Spaces:
Runtime error
Runtime error
import gradio as gr | |
import time | |
from transformers import pipeline, AutoTokenizer, AutoModelForSequenceClassification | |
model_name = "ethanrom/a2" | |
tokenizer = AutoTokenizer.from_pretrained(model_name) | |
model = AutoModelForSequenceClassification.from_pretrained(model_name) | |
pretrained_model_name = "roberta-large-mnli" | |
pretrained_tokenizer = AutoTokenizer.from_pretrained(pretrained_model_name) | |
pretrained_model = pipeline("zero-shot-classification", model=pretrained_model_name, tokenizer=pretrained_tokenizer) | |
candidate_labels = ["negative", "positive", "no impact", "mixed"] | |
accuracy_scores = {"Fine-tuned": 0.0, "Pretrained": 0.0} | |
model_sizes = {"Fine-tuned": 0, "Pretrained": 0} | |
inference_times = {"Fine-tuned": 0.0, "Pretrained": 0.0} | |
def predict_sentiment(text_input, model_selection): | |
global accuracy_scores, model_sizes, inference_times | |
start_time = time.time() | |
if model_selection == "Fine-tuned": | |
inputs = tokenizer.encode_plus(text_input, return_tensors='pt') | |
outputs = model(**inputs) | |
logits = outputs.logits.detach().cpu().numpy()[0] | |
predicted_class = int(logits.argmax()) | |
accuracy_scores[model_selection] += 1 if candidate_labels[predicted_class] == "positive" else 0 | |
model_sizes[model_selection] = model.num_parameters() | |
else: | |
result = pretrained_model(text_input, candidate_labels) | |
predicted_class = result["labels"][0] | |
accuracy_scores[model_selection] += 1 if predicted_class == 1 else 0 | |
model_sizes[model_selection] = pretrained_model.model.num_parameters() | |
end_time = time.time() | |
inference_times[model_selection] = end_time - start_time | |
return candidate_labels[predicted_class] | |
def accuracy(model_selection): | |
return accuracy_scores[model_selection]/10 | |
def model_size(model_selection): | |
return str(model_sizes[model_selection]//(1024*1024)) + " MB" | |
def inference_time(model_selection): | |
return str(inference_times[model_selection]*1000) + " ms" | |
inputs = [ | |
gr.inputs.Textbox("Enter text"), | |
gr.inputs.Dropdown(["Pretrained", "Fine-tuned"], label="Select model"), | |
] | |
outputs = [ | |
gr.outputs.Textbox(label="Predicted Sentiment"), | |
gr.outputs.Label(label="Accuracy:"), | |
gr.outputs.Label(label="Model Size:"), | |
gr.outputs.Label(label="Inference Time:") | |
] | |
gr.Interface(fn=predict_sentiment, inputs=inputs, outputs=outputs, | |
title="Sentiment Analysis", description="Compare the output of two models", | |
live=True, | |
examples=[["on us lift up the light", "Fine-tuned"], ["max laid his hand upon the old man's arm", "Pretrained"]] | |
).launch(); | |