Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -1,41 +1,56 @@
|
|
1 |
import streamlit as st
|
2 |
-
from transformers import T5Tokenizer, T5ForConditionalGeneration
|
3 |
import torch
|
|
|
4 |
|
5 |
-
#
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
6 |
@st.cache_resource
|
7 |
def load_model():
|
8 |
model_name = "flax-community/t5-recipe-generation"
|
9 |
-
|
10 |
-
|
11 |
-
|
12 |
-
|
13 |
-
|
14 |
-
|
15 |
-
|
|
|
|
|
|
|
|
|
16 |
|
17 |
-
# Generate recipe function with
|
18 |
def generate_recipe(ingredients, tokenizer, model, max_length=512):
|
19 |
# Prepare input
|
20 |
input_text = f"Generate recipe with: {ingredients}"
|
21 |
|
22 |
-
|
23 |
-
|
24 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
25 |
|
26 |
-
#
|
27 |
-
|
28 |
-
|
29 |
-
|
30 |
-
|
31 |
-
|
32 |
-
num_beams=4, # Reduced beam search for faster CPU processing
|
33 |
-
early_stopping=True
|
34 |
-
)
|
35 |
-
|
36 |
-
# Decode and clean the output
|
37 |
-
recipe = tokenizer.decode(output_ids[0], skip_special_tokens=True)
|
38 |
-
return recipe
|
39 |
|
40 |
# Streamlit app
|
41 |
def main():
|
@@ -57,9 +72,12 @@ def main():
|
|
57 |
with st.spinner("Generating recipe..."):
|
58 |
recipe = generate_recipe(ingredients_input, tokenizer, model)
|
59 |
|
60 |
-
|
61 |
-
|
62 |
-
|
|
|
|
|
|
|
63 |
else:
|
64 |
st.warning("Please enter some ingredients!")
|
65 |
|
|
|
1 |
import streamlit as st
|
|
|
2 |
import torch
|
3 |
+
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
|
4 |
|
5 |
+
# Ensure SentencePiece is installed
|
6 |
+
try:
|
7 |
+
import sentencepiece
|
8 |
+
except ImportError:
|
9 |
+
st.error("SentencePiece is not installed. Please install it using: pip install sentencepiece")
|
10 |
+
st.stop()
|
11 |
+
|
12 |
+
# Load the model and tokenizer with caching
|
13 |
@st.cache_resource
|
14 |
def load_model():
|
15 |
model_name = "flax-community/t5-recipe-generation"
|
16 |
+
try:
|
17 |
+
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
18 |
+
model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
|
19 |
+
|
20 |
+
# Explicitly set to CPU and use float32 to reduce memory usage
|
21 |
+
model = model.to('cpu').float()
|
22 |
+
|
23 |
+
return tokenizer, model
|
24 |
+
except Exception as e:
|
25 |
+
st.error(f"Error loading model: {e}")
|
26 |
+
st.stop()
|
27 |
|
28 |
+
# Generate recipe function with error handling
|
29 |
def generate_recipe(ingredients, tokenizer, model, max_length=512):
|
30 |
# Prepare input
|
31 |
input_text = f"Generate recipe with: {ingredients}"
|
32 |
|
33 |
+
try:
|
34 |
+
# Use torch no_grad to reduce memory consumption
|
35 |
+
with torch.no_grad():
|
36 |
+
input_ids = tokenizer.encode(input_text, return_tensors="pt", max_length=max_length, truncation=True)
|
37 |
+
|
38 |
+
# Adjust generation parameters for faster CPU inference
|
39 |
+
output_ids = model.generate(
|
40 |
+
input_ids,
|
41 |
+
max_length=max_length,
|
42 |
+
num_return_sequences=1,
|
43 |
+
no_repeat_ngram_size=2,
|
44 |
+
num_beams=4, # Reduced beam search for faster CPU processing
|
45 |
+
early_stopping=True
|
46 |
+
)
|
47 |
|
48 |
+
# Decode and clean the output
|
49 |
+
recipe = tokenizer.decode(output_ids[0], skip_special_tokens=True)
|
50 |
+
return recipe
|
51 |
+
except Exception as e:
|
52 |
+
st.error(f"Error generating recipe: {e}")
|
53 |
+
return None
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
54 |
|
55 |
# Streamlit app
|
56 |
def main():
|
|
|
72 |
with st.spinner("Generating recipe..."):
|
73 |
recipe = generate_recipe(ingredients_input, tokenizer, model)
|
74 |
|
75 |
+
if recipe:
|
76 |
+
# Display recipe sections
|
77 |
+
st.subheader("π₯ Generated Recipe")
|
78 |
+
st.write(recipe)
|
79 |
+
else:
|
80 |
+
st.error("Failed to generate recipe. Please try again.")
|
81 |
else:
|
82 |
st.warning("Please enter some ingredients!")
|
83 |
|