chavinlo

# Install sudo first
sudo apt update
sudo apt install software-properties-common -y
sudo add-apt-repository --yes ppa:deadsnakes/ppa
sudo apt update && sudo apt install python3.10 python3.10-venv python3.10-dev -y
python3.10 -m venv env
source env/bin/activate

chavinlo

#This module is meant for direct use only. For API-usage please check SDA-TRAINER.
#Based off NVIDIA's demo
import argparse
from threads.trt.models import CLIP, UNet, VAE
import os
import onnx
import torch
from diffusers import UNet2DConditionModel, AutoencoderKL
from transformers import CLIPTextModel
from threads.trt.utilities import Engine

chavinlo

diffusers>=0.5.1
numpy==1.23.4
wandb==0.13.4
torch
torchvision
transformers>=4.21.0
huggingface-hub>=0.10.0
Pillow==9.2.0
tqdm==4.64.1
ftfy==6.1.1

chavinlo

# Install bitsandbytes:
# `nvcc --version` to get CUDA version.
# `pip install -i https://test.pypi.org/simple/ bitsandbytes-cudaXXX` to install for current CUDA.
# Example Usage:
# Single GPU: torchrun --nproc_per_node=1 trainer/diffusers_trainer.py --model="CompVis/stable-diffusion-v1-4" --run_name="liminal" --dataset="liminal-dataset" --hf_token="hf_blablabla" --bucket_side_min=64 --use_8bit_adam=True --gradient_checkpointing=True --batch_size=1 --fp16=True --image_log_steps=250 --epochs=20 --resolution=768 --use_ema=True
# Multiple GPUs: torchrun --nproc_per_node=N trainer/diffusers_trainer.py --model="CompVis/stable-diffusion-v1-4" --run_name="liminal" --dataset="liminal-dataset" --hf_token="hf_blablabla" --bucket_side_min=64 --use_8bit_adam=True --gradient_checkpointing=True --batch_size=10 --fp16=True --image_log_steps=250 --epochs=20 --resolution=768 --use_ema=True

import argparse
import socket
import sys

chavinlo

November 26th Hivemind Test Run

Instructions:

For Windows Users:

Hivemind only works on Linux. Thankfully you can still use WSL to run the training.

Follow this guide: https://learn.microsoft.com/es-es/windows/wsl/install

Basically, you should open a CMD window (the black window with white words) and type:

chavinlo

torchrun --nproc_per_node=1 \
	train.py \
	--workingdirectory hivemindtemp \
	--wantedimages 500 \
	--datasetserver="DATASET_SERVER_IP" \
	--node="true" \
	--o_port1=LOCAL_TCP_PORT \
	--o_port2=LOCAL_UDP_PORT \
	--ip_is_different="true" \
	--p_ip="PUBLIC_IP" \

chavinlo

# Install bitsandbytes:
# `nvcc --version` to get CUDA version.
# `pip install -i https://test.pypi.org/simple/ bitsandbytes-cudaXXX` to install for current CUDA.
# Example Usage:
# Single GPU: torchrun --nproc_per_node=1 trainer/diffusers_trainer.py --model="CompVis/stable-diffusion-v1-4" --run_name="liminal" --dataset="liminal-dataset" --hf_token="hf_blablabla" --bucket_side_min=64 --use_8bit_adam=True --gradient_checkpointing=True --batch_size=1 --fp16=True --image_log_steps=250 --epochs=20 --resolution=768 --use_ema=True
# Multiple GPUs: torchrun --nproc_per_node=N trainer/diffusers_trainer.py --model="CompVis/stable-diffusion-v1-4" --run_name="liminal" --dataset="liminal-dataset" --hf_token="hf_blablabla" --bucket_side_min=64 --use_8bit_adam=True --gradient_checkpointing=True --batch_size=10 --fp16=True --image_log_steps=250 --epochs=20 --resolution=768 --use_ema=True

import argparse
import socket
import torch

chavinlo

#!/bin/bash

#Install deps
apt-get update -y
apt-get install htop screen psmisc python3-pip unzip wget gcc g++ nano -y

#Install Python deps
wget https://gist.githubusercontent.com/chavinlo/fe8afc02e03d9cc4eb545c4c306c8a73/raw/d9a5ad446fe662dc3e6597163a1f8d5546a8a795/requirements.txt
pip install -r requirements.txt OmegaConf 
pip install triton==2.0.0.dev20221120

chavinlo

# Install bitsandbytes:
# `nvcc --version` to get CUDA version.
# `pip install -i https://test.pypi.org/simple/ bitsandbytes-cudaXXX` to install for current CUDA.
# Example Usage:
# Single GPU: torchrun --nproc_per_node=1 trainer/diffusers_trainer.py --model="CompVis/stable-diffusion-v1-4" --run_name="liminal" --dataset="liminal-dataset" --hf_token="hf_blablabla" --bucket_side_min=64 --use_8bit_adam=True --gradient_checkpointing=True --batch_size=1 --fp16=True --image_log_steps=250 --epochs=20 --resolution=768 --use_ema=True
# Multiple GPUs: torchrun --nproc_per_node=N trainer/diffusers_trainer.py --model="CompVis/stable-diffusion-v1-4" --run_name="liminal" --dataset="liminal-dataset" --hf_token="hf_blablabla" --bucket_side_min=64 --use_8bit_adam=True --gradient_checkpointing=True --batch_size=10 --fp16=True --image_log_steps=250 --epochs=20 --resolution=768 --use_ema=True

import argparse
import socket
import torch

chavinlo

#!/bin/bash

#Install deps
apt-get update -y
apt-get install htop screen psmisc python3-pip unzip wget gcc g++ nano -y

#Install Python deps
wget https://gist.githubusercontent.com/chavinlo/fe8afc02e03d9cc4eb545c4c306c8a73/raw/d9a5ad446fe662dc3e6597163a1f8d5546a8a795/requirements.txt
pip install -r requirements.txt OmegaConf 
pip install triton==2.0.0.dev20221120