Upload 3 files
Browse files- app.py +63 -0
- gpt_model_quantized.pt +3 -0
- requirements.txt +1 -0
app.py
ADDED
|
@@ -0,0 +1,63 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import gradio as gr
|
| 2 |
+
import torch
|
| 3 |
+
import tiktoken
|
| 4 |
+
import numpy as np
|
| 5 |
+
from train import GPT, GPTConfig # Make sure to upload train.py to the Space
|
| 6 |
+
|
| 7 |
+
def load_quantized_model():
|
| 8 |
+
model = GPT(GPTConfig())
|
| 9 |
+
quantized_dict = torch.load("gpt_model_quantized.pt")
|
| 10 |
+
|
| 11 |
+
# Dequantize model
|
| 12 |
+
state_dict = {}
|
| 13 |
+
for key, value in quantized_dict.items():
|
| 14 |
+
if isinstance(value, dict):
|
| 15 |
+
state_dict[key] = torch.tensor(
|
| 16 |
+
value['data'].astype(np.float32) * value['scale']
|
| 17 |
+
)
|
| 18 |
+
else:
|
| 19 |
+
state_dict[key] = value
|
| 20 |
+
|
| 21 |
+
model.load_state_dict(state_dict)
|
| 22 |
+
model.eval()
|
| 23 |
+
return model
|
| 24 |
+
|
| 25 |
+
def generate_text(input_text):
|
| 26 |
+
try:
|
| 27 |
+
# Set device
|
| 28 |
+
device = 'cuda' if torch.cuda.is_available() else 'cpu'
|
| 29 |
+
|
| 30 |
+
# Load model
|
| 31 |
+
model = load_quantized_model()
|
| 32 |
+
model = model.to(device)
|
| 33 |
+
|
| 34 |
+
# Tokenize input
|
| 35 |
+
tokenizer = tiktoken.get_encoding('gpt2')
|
| 36 |
+
input_tokens = torch.tensor([tokenizer.encode(input_text)]).to(device)
|
| 37 |
+
|
| 38 |
+
# Generate
|
| 39 |
+
with torch.no_grad():
|
| 40 |
+
output_tokens = model.generate(input_tokens, max_new_tokens=500)[0].tolist()
|
| 41 |
+
|
| 42 |
+
# Decode and return
|
| 43 |
+
generated_text = tokenizer.decode(output_tokens)
|
| 44 |
+
return generated_text
|
| 45 |
+
except Exception as e:
|
| 46 |
+
return f"Error generating text: {e}"
|
| 47 |
+
|
| 48 |
+
# Create Gradio interface
|
| 49 |
+
iface = gr.Interface(
|
| 50 |
+
fn=generate_text,
|
| 51 |
+
inputs=gr.Textbox(lines=5, label="Input Text"),
|
| 52 |
+
outputs=gr.Textbox(lines=10, label="Generated Text"),
|
| 53 |
+
title="GPT Text Generator",
|
| 54 |
+
description="Enter some text and the model will generate a continuation.",
|
| 55 |
+
examples=[
|
| 56 |
+
["The quick brown fox"],
|
| 57 |
+
["In a world where AI"],
|
| 58 |
+
["Once upon a time"]
|
| 59 |
+
]
|
| 60 |
+
)
|
| 61 |
+
|
| 62 |
+
# Launch the interface
|
| 63 |
+
iface.launch()
|
gpt_model_quantized.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:84ddd50901ce047532e6587705beb5768740afc9879c4652094b842dd913c41c
|
| 3 |
+
size 255934032
|
requirements.txt
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
|