Spaces:
Sleeping
Sleeping
File size: 1,914 Bytes
ac83605 |
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 |
import gradio as gr
from transformers import AutoModel
from safetensors.torch import load_file, save_file
import torch
import os
# Function to download and modify model
def add_fuzz_to_model(model_name, min_val, max_val):
# Download model weights from Hugging Face
model = AutoModel.from_pretrained(model_name, use_safetensors=True)
model_file = f"{model_name}.safetensors"
model.save_pretrained(".", safe_serialization=True) # Save locally for access
# Load safetensors
weights = load_file(model_file)
# Apply fuzz to each weight tensor
def add_fuzz(tensor, min_val, max_val):
noise = torch.empty_like(tensor).uniform_(min_val, max_val)
return tensor + noise
modified_weights = {}
for key, tensor in weights.items():
modified_weights[key] = add_fuzz(tensor, min_val, max_val)
# Save modified weights
modified_file = f"{model_name}_fuzzed.safetensors"
save_file(modified_weights, modified_file)
return f"Model processed and saved as {modified_file}"
# Gradio UI
def fuzz_weights_ui():
with gr.Blocks() as interface:
gr.Markdown("### Add Fuzz to Hugging Face Model Weights")
with gr.Row():
model_name = gr.Textbox(label="Hugging Face Model Name", placeholder="e.g., gpt2")
with gr.Row():
min_val = gr.Slider(-0.5, 0.0, value=-0.1, step=0.01, label="Minimum Fuzz")
max_val = gr.Slider(0.0, 0.5, value=0.1, step=0.01, label="Maximum Fuzz")
with gr.Row():
process_button = gr.Button("Apply Fuzz")
result = gr.Textbox(label="Result")
# Define interaction
process_button.click(
fn=add_fuzz_to_model,
inputs=[model_name, min_val, max_val],
outputs=[result],
)
return interface
# Launch the app
interface = fuzz_weights_ui()
interface.launch()
|