Spaces:
Sleeping
Sleeping
File size: 2,878 Bytes
f72ef0e 612c19e f72ef0e 612c19e f72ef0e 612c19e f72ef0e 612c19e |
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 |
import gradio as gr
from transformers import Trainer, TrainingArguments, AutoTokenizer, AutoModelForSeq2SeqLM
from datasets import load_dataset
def fine_tune_model(model_name, dataset_name, hub_id, num_epochs, batch_size, lr, grad):
# Load the dataset
dataset = load_dataset(dataset_name)
# Load the model and tokenizer
model = AutoModelForSeq2SeqLM.from_pretrained(model_name, num_labels=2)
tokenizer = AutoTokenizer.from_pretrained(model_name)
# Tokenize the dataset
def tokenize_function(examples):
return tokenizer(examples['text'], padding="max_length", truncation=True)
tokenized_datasets = dataset.map(tokenize_function, batched=True)
# Set training arguments
training_args = TrainingArguments(
output_dir='./results',
evaluation_strategy="epoch",
learning_rate=lr,
per_device_train_batch_size=batch_size,
per_device_eval_batch_size=batch_size,
num_train_epochs=num_epochs,
weight_decay=0.01,
evaluation_strategy='epoch',
gradient_accumulation_steps=grad,
load_best_model_at_end=True,
metric_for_best_model="accuracy",
greater_is_better=True,
logging_dir='./logs',
logging_steps=10,
push_to_hub=True,
hub_model_id=hub_id,
)
# Create Trainer
trainer = Trainer(
model=model,
args=training_args,
train_dataset=tokenized_datasets['train'],
eval_dataset=tokenized_datasets['validation'],
)
# Fine-tune the model
trainer.train()
trainer.push_to_hub(commit_message="Training complete!")
return 'DONE!'#model
# Define Gradio interface
def predict(text):
inputs = tokenizer(text, return_tensors="pt", padding=True, truncation=True)
outputs = model(inputs)
predictions = outputs.logits.argmax(dim=-1)
return "Positive" if predictions.item() == 1 else "Negative"
# Create Gradio interface
iface = gr.Interface(
fn=fine_tune_model,
inputs=[
gr.inputs.Textbox(label="Model Name (e.g., 'google/t5-efficient-tiny-nh8')"),
gr.inputs.Textbox(label="Dataset Name (e.g., 'imdb')"),
gr.inputs.Textbox(label="HF hub to push to after training"),
gr.inputs.Slider(minimum=1, maximum=10, default=3, label="Number of Epochs"),
gr.inputs.Slider(minimum=1, maximum=16, default=4, label="Batch Size"),
gr.inputs.Slider(minimum=1, maximum=16, default=4, label="Batch Size"),
gr.inputs.Slider(minimum=1, maximum=1000, default=50, label="Learning Rate (e-6)"),
gr.inputs.Slider(minimum=1, maximum=100, default=1, label="Gradient accumulation (e-1)"),
],
outputs="text",
title="Fine-Tune Hugging Face Model",
description="This interface allows you to fine-tune a Hugging Face model on a specified dataset."
)
# Launch the interface
iface.launch() |