CoffeeVampir3

CoffeeVampir3

for text, inst, summary in zip(text, instructions, previous_summary):
    if summary != "":
        summarized_augmentation = random.choice(augmented_continuation)
        inst = f"{inst}\n\n{summarized_augmentation} {summary}"

    next_prompt = copy.deepcopy(enc_sys_prompt)
    next_message = encode_message(tokenizer, "user", inst)
    next_prompt.extend(next_message)
    mask_length = len(next_prompt)
    next_prompt.extend(encode_message_english_sentence_truncate(tokenizer, "assistant", text, mask_length, 8150))

CoffeeVampir3

from flask import Flask, render_template
import torch
from flask_socketio import SocketIO, emit
from generation.make_instruct import get_generator_func
from generation.exllama_generator_wrapper import encode_message, encode_system, encode_header
import os,sys

app = Flask(__name__)
socketio = SocketIO(app)
system_prompt = "Respond to all inputs with EEE"

CoffeeVampir3

from transformers import AutoTokenizer
import json
import sys

model = "/home/blackroot/Desktop/llama3-8b/llama-3-8b"
max_tokens = 8192

def count_tokens_hf(text: str, model_name: str) -> int:
    tokenizer = AutoTokenizer.from_pretrained(model_name)
    encoded_input = tokenizer.encode(text)

CoffeeVampir3

CoffeeVampir3

def load_cfg_from_json(json_file):
    with open(json_file, "r", encoding="utf-8") as reader:
        text = reader.read()
    return json.loads(text)

def load_cfg(model_id, cfg_path):
    hf_config = load_cfg_from_json(cfg_path)
    if 'pretrained_cfg' not in hf_config:
        # old form, pull pretrain_cfg out of the base dict
        pretrained_cfg = hf_config

CoffeeVampir3

#!/bin/bash

# Config Start

# Configurations
ckpt="/home/blackroot/Desktop/SD/sd-scripts/models/model.safetensors" # base checkpoint to finetune
image_dir="/home/blackroot/Desktop/SD/sd-scripts/train_data" # folder containing folders with repeats_conceptname
reg_dir="" #optional, just point this to an empty folder if you don't care
output="/home/blackroot/Desktop/SD/sd-scripts/outputs" # safetensors output folder

CoffeeVampir3

def per_polygon_translate(generate_fn, input):
    prompt = (
f"""::JAPANESE TEXT::
むし、 ある ところ に
[[0.0, 8.0], [252.0, 8.0], [252.0, 27.0], [0.0, 26.0]]
おじいさん と おばあさん が いました。
[[0.0, 33.0], [289.0, 32.0], [289.0, 50.0], [0.0, 50.0]]
おじいさん が 山（やま） へ 木（き） を きり に いけば、
[[0.0, 57.0], [416.0, 56.0], [417.0, 74.0], [0.0, 75.0]]
おばあさん は 川（かわ）へ せんたく に でかけます。

CoffeeVampir3

from diffusers import StableDiffusionXLPipeline, StableDiffusionXLImg2ImgPipeline
from diffusers import EulerAncestralDiscreteScheduler
import torch
import torch.nn.functional
import gc
from PIL import Image

from mechanisms.tokenizer_utils import encode_from_pipe

@torch.no_grad()

CoffeeVampir3

import re, os, random, csv

class WildCLIP:
    @classmethod
    def INPUT_TYPES(s):
        return {
        "required": {
            "text": ("STRING", {"multiline": True}), "clip": ("CLIP", )
        }
        }