Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -1,43 +1,41 @@
|
|
1 |
from peft import PeftModel
|
2 |
from transformers import AutoModelForCausalLM, AutoTokenizer
|
3 |
import gradio as gr
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
4 |
|
5 |
# Load the base model, fine-tuned model, and tokenizer
|
6 |
try:
|
7 |
-
# Base model and fine-tuned model names
|
8 |
-
base_model_name = "meta-llama/Meta-Llama-3-8B"
|
9 |
-
fine_tuned_model_name = "VinitT/Sanskrit-llama"
|
10 |
-
|
11 |
print("Loading base model...")
|
12 |
-
base_model = AutoModelForCausalLM.from_pretrained(base_model_name)
|
13 |
-
|
14 |
print("Loading fine-tuned model...")
|
15 |
-
model = PeftModel.from_pretrained(base_model, fine_tuned_model_name)
|
16 |
-
|
17 |
print("Loading tokenizer...")
|
18 |
-
|
19 |
-
tokenizer = AutoTokenizer.from_pretrained(base_model_name)
|
20 |
-
except Exception as e:
|
21 |
-
print(f"Failed to load tokenizer for {base_model_name}: {e}")
|
22 |
-
print("Falling back to a generic tokenizer...")
|
23 |
-
tokenizer = AutoTokenizer.from_pretrained("EleutherAI/gpt-neo-1.3B")
|
24 |
|
25 |
except Exception as e:
|
26 |
-
|
27 |
-
tokenizer = None # Ensure the tokenizer is defined even if loading fails
|
28 |
|
29 |
# Function to generate text using the model
|
30 |
def generate_text(input_text):
|
31 |
-
if tokenizer is None:
|
32 |
-
return "Tokenizer failed to load. Please check your model configuration."
|
33 |
try:
|
34 |
# Tokenize the input
|
35 |
inputs = tokenizer(input_text, return_tensors="pt")
|
36 |
-
|
37 |
# Generate response
|
38 |
outputs = model.generate(**inputs, max_length=200, num_return_sequences=1)
|
39 |
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
40 |
-
|
41 |
return response
|
42 |
except Exception as e:
|
43 |
return f"Error during generation: {e}"
|
@@ -45,6 +43,7 @@ def generate_text(input_text):
|
|
45 |
# Create Gradio Interface
|
46 |
with gr.Blocks() as demo:
|
47 |
gr.Markdown("## Sanskrit Text Generation with Fine-Tuned LLaMA")
|
|
|
48 |
|
49 |
input_text = gr.Textbox(label="Enter your prompt in Sanskrit", placeholder="Type your text here...")
|
50 |
output_text = gr.Textbox(label="Generated Text", interactive=False)
|
|
|
1 |
from peft import PeftModel
|
2 |
from transformers import AutoModelForCausalLM, AutoTokenizer
|
3 |
import gradio as gr
|
4 |
+
import os
|
5 |
+
|
6 |
+
# Fetch Hugging Face token from environment variables
|
7 |
+
token = os.getenv("HUGGING_FACE_HUB_TOKEN")
|
8 |
+
if not token:
|
9 |
+
raise ValueError("Hugging Face token not found. Please add it as a secret in your Hugging Face Space.")
|
10 |
+
|
11 |
+
# Base model and fine-tuned model names
|
12 |
+
base_model_name = "meta-llama/Meta-Llama-3-8B"
|
13 |
+
fine_tuned_model_name = "VinitT/Sanskrit-llama"
|
14 |
|
15 |
# Load the base model, fine-tuned model, and tokenizer
|
16 |
try:
|
|
|
|
|
|
|
|
|
17 |
print("Loading base model...")
|
18 |
+
base_model = AutoModelForCausalLM.from_pretrained(base_model_name, use_auth_token=token)
|
19 |
+
|
20 |
print("Loading fine-tuned model...")
|
21 |
+
model = PeftModel.from_pretrained(base_model, fine_tuned_model_name, use_auth_token=token)
|
22 |
+
|
23 |
print("Loading tokenizer...")
|
24 |
+
tokenizer = AutoTokenizer.from_pretrained(base_model_name, use_auth_token=token)
|
|
|
|
|
|
|
|
|
|
|
25 |
|
26 |
except Exception as e:
|
27 |
+
raise RuntimeError(f"Error loading the model or tokenizer: {e}")
|
|
|
28 |
|
29 |
# Function to generate text using the model
|
30 |
def generate_text(input_text):
|
|
|
|
|
31 |
try:
|
32 |
# Tokenize the input
|
33 |
inputs = tokenizer(input_text, return_tensors="pt")
|
34 |
+
|
35 |
# Generate response
|
36 |
outputs = model.generate(**inputs, max_length=200, num_return_sequences=1)
|
37 |
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
38 |
+
|
39 |
return response
|
40 |
except Exception as e:
|
41 |
return f"Error during generation: {e}"
|
|
|
43 |
# Create Gradio Interface
|
44 |
with gr.Blocks() as demo:
|
45 |
gr.Markdown("## Sanskrit Text Generation with Fine-Tuned LLaMA")
|
46 |
+
gr.Markdown("### Enter your Sanskrit prompt below and generate text using the fine-tuned LLaMA model.")
|
47 |
|
48 |
input_text = gr.Textbox(label="Enter your prompt in Sanskrit", placeholder="Type your text here...")
|
49 |
output_text = gr.Textbox(label="Generated Text", interactive=False)
|