Spaces:
Sleeping
Sleeping
Upload 2 files
Browse files- app.py +37 -26
- requirements.txt +2 -6
app.py
CHANGED
@@ -2,40 +2,51 @@ import gradio as gr
|
|
2 |
import torch
|
3 |
from transformers import GPT2LMHeadModel, GPT2Tokenizer
|
4 |
|
5 |
-
# Load model and tokenizer
|
6 |
def load_model():
|
7 |
-
|
8 |
-
|
9 |
-
|
10 |
-
|
11 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
12 |
|
13 |
# Text generation function
|
14 |
def generate_text(prompt, max_length=500, temperature=0.8, top_k=40, top_p=0.9):
|
15 |
-
|
16 |
-
|
17 |
-
|
18 |
-
|
19 |
-
|
20 |
-
|
21 |
-
|
22 |
-
|
23 |
-
|
24 |
-
|
25 |
-
|
26 |
-
|
27 |
-
|
28 |
-
|
29 |
-
|
30 |
-
|
31 |
-
|
32 |
-
|
33 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
34 |
|
35 |
# Load model and tokenizer globally
|
36 |
print("Loading model and tokenizer...")
|
37 |
model, tokenizer = load_model()
|
38 |
-
print("Model loaded successfully!")
|
39 |
|
40 |
# Create Gradio interface
|
41 |
demo = gr.Interface(
|
|
|
2 |
import torch
|
3 |
from transformers import GPT2LMHeadModel, GPT2Tokenizer
|
4 |
|
5 |
+
# Load model and tokenizer
|
6 |
def load_model():
|
7 |
+
try:
|
8 |
+
# Load the fine-tuned model
|
9 |
+
model = GPT2LMHeadModel.from_pretrained("aayushraina/gpt2shakespeare")
|
10 |
+
# Use the base GPT-2 tokenizer
|
11 |
+
tokenizer = GPT2Tokenizer.from_pretrained("gpt2")
|
12 |
+
model.eval()
|
13 |
+
print("Model and tokenizer loaded successfully!")
|
14 |
+
return model, tokenizer
|
15 |
+
except Exception as e:
|
16 |
+
print(f"Error loading model: {e}")
|
17 |
+
return None, None
|
18 |
|
19 |
# Text generation function
|
20 |
def generate_text(prompt, max_length=500, temperature=0.8, top_k=40, top_p=0.9):
|
21 |
+
if model is None or tokenizer is None:
|
22 |
+
return "Error: Model not loaded properly"
|
23 |
+
|
24 |
+
try:
|
25 |
+
# Encode the input prompt
|
26 |
+
input_ids = tokenizer.encode(prompt, return_tensors='pt')
|
27 |
+
|
28 |
+
# Generate text
|
29 |
+
with torch.no_grad():
|
30 |
+
output = model.generate(
|
31 |
+
input_ids,
|
32 |
+
max_length=max_length,
|
33 |
+
temperature=temperature,
|
34 |
+
top_k=top_k,
|
35 |
+
top_p=top_p,
|
36 |
+
do_sample=True,
|
37 |
+
pad_token_id=tokenizer.eos_token_id,
|
38 |
+
num_return_sequences=1
|
39 |
+
)
|
40 |
+
|
41 |
+
# Decode and return the generated text
|
42 |
+
generated_text = tokenizer.decode(output[0], skip_special_tokens=True)
|
43 |
+
return generated_text
|
44 |
+
except Exception as e:
|
45 |
+
return f"Error during generation: {str(e)}"
|
46 |
|
47 |
# Load model and tokenizer globally
|
48 |
print("Loading model and tokenizer...")
|
49 |
model, tokenizer = load_model()
|
|
|
50 |
|
51 |
# Create Gradio interface
|
52 |
demo = gr.Interface(
|
requirements.txt
CHANGED
@@ -1,7 +1,3 @@
|
|
1 |
-
wandb
|
2 |
-
tiktoken
|
3 |
torch>=2.0.0
|
4 |
-
|
5 |
-
|
6 |
-
transformers
|
7 |
-
gradio>=4.0.0
|
|
|
|
|
|
|
1 |
torch>=2.0.0
|
2 |
+
transformers>=4.30.0
|
3 |
+
gradio>=4.0.0
|
|
|
|