matrix-pygmalion-bot/matrix_pygmalion_bot/ai/stablehorde.py

import asyncio
import os, tempfile
import logging

import requests

from transformers import AutoTokenizer, AutoConfig
from huggingface_hub import hf_hub_download

logger = logging.getLogger(__name__)


async def generate_sync(
    prompt: str,
    api_key: str,
    bot_name: str,
):
    # Set the API endpoint URL
    endpoint = "https://koboldai.net/api/v2/generate/async"

    # Set the headers for the request
    headers = {
        "Content-Type": "application/json",
        "accept": "application/json",
        "apikey": f"{api_key}"
    }

    max_new_tokens = 200
    prompt_num_tokens = await num_tokens(prompt)

    # Define your inputs
    input_data = {
        "prompt": prompt,
        "params": {
            "n": 1,
#            "frmtadsnsp": False,
#            "frmtrmblln": False,
#            "frmtrmspch": False,
#            "frmttriminc": False,
            "max_context_length": 1024,
            "max_length": 512,
            "rep_pen": 1.1,
            "rep_pen_range": 1024,
            "rep_pen_slope": 0.7,
#            "singleline": False,
#            "soft_prompt": "",
            "temperature": 0.75,
            "tfs": 1.0,
            "top_a": 0.0,
            "top_k": 0,
            "top_p": 0.9,
            "typical": 1.0,
#            "sampler_order": [0],
        },
        "softprompts": [],
        "trusted_workers": False,
        "nsfw": True,
#        "workers": [],
        "models": ["PygmalionAI/pygmalion-6b"]
    }


    logger.info(f"sending request to koboldai.net")

    # Make the request
    r = requests.post(endpoint, json=input_data, headers=headers, timeout=180)

    r_json = r.json()
    logger.info(r_json)
    status = r_json["message"]

    if "id" in r_json:
        job_id = r_json["id"]
        TIMEOUT = 360
        DELAY = 11
        for i in range(TIMEOUT//DELAY):
            endpoint = "https://koboldai.net/api/v2/generate/text/status/" + job_id
            r = requests.get(endpoint, headers=headers)
            r_json = r.json()
            logger.info(r_json)
            if "done" not in r_json:
                return "<ERROR>"
            if r_json["done"] == True:
                text = r_json["generations"][0]["text"]
                answer = text.removeprefix(prompt)
                idx = answer.find(f"\nYou:")
                if idx != -1:
                    reply = answer[:idx].strip()
                else:
                    reply = answer.removesuffix('<|endoftext|>').strip()
                reply = reply.replace(f"\n{bot_name}: ", " ")
                reply.replace(f"\n<BOT>: ", " ")
                return reply
            else:
                await asyncio.sleep(DELAY)
    else:
        return f"<ERROR> {status}"

async def get_full_prompt(simple_prompt: str, bot, chat_history):

    # Prompt without history
    prompt = bot.name + "'s Persona: " + bot.get_persona() + "\n"
    prompt += "Scenario: " + bot.get_scenario() + "\n"
    prompt += "<START>" + "\n"
    #prompt += bot.name + ": " + bot.greeting + "\n"
    prompt += "You: " + simple_prompt + "\n"
    prompt += bot.name + ":"

    MAX_TOKENS = 2048
    max_new_tokens = 200
    total_num_tokens = await num_tokens(prompt)
    visible_history = []
    current_message = True
    for key, chat_item in reversed(chat_history.chat_history.items()):
        if current_message:
            current_message = False
            continue
        if chat_item.message["en"].startswith('!begin'):
            break
        if chat_item.message["en"].startswith('!'):
            continue
        if chat_item.message["en"].startswith('<ERROR>'):
            continue
        #if chat_item.message["en"] == bot.greeting:
        #    continue
        if chat_item.num_tokens == None:
            chat_item.num_tokens = await num_tokens("{}: {}".format(chat_item.user_name, chat_item.message["en"]))
        # TODO: is it MAX_TOKENS or MAX_TOKENS - max_new_tokens??
        logger.debug(f"History: " + str(chat_item) + " [" + str(chat_item.num_tokens) + "]")
        if total_num_tokens + chat_item.num_tokens < MAX_TOKENS - max_new_tokens:
            visible_history.append(chat_item)
            total_num_tokens += chat_item.num_tokens
        else:
            break
    visible_history = reversed(visible_history)

