Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -5,8 +5,8 @@ from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
|
|
5 |
import gradio as gr
|
6 |
import json
|
7 |
import faiss
|
8 |
-
import spaces
|
9 |
import numpy as np
|
|
|
10 |
# Ensure you have GPU support
|
11 |
device = 'cuda' if torch.cuda.is_available() else 'cpu'
|
12 |
|
@@ -32,7 +32,6 @@ llama_model = AutoModelForCausalLM.from_pretrained("openai-community/gpt2").to(d
|
|
32 |
summarizer = pipeline("summarization", model="facebook/bart-large-cnn", device=0 if device == 'cuda' else -1)
|
33 |
|
34 |
# Define the function to find the most relevant document using FAISS
|
35 |
-
@spaces.GPU(duration=120)
|
36 |
def retrieve_relevant_doc(query):
|
37 |
query_embedding = sentence_model.encode(query, convert_to_tensor=False)
|
38 |
_, indices = index.search(np.array([query_embedding]), k=1)
|
@@ -40,14 +39,23 @@ def retrieve_relevant_doc(query):
|
|
40 |
return df.iloc[best_match_idx]['Abstract']
|
41 |
|
42 |
# Define the function to generate a response
|
43 |
-
@spaces.GPU(duration=120)
|
44 |
def generate_response(query):
|
45 |
relevant_doc = retrieve_relevant_doc(query)
|
46 |
if len(relevant_doc) > 512: # Truncate long documents
|
47 |
relevant_doc = summarizer(relevant_doc, max_length=150, min_length=50, do_sample=False)[0]['summary_text']
|
|
|
48 |
input_text = f"Document: {relevant_doc}\n\nQuestion: {query}\n\nAnswer:"
|
49 |
inputs = llama_tokenizer(input_text, return_tensors="pt").to(device)
|
50 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
51 |
response = llama_tokenizer.decode(outputs[0], skip_special_tokens=True)
|
52 |
return response
|
53 |
|
|
|
5 |
import gradio as gr
|
6 |
import json
|
7 |
import faiss
|
|
|
8 |
import numpy as np
|
9 |
+
import spaces
|
10 |
# Ensure you have GPU support
|
11 |
device = 'cuda' if torch.cuda.is_available() else 'cpu'
|
12 |
|
|
|
32 |
summarizer = pipeline("summarization", model="facebook/bart-large-cnn", device=0 if device == 'cuda' else -1)
|
33 |
|
34 |
# Define the function to find the most relevant document using FAISS
|
|
|
35 |
def retrieve_relevant_doc(query):
|
36 |
query_embedding = sentence_model.encode(query, convert_to_tensor=False)
|
37 |
_, indices = index.search(np.array([query_embedding]), k=1)
|
|
|
39 |
return df.iloc[best_match_idx]['Abstract']
|
40 |
|
41 |
# Define the function to generate a response
|
|
|
42 |
def generate_response(query):
|
43 |
relevant_doc = retrieve_relevant_doc(query)
|
44 |
if len(relevant_doc) > 512: # Truncate long documents
|
45 |
relevant_doc = summarizer(relevant_doc, max_length=150, min_length=50, do_sample=False)[0]['summary_text']
|
46 |
+
|
47 |
input_text = f"Document: {relevant_doc}\n\nQuestion: {query}\n\nAnswer:"
|
48 |
inputs = llama_tokenizer(input_text, return_tensors="pt").to(device)
|
49 |
+
|
50 |
+
# Set pad_token_id to eos_token_id to avoid the warning
|
51 |
+
pad_token_id = llama_tokenizer.eos_token_id
|
52 |
+
outputs = llama_model.generate(
|
53 |
+
inputs["input_ids"],
|
54 |
+
attention_mask=inputs["attention_mask"],
|
55 |
+
max_length=150,
|
56 |
+
pad_token_id=pad_token_id
|
57 |
+
)
|
58 |
+
|
59 |
response = llama_tokenizer.decode(outputs[0], skip_special_tokens=True)
|
60 |
return response
|
61 |
|