Jackss
commited on
Commit
•
1dd5fed
1
Parent(s):
12bb371
Added HTML
Browse files- Dockerfile +10 -1
- main.py +28 -6
- requirements.txt +2 -1
- static/index.html +10 -0
Dockerfile
CHANGED
|
@@ -6,6 +6,15 @@ COPY ./requirements.txt /code/requirements.txt
|
|
| 6 |
|
| 7 |
RUN pip install --no-cache-dir --upgrade -r /code/requirements.txt
|
| 8 |
|
| 9 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 10 |
|
| 11 |
CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "7860"]
|
|
|
|
| 6 |
|
| 7 |
RUN pip install --no-cache-dir --upgrade -r /code/requirements.txt
|
| 8 |
|
| 9 |
+
RUN useradd -m -u 1000 user
|
| 10 |
+
|
| 11 |
+
USER user
|
| 12 |
+
|
| 13 |
+
ENV HOME=/home/user \
|
| 14 |
+
PATH=/home/user/.local/bin:$PATH
|
| 15 |
+
|
| 16 |
+
WORKDIR $HOME/app
|
| 17 |
+
|
| 18 |
+
COPY --chown=user . $HOME/app
|
| 19 |
|
| 20 |
CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "7860"]
|
main.py
CHANGED
|
@@ -1,14 +1,36 @@
|
|
| 1 |
from fastapi import FastAPI
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
|
| 3 |
-
model_name = 'allenai/specter'
|
| 4 |
|
| 5 |
app = FastAPI()
|
| 6 |
|
| 7 |
-
|
| 8 |
-
|
| 9 |
-
|
|
|
|
|
|
|
| 10 |
|
| 11 |
@app.get('/similarity')
|
| 12 |
def similarity(input):
|
| 13 |
-
|
| 14 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
from fastapi import FastAPI
|
| 2 |
+
from fastapi.staticfiles import StaticFiles
|
| 3 |
+
from fastapi.responses import FileResponse
|
| 4 |
+
from transformers import AutoTokenizer, AutoModel
|
| 5 |
+
import numpy as np
|
| 6 |
+
from sklearn.metrics.pairwise import cosine_similarity
|
| 7 |
+
|
| 8 |
+
# load model and tokenizer
|
| 9 |
+
tokenizer = AutoTokenizer.from_pretrained('allenai/specter')
|
| 10 |
+
model = AutoModel.from_pretrained('allenai/specter')
|
| 11 |
+
|
| 12 |
+
# papers = [{'title': 'BERT', 'abstract': 'We introduce a new language representation model called BERT'},
|
| 13 |
+
# {'title': 'Attention is all you need', 'abstract': ' The dominant sequence transduction models are based on complex recurrent or convolutional neural networks'}]
|
| 14 |
+
|
| 15 |
+
# concatenate title and abstract
|
| 16 |
|
|
|
|
| 17 |
|
| 18 |
app = FastAPI()
|
| 19 |
|
| 20 |
+
app.mount("/", StaticFiles(directory="static", html=True), name="static")
|
| 21 |
+
|
| 22 |
+
@app.get("/")
|
| 23 |
+
def index() -> FileResponse:
|
| 24 |
+
return FileResponse(path="/app/static/index.html", media_type="text/html")
|
| 25 |
|
| 26 |
@app.get('/similarity')
|
| 27 |
def similarity(input):
|
| 28 |
+
papers = input['papers']
|
| 29 |
+
title_abs = [d['title'] + tokenizer.sep_token + (d.get('abstract') or '') for d in papers]
|
| 30 |
+
# preprocess the input
|
| 31 |
+
inputs = tokenizer(title_abs, padding=True, truncation=True, return_tensors="pt", max_length=512)
|
| 32 |
+
result = model(**inputs)
|
| 33 |
+
# take the first token in the batch as the embedding
|
| 34 |
+
embeddings = result.last_hidden_state[:, 0, :]
|
| 35 |
+
res = cosine_similarity(embeddings, embeddings).tolist()
|
| 36 |
+
return {"output": res}
|
requirements.txt
CHANGED
|
@@ -3,4 +3,5 @@ requests==2.27.*
|
|
| 3 |
sentencepiece==0.1.*
|
| 4 |
torch==1.11.*
|
| 5 |
transformers==4.*
|
| 6 |
-
uvicorn[standard]==0.17.*
|
|
|
|
|
|
| 3 |
sentencepiece==0.1.*
|
| 4 |
torch==1.11.*
|
| 5 |
transformers==4.*
|
| 6 |
+
uvicorn[standard]==0.17.*
|
| 7 |
+
scikit-learn==1.2.*
|
static/index.html
ADDED
|
@@ -0,0 +1,10 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
<html lang="en">
|
| 2 |
+
<head>
|
| 3 |
+
<meta charset="UTF-8">
|
| 4 |
+
<meta name="viewport" content="width=device-width, initial-scale=1.0">
|
| 5 |
+
<title>Document</title>
|
| 6 |
+
</head>
|
| 7 |
+
<body>
|
| 8 |
+
Hello world!
|
| 9 |
+
</body>
|
| 10 |
+
</html>
|