    prompt = bot.name + "'s Persona: " + bot.get_persona() + "\n"
    prompt += "Scenario: " + bot.get_scenario() + "\n"
    prompt += "<START>" + "\n"
    #prompt += bot.name + ": " + bot.greeting + "\n"
    for chat_item in visible_history:
        if chat_item.is_own_message:
            prompt += bot.name + ": " + chat_item.message["en"] + "\n"
        else:
            prompt += "You" + ": " + chat_item.message["en"] + "\n"
    prompt += "You: " + simple_prompt + "\n"
    prompt += bot.name + ":"

    return prompt


async def num_tokens(input_text: str):
#    os.makedirs("./models/pygmalion-6b", exist_ok=True)
#    hf_hub_download(repo_id="PygmalionAI/pygmalion-6b", filename="config.json", cache_dir="./models/pygmalion-6b")
#    config = AutoConfig.from_pretrained("./models/pygmalion-6b/config.json")
    tokenizer = AutoTokenizer.from_pretrained("PygmalionAI/pygmalion-6b")
    encoding = tokenizer.encode(input_text, add_special_tokens=False)
    max_input_size = tokenizer.max_model_input_sizes
    return len(encoding)

async def estimate_num_tokens(input_text: str):
    return len(input_text)//4+1


async def generate_image(input_prompt: str, negative_prompt: str, model: str, api_key: str):

    # Set the API endpoint URL
    endpoint = "https://stablehorde.net/api/v2/generate/async"

    # Set the headers for the request
    headers = {
        "Content-Type": "application/json",
        "accept": "application/json",
        "apikey": f"{api_key}"
    }

    # Define your inputs
    input_data = {
        "prompt": input_prompt,
        "params": {
#            "negative_prompt": negative_prompt,
            "width": 512,
            "height": 512,
        },
        "nsfw": True,
        "trusted_workers": False,
#        "workers": [],
        "models": [f"{model}"]
    }

    logger.info(f"sending request to stablehorde.net")

    # Make the request
    r = requests.post(endpoint, json=input_data, headers=headers)
    r_json = r.json()
    logger.info(r_json)

    if r.status_code == 202:
        #status = r_json["message"]
        job_id = r_json["id"]
        TIMEOUT = 360
        DELAY = 11
        output = None
        for i in range(TIMEOUT//DELAY):
            endpoint = "https://stablehorde.net/api/v2/generate/status/" + job_id
            r = requests.get(endpoint, headers=headers)
            r_json = r.json()
            logger.info(r_json)
            #status = r_json["message"]
            if "done" not in r_json:
                return "<ERROR>"
            if "faulted" in r_json and r_json["faulted"] == True:
                return "<ERROR>"
            if r_json["done"] == True:
                output = r_json["generations"]
                break
            else:
                await asyncio.sleep(DELAY)

        if not output:
            raise ValueError(f"<ERROR>")

        os.makedirs("./images", exist_ok=True)
        files = []
        for image in output:
            temp_name = next(tempfile._get_candidate_names())
            filename = "./images/" + temp_name + ".jpg"
            await download_image(image["img"], filename)
            files.append(filename)

        return files

async def generate_image1(input_prompt: str, negative_prompt: str, api_key: str):
    return await generate_image(input_prompt, negative_prompt, "Deliberate", api_key)

async def generate_image2(input_prompt: str, negative_prompt: str, api_key: str):
    return await generate_image(input_prompt, negative_prompt, "PFG", api_key)

async def generate_image3(input_prompt: str, negative_prompt: str, api_key: str):
    return await generate_image(input_prompt, negative_prompt, "Hassanblend", api_key)

async def download_image(url, path):
    r = requests.get(url, stream=True)
    if r.status_code == 200:
        with open(path, 'wb') as f:
            for chunk in r:
                f.write(chunk)