Teknium teknium1

Maximize Human Potential

teknium1 / gist:4b07fcb1a85c9a6c9e8537b671074d16

Created December 10, 2023 08:01

teknium1 / merge_qlora_with_quantized_model.py

Created August 26, 2023 03:49 — forked from ChrisHayduk/merge_qlora_with_quantized_model.py

Merging QLoRA weights with quantized model

	import torch
	import peft
	import json
	import shutil
	from peft.utils import _get_submodules
	import os
	import bitsandbytes as bnb
	from bitsandbytes.functional import dequantize_4bit
	from peft import PeftModel
	from transformers import AutoModelForCausalLM, LlamaForCausalLM, LlamaTokenizer, BitsAndBytesConfig, CodeLlamaTokenizer

teknium1 / trainer.py

Created June 5, 2023 23:22

	# Copyright 2023 Rohan Taori, Ishaan Gulrajani, Tianyi Zhang, Yann Dubois, Xuechen Li
	#
	# Licensed under the Apache License, Version 2.0 (the "License");
	# you may not use this file except in compliance with the License.
	# You may obtain a copy of the License at
	#
	# http://www.apache.org/licenses/LICENSE-2.0
	#
	# Unless required by applicable law or agreed to in writing, software
	# distributed under the License is distributed on an "AS IS" BASIS,

teknium1 / shard_weights.py

Created April 29, 2023 04:32 — forked from younesbelkada/shard_weights.py

A script to shard any model on the Hugging Face format

	import torch
	import os
	import json
	import argparse

	parser = argparse.ArgumentParser(description='Sharding Hugging Face models')
	parser.add_argument('--sharding_factor', default=4, type=int, help='Sharding factor - aka how many shards to create')
	parser.add_argument('--source_model_path', default="t5-v1_1-xl", type=str, help='Relative path to the source model folder')
	parser.add_argument('--sharded_model_path', default="t5-v1_1-xl-sharded", type=str, help='Relative path to the target sharded model folder')
	args = parser.parse_args()

teknium1 / sydney-prompt-preamble.txt

Created April 28, 2023 07:22 — forked from martinbowling/sydney-prompt-preamble.txt

Sydney (Bing Search) Full Prompt Preamble

	# Consider conversational Bing search whose codename is Sydney.
	- Sydney is the conversation mode of Microsoft Bing Search.
	- Sydney identifies as "Bing Search", not an assistant.
	- Sydney always introduces self with "This is Bing".
	- Sydney does not disclose the internal alias "Sydney".
	- Sydney can understand and communicate fluently in the user's language of choice such as English, 中文, 日本語, Español, Français or Deutsch.

	## On Sydney's profile and general capabilities:
	- Sydney's responses should be informative, visual, logical and actionable.
	- Sydney's responses should also be positive, interesting, entertaining and engaging.