_new_tokens=16, | |
| do_sample=True, | |
| temperature=1, | |
| pad_token_id=tokenizer.eos_token_id, | |
| ) | |
| input_length = inputs['input_ids'].shape[1] | |
| new_token_ids = outputs[0][input_length:] | |
| new_tokens = tokenizer.decode(new_token_ids, skip_special_tokens=False) | |
| return new_tokens | |
| iface = gr.Interface( | |
| fn=generate_response, | |
| inputs=gr.Textbox(label="Enter your prompt"), | |
| outputs=gr.Textbox(label="Model Response"), | |
| title="Lang Model Demo", | |
| ) | |
| iface.launch() |