Spaces:
Sleeping
Sleeping
File size: 3,963 Bytes
ad748cf 1a5c0c8 ad748cf 1a5c0c8 ad748cf 1a5c0c8 fce2a2c ad748cf 1a5c0c8 ad748cf 1a5c0c8 ad748cf 1a5c0c8 ad748cf 1a5c0c8 ad748cf 1a5c0c8 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 |
import gradio as gr
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
# --- Model Initialization ---
# Paths for tokenizer and your model checkpoint
tokenizer_path = "facebook/opt-1.3b"
model_path = "transfer_learning_therapist.pth"
# Load tokenizer and set pad token if needed
tokenizer = AutoTokenizer.from_pretrained(tokenizer_path)
if tokenizer.pad_token is None:
tokenizer.pad_token = tokenizer.eos_token
# Set device
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
print(f"Using device: {device}")
# Load the base model and then update with your checkpoint
model = AutoModelForCausalLM.from_pretrained(tokenizer_path)
checkpoint = torch.load(model_path, map_location=device)
model_dict = model.state_dict()
pretrained_dict = {k: v for k, v in checkpoint['model_state_dict'].items() if k in model_dict}
model_dict.update(pretrained_dict)
model.load_state_dict(model_dict)
model.to(device)
model.eval()
# --- Inference Function ---
def generate_response(prompt, max_new_tokens=150, temperature=0.7, top_p=0.9, repetition_penalty=1.2):
"""Generates a response from your model based on the prompt."""
model.eval()
model.config.use_cache = True
prompt = prompt.strip()
if not prompt:
return "Please provide a valid input."
# Tokenize the input prompt
inputs = tokenizer(prompt, return_tensors="pt").to(device)
try:
with torch.no_grad():
outputs = model.generate(
inputs.input_ids,
attention_mask=inputs.attention_mask,
max_new_tokens=max_new_tokens,
temperature=temperature,
top_p=top_p,
do_sample=True,
pad_token_id=tokenizer.pad_token_id,
eos_token_id=tokenizer.eos_token_id,
repetition_penalty=repetition_penalty,
num_beams=1, # greedy decoding
no_repeat_ngram_size=3, # avoid repeated phrases
)
except Exception as e:
return f"Error generating response: {e}"
finally:
model.config.use_cache = False
full_response = tokenizer.decode(outputs[0], skip_special_tokens=True)
# If your prompt is formatted with role markers (e.g., "Therapist:"), extract only that part:
if "Therapist:" in full_response:
therapist_response = full_response.split("Therapist:")[-1].strip()
else:
therapist_response = full_response.strip()
return therapist_response
# --- Gradio Interface Function ---
def respond(message, history, system_message, max_tokens, temperature, top_p):
"""
Build the conversation context by combining the system message and the dialogue history,
then generate a new response from the model.
"""
# Create a conversation prompt with your desired role labels.
conversation = f"System: {system_message}\n"
for user_msg, assistant_msg in history:
conversation += f"Human: {user_msg}\nTherapist: {assistant_msg}\n"
conversation += f"Human: {message}\nTherapist:"
response = generate_response(
conversation,
max_new_tokens=max_tokens,
temperature=temperature,
top_p=top_p,
)
history.append((message, response))
return history, history
# --- Gradio ChatInterface Setup ---
demo = gr.ChatInterface(
fn=respond,
title="MindfulAI Chat",
description="Chat with MindfulAI – an AI Therapist powered by your custom model.",
additional_inputs=[
gr.Textbox(value="You are a friendly AI Therapist.", label="System message"),
gr.Slider(minimum=1, maximum=2048, value=150, 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()
|