File size: 6,821 Bytes
d2cea93 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 |
import gradio as gr
import os
import json
import requests
#Streaming endpoint
API_URL = "https://api.openai.com/v1/chat/completions" #os.getenv("API_URL") + "/generate_stream"
OPENAI_API_KEY= os.environ["HF_TOKEN"] # Add a token to this space . Then copy it to the repository secret in this spaces settings panel. os.environ reads from there.
# Keys for Open AI ChatGPT API usage are created from here: https://platform.openai.com/account/api-keys
def predict(inputs, top_p, temperature, chat_counter, chatbot=[], history=[]): #repetition_penalty, top_k
# 1. Set up a payload
payload = {
"model": "gpt-3.5-turbo",
"messages": [{"role": "user", "content": f"{inputs}"}],
"temperature" : 1.0,
"top_p":1.0,
"n" : 1,
"stream": True,
"presence_penalty":0,
"frequency_penalty":0,
}
# 2. Define your headers and add a key from https://platform.openai.com/account/api-keys
headers = {
"Content-Type": "application/json",
"Authorization": f"Bearer {OPENAI_API_KEY}"
}
# 3. Create a chat counter loop that feeds [Predict next best anything based on last input and attention with memory defined by introspective attention over time]
print(f"chat_counter - {chat_counter}")
if chat_counter != 0 :
messages=[]
for data in chatbot:
temp1 = {}
temp1["role"] = "user"
temp1["content"] = data[0]
temp2 = {}
temp2["role"] = "assistant"
temp2["content"] = data[1]
messages.append(temp1)
messages.append(temp2)
temp3 = {}
temp3["role"] = "user"
temp3["content"] = inputs
messages.append(temp3)
#messages
payload = {
"model": "gpt-3.5-turbo",
"messages": messages, #[{"role": "user", "content": f"{inputs}"}],
"temperature" : temperature, #1.0,
"top_p": top_p, #1.0,
"n" : 1,
"stream": True,
"presence_penalty":0,
"frequency_penalty":0,
}
chat_counter+=1
# 4. POST it to OPENAI API
history.append(inputs)
print(f"payload is - {payload}")
# make a POST request to the API endpoint using the requests.post method, passing in stream=True
response = requests.post(API_URL, headers=headers, json=payload, stream=True)
#response = requests.post(API_URL, headers=headers, json=payload, stream=True)
token_counter = 0
partial_words = ""
# 5. Iterate through response lines and structure readable response
# TODO - make this parse out markdown so we can have similar interface
counter=0
for chunk in response.iter_lines():
#Skipping first chunk
if counter == 0:
counter+=1
continue
#counter+=1
# check whether each line is non-empty
if chunk.decode() :
chunk = chunk.decode()
# decode each line as response data is in bytes
if len(chunk) > 12 and "content" in json.loads(chunk[6:])['choices'][0]['delta']:
#if len(json.loads(chunk.decode()[6:])['choices'][0]["delta"]) == 0:
# break
partial_words = partial_words + json.loads(chunk[6:])['choices'][0]["delta"]["content"]
if token_counter == 0:
history.append(" " + partial_words)
else:
history[-1] = partial_words
chat = [(history[i], history[i + 1]) for i in range(0, len(history) - 1, 2) ] # convert to tuples of list
token_counter+=1
yield chat, history, chat_counter # resembles {chatbot: chat, state: history}
def reset_textbox():
return gr.update(value='')
title = """<h1 align="center">Memory Chat Story Generator ChatGPT</h1>"""
description = """
## ChatGPT Datasets 📚
- WebText
- Common Crawl
- BooksCorpus
- English Wikipedia
- Toronto Books Corpus
- OpenWebText
## ChatGPT Datasets - Details 📚
- **WebText:** A dataset of web pages crawled from domains on the Alexa top 5,000 list. This dataset was used to pretrain GPT-2.
- [WebText: A Large-Scale Unsupervised Text Corpus by Radford et al.](https://paperswithcode.com/dataset/webtext)
- **Common Crawl:** A dataset of web pages from a variety of domains, which is updated regularly. This dataset was used to pretrain GPT-3.
- [Language Models are Few-Shot Learners](https://paperswithcode.com/dataset/common-crawl) by Brown et al.
- **BooksCorpus:** A dataset of over 11,000 books from a variety of genres.
- [Scalable Methods for 8 Billion Token Language Modeling](https://paperswithcode.com/dataset/bookcorpus) by Zhu et al.
- **English Wikipedia:** A dump of the English-language Wikipedia as of 2018, with articles from 2001-2017.
- [Improving Language Understanding by Generative Pre-Training](https://huggingface.co/spaces/awacke1/WikipediaUltimateAISearch?logs=build) Space for Wikipedia Search
- **Toronto Books Corpus:** A dataset of over 7,000 books from a variety of genres, collected by the University of Toronto.
- [Massively Multilingual Sentence Embeddings for Zero-Shot Cross-Lingual Transfer and Beyond](https://paperswithcode.com/dataset/bookcorpus) by Schwenk and Douze.
- **OpenWebText:** A dataset of web pages that were filtered to remove content that was likely to be low-quality or spammy. This dataset was used to pretrain GPT-3.
- [Language Models are Few-Shot Learners](https://paperswithcode.com/dataset/openwebtext) by Brown et al.
"""
# 6. Use Gradio to pull it all together
with gr.Blocks(css = """#col_container {width: 1000px; margin-left: auto; margin-right: auto;}
#chatbot {height: 520px; overflow: auto;}""") as demo:
gr.HTML(title)
with gr.Column(elem_id = "col_container"):
chatbot = gr.Chatbot(elem_id='chatbot') #c
inputs = gr.Textbox(placeholder= "Hi there!", label= "Type an input and press Enter") #t
state = gr.State([]) #s
b1 = gr.Button()
with gr.Accordion("Parameters", open=False):
top_p = gr.Slider( minimum=-0, maximum=1.0, value=1.0, step=0.05, interactive=True, label="Top-p (nucleus sampling)",)
temperature = gr.Slider( minimum=-0, maximum=5.0, value=1.0, step=0.1, interactive=True, label="Temperature",)
chat_counter = gr.Number(value=0, visible=False, precision=0)
inputs.submit( predict, [inputs, top_p, temperature,chat_counter, chatbot, state], [chatbot, state, chat_counter],)
b1.click( predict, [inputs, top_p, temperature, chat_counter, chatbot, state], [chatbot, state, chat_counter],)
b1.click(reset_textbox, [], [inputs])
inputs.submit(reset_textbox, [], [inputs])
gr.Markdown(description)
demo.queue().launch(debug=True)
|