Spaces:
Sleeping
Sleeping
Commit
·
c51e482
1
Parent(s):
a05b3ab
added app.py
Browse files
app.py
ADDED
@@ -0,0 +1,48 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
from llm2vec import LLM2Vec
|
3 |
+
from transformers import AutoTokenizer, AutoModel, AutoConfig
|
4 |
+
from peft import PeftModel
|
5 |
+
import torch
|
6 |
+
import os
|
7 |
+
|
8 |
+
torch.backends.cuda.enable_mem_efficient_sdp(False)
|
9 |
+
torch.backends.cuda.enable_flash_sdp(False)
|
10 |
+
|
11 |
+
# Read tokens from environment variables
|
12 |
+
GROQ_API_KEY = os.getenv('GROQ_API_KEY')
|
13 |
+
HF_TOKEN = os.getenv('HF_TOKEN')
|
14 |
+
|
15 |
+
if not GROQ_API_KEY or not HF_TOKEN:
|
16 |
+
raise ValueError("GROQ_API_KEY and HF_TOKEN must be set as environment variables.")
|
17 |
+
|
18 |
+
os.environ['GROQ_API_KEY'] = GROQ_API_KEY
|
19 |
+
os.environ['HF_TOKEN'] = HF_TOKEN
|
20 |
+
|
21 |
+
|
22 |
+
# Load tokenizer and model
|
23 |
+
tokenizer = AutoTokenizer.from_pretrained("McGill-NLP/LLM2Vec-Sheared-LLaMA-mntp")
|
24 |
+
config = AutoConfig.from_pretrained("McGill-NLP/LLM2Vec-Sheared-LLaMA-mntp", trust_remote_code=True)
|
25 |
+
model = AutoModel.from_pretrained("McGill-NLP/LLM2Vec-Sheared-LLaMA-mntp", trust_remote_code=True, config=config, torch_dtype=torch.bfloat16, device_map="cuda" if torch.cuda.is_available() else "cpu")
|
26 |
+
|
27 |
+
model = PeftModel.from_pretrained(model, "McGill-NLP/LLM2Vec-Sheared-LLaMA-mntp")
|
28 |
+
model = model.merge_and_unload()
|
29 |
+
|
30 |
+
# Load unsupervised SimCSE model
|
31 |
+
model = PeftModel.from_pretrained(model, "McGill-NLP/LLM2Vec-Sheared-LLaMA-mntp-unsup-simcse")
|
32 |
+
|
33 |
+
# Wrapper for encoding and pooling operations
|
34 |
+
l2v = LLM2Vec(model, tokenizer, pooling_mode="mean", max_length=512)
|
35 |
+
|
36 |
+
def encode_text(input_text):
|
37 |
+
encoding = l2v.encode(input_text)
|
38 |
+
return encoding
|
39 |
+
|
40 |
+
# Define Gradio interface
|
41 |
+
iface = gr.Interface(
|
42 |
+
fn=encode_text,
|
43 |
+
inputs=gr.inputs.Textbox(lines=2, placeholder="Enter text here..."),
|
44 |
+
outputs=gr.outputs.JSON()
|
45 |
+
)
|
46 |
+
|
47 |
+
# Launch Gradio app
|
48 |
+
iface.launch(share=True)
|