summaryrefslogtreecommitdiffstats
path: root/g4f
diff options
context:
space:
mode:
authorabc <98614666+xtekky@users.noreply.github.com>2023-10-06 20:52:39 +0200
committerabc <98614666+xtekky@users.noreply.github.com>2023-10-06 20:52:39 +0200
commitb04ff7fb53e0370ebe6d558060bd0495f41273de (patch)
tree96119382765f4ba54196bf5da0036f1a0902d339 /g4f
parent~ | new g4f GUI (diff)
downloadgpt4free-b04ff7fb53e0370ebe6d558060bd0495f41273de.tar
gpt4free-b04ff7fb53e0370ebe6d558060bd0495f41273de.tar.gz
gpt4free-b04ff7fb53e0370ebe6d558060bd0495f41273de.tar.bz2
gpt4free-b04ff7fb53e0370ebe6d558060bd0495f41273de.tar.lz
gpt4free-b04ff7fb53e0370ebe6d558060bd0495f41273de.tar.xz
gpt4free-b04ff7fb53e0370ebe6d558060bd0495f41273de.tar.zst
gpt4free-b04ff7fb53e0370ebe6d558060bd0495f41273de.zip
Diffstat (limited to '')
-rw-r--r--g4f/interference/__init__.py (renamed from interference/app.py)81
1 files changed, 6 insertions, 75 deletions
diff --git a/interference/app.py b/g4f/interference/__init__.py
index 5abbcff2..d756faa7 100644
--- a/interference/app.py
+++ b/g4f/interference/__init__.py
@@ -2,17 +2,19 @@ import json
import time
import random
import string
-import requests
from typing import Any
from flask import Flask, request
from flask_cors import CORS
-from transformers import AutoTokenizer
from g4f import ChatCompletion
app = Flask(__name__)
CORS(app)
+@app.route('/')
+def index():
+ return 'interference api, url: http://127.0.0.1:1337'
+
@app.route('/chat/completions', methods=['POST'])
def chat_completions():
model = request.get_json().get('model', 'gpt-3.5-turbo')
@@ -88,76 +90,5 @@ def chat_completions():
return app.response_class(streaming(), mimetype='text/event-stream')
-
-# Get the embedding from huggingface
-def get_embedding(input_text, token):
- huggingface_token = token
- embedding_model = 'sentence-transformers/all-mpnet-base-v2'
- max_token_length = 500
-
- # Load the tokenizer for the 'all-mpnet-base-v2' model
- tokenizer = AutoTokenizer.from_pretrained(embedding_model)
- # Tokenize the text and split the tokens into chunks of 500 tokens each
- tokens = tokenizer.tokenize(input_text)
- token_chunks = [tokens[i:i + max_token_length]
- for i in range(0, len(tokens), max_token_length)]
-
- # Initialize an empty list
- embeddings = []
-
- # Create embeddings for each chunk
- for chunk in token_chunks:
- # Convert the chunk tokens back to text
- chunk_text = tokenizer.convert_tokens_to_string(chunk)
-
- # Use the Hugging Face API to get embeddings for the chunk
- api_url = f'https://api-inference.huggingface.co/pipeline/feature-extraction/{embedding_model}'
- headers = {'Authorization': f'Bearer {huggingface_token}'}
- chunk_text = chunk_text.replace('\n', ' ')
-
- # Make a POST request to get the chunk's embedding
- response = requests.post(api_url, headers=headers, json={
- 'inputs': chunk_text, 'options': {'wait_for_model': True}})
-
- # Parse the response and extract the embedding
- chunk_embedding = response.json()
- # Append the embedding to the list
- embeddings.append(chunk_embedding)
-
- # averaging all the embeddings
- # this isn't very effective
- # someone a better idea?
- num_embeddings = len(embeddings)
- average_embedding = [sum(x) / num_embeddings for x in zip(*embeddings)]
- embedding = average_embedding
- return embedding
-
-
-@app.route('/embeddings', methods=['POST'])
-def embeddings():
- input_text_list = request.get_json().get('input')
- input_text = ' '.join(map(str, input_text_list))
- token = request.headers.get('Authorization').replace('Bearer ', '')
- embedding = get_embedding(input_text, token)
-
- return {
- 'data': [
- {
- 'embedding': embedding,
- 'index': 0,
- 'object': 'embedding'
- }
- ],
- 'model': 'text-embedding-ada-002',
- 'object': 'list',
- 'usage': {
- 'prompt_tokens': None,
- 'total_tokens': None
- }
- }
-
-def main():
- app.run(host='0.0.0.0', port=1337, debug=True)
-
-if __name__ == '__main__':
- main() \ No newline at end of file
+def run_interference():
+ app.run(host='0.0.0.0', port=1337, debug=True) \ No newline at end of file