Add files via upload

This commit is contained in:
Madiator2011 2024-07-14 16:20:13 +02:00 committed by GitHub
parent 05c55ecec2
commit 7fdd7605ce
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
4 changed files with 554 additions and 0 deletions

View file

@ -0,0 +1,51 @@
FROM madiator2011/better-pytorch:cuda12.1
ARG DEBIAN_FRONTEND=noninteractive
ARG TORCH
ARG TORCHVISION
ARG TORCHAUDIO
ARG PYTHON_VERSION
# Use bash shell
SHELL ["/bin/bash", "-o", "pipefail", "-c"]
# Set environment variables
ENV DEBIAN_FRONTEND=noninteractive
ENV SHELL=/bin/bash
ENV PYTHONUNBUFFERED=1
ENV LD_LIBRARY_PATH="/usr/local/cuda-12.1/targets/x86_64-linux/lib/"
# Create workspace directory
WORKDIR /
# Install any additional packages not included in the base image
RUN apt-get update --yes && \
apt-get upgrade --yes && \
apt-get autoremove -y && \
apt-get clean && \
rm -rf /var/lib/apt/lists/* && \
echo "en_US.UTF-8 UTF-8" > /etc/locale.gen
# Clone EveryDream2trainer repository
WORKDIR /root
RUN git clone https://github.com/victorchall/EveryDream2trainer
# Add requirements files and install Python packages from the cloned repository
RUN pip install --upgrade --no-cache-dir pip && \
pip install ${TORCH} torchvision torchaudio && \
pip install --no-deps xformers==0.0.22.post7 && \
pip install --no-cache-dir -r /root/EveryDream2trainer/docker/requirements-build.txt && \
pip install --no-cache-dir -r /root/EveryDream2trainer/docker/requirements-runtime.txt
# Copy pre_start.sh script to the image
COPY pre_start.sh /pre_start.sh
RUN chmod +x /pre_start.sh
# Copy notebook
COPY Train_RunPod.ipynb /root/EveryDream2trainer/Train_RunPod.ipynb
# Start Scripts
COPY --from=scripts start.sh /
RUN chmod +x /start.sh
# Set the default command for the container
CMD [ "/start.sh" ]

View file

@ -0,0 +1,454 @@
{
"cells": [
{
"attachments": {},
"cell_type": "markdown",
"id": "2c831b5b-3025-4177-bef5-25aaec89573a",
"metadata": {},
"source": [
"## Every Dream v2 RunPod Notebook\n",
"\n",
"### [General Instructions](https://github.com/victorchall/EveryDream2trainer/blob/main/README.md)\n",
"\n",
"### What's your plan?\n",
"You will want to have your data prepared before starting, and have a rough training plan in mind. \n",
"\n",
"**Make sure your images are captioned!**\n",
"\n",
"By default the name of your image files are assumed to be captions. If you want to get fancy, there are [more sophisticated techniques](https://github.com/victorchall/EveryDream2trainer/blob/main/doc/DATA.md)\n",
"\n",
"**If this is your first time trying a full fine-tune, start small!** \n",
"\n",
"Pick a single concept and 30-100 images, and see what happens. \n",
"\n",
"Training a small dataset like this is fast, and will give you a feel for how quickly your model (over-)trains depending on your training schedule, captioning schema, knob twiddling. This notebook provides some sensible defaults, there are more questions than answers in how best to fine tune anything. \n",
"\n",
"**_When_ you have questions...**\n",
"\n",
"Come visit us at [EveryDream Discord](https://discord.gg/uheqxU6sXN)"
]
},
{
"attachments": {},
"cell_type": "markdown",
"id": "ffff47f7",
"metadata": {},
"source": [
"## Optional Speed Test\n",
"If all goes well you may find yourself downloading (or pushing to the cloud) 2-8GB of model data per saved checkpoint. Make sure your pod is not a dud. ~1000Mbit/s up/dn is probably good, though the location of the pod also makes a difference.\n"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "934ba107",
"metadata": {},
"outputs": [],
"source": [
"import speedtest\n",
"st = speedtest.Speedtest()\n",
"print(f\"Your download speed: {round(st.download() / 1000 / 1000, 1)} Mbit/s\")\n",
"print(f\"Your upload speed: {round(st.upload() / 1000 / 1000, 1)} Mbit/s\")"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "3034c47a",
"metadata": {},
"outputs": [],
"source": [
"# Check your GPU\n",
"!nvidia-smi"
]
},
{
"cell_type": "markdown",
"id": "79c17082",
"metadata": {},
"source": [
"## Run this first to go to Workspace folder"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "329f522d",
"metadata": {},
"outputs": [],
"source": [
"%cd /workspace"
]
},
{
"attachments": {},
"cell_type": "markdown",
"id": "7c73894e-3b5e-4268-9f83-ed89bd4569f2",
"metadata": {
"tags": []
},
"source": [
"# Upload training files\n",
"\n",
"Use the navigation on the left to open the **/workspace/EveryDream2trainer/input** folder and upload your training files using the **up arrow button** above the file explorer, or by dragging and dropping the files from your local machine onto the file explorer.\n",
"\n",
"If you have many training files, or nested folders of training data, create a zip archive of your training data, upload this file to the input folder, then unzip via terminal.\n",
"\n",
"### (Optional) Configure sample prompts\n",
"You can set your own sample prompts by adding them, one line at a time, to sample_prompts.txt.### (Optional) Weights and Biases login. "
]
},
{
"attachments": {},
"cell_type": "markdown",
"id": "f15fcd56-0418-4be1-a5c3-38aa679b1aaf",
"metadata": {},
"source": [
"# Start Training\n",
"Naming your project will help you track what the heck you're doing when you're floating in checkpoint files later.\n",
"\n",
"You may wish to consider adding \"sd1\" or \"sd2v\" or similar to remember what the base was, as you'll also have to tell your inference app what you were using, as its difficult for programs to know what inference YAML to use automatically. For instance, Automatic1111 webui requires you to copy the v2 inference YAML and rename it to match your checkpoint name so it knows how to load the file, tough it assumes SD 1.x compatible. Something to keep in mind if you start training on SD2.1.\n",
"\n",
"`max_epochs`, `sample_steps`, and `save_every_n_epochs` should be tuned to your dataset. I like to generate one or two sets of samples per save, and aim for 5 (give or take 2) saved checkpoints.\n",
"\n",
"Next cell runs training. This will take a while depending on your number of images, repeats, and max_epochs.\n",
"\n",
"You can watch for test images in the logs folder.\n",
"\n",
"#### Weights and Balanaces\n",
"I you pass the `--wandb` flag you will be prompted for your W&B `API Key`. W&B is a free online logging utility. If you don't have a W&B account, you can create one for free at https://wandb.ai/site. Your key is on this page: https://wandb.ai/settings under \"Danger Zone\" \"API Keys\""
]
},
{
"cell_type": "markdown",
"id": "95cac0bd",
"metadata": {},
"source": [
"### Note this section might look wierd but training works fine"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "6f73fb86-ebef-41e2-9382-4aa11be84be6",
"metadata": {
"scrolled": true,
"tags": []
},
"outputs": [],
"source": [
"%run train.py --config train.json \\\n",
"--resume_ckpt \"panopstor/EveryDream\" \\\n",
"--project_name \"sd1_mymodel\" \\\n",
"--data_root \"input\" \\\n",
"--max_epochs 200 \\\n",
"--sample_steps 150 \\\n",
"--save_every_n_epochs 35 \\\n",
"--lr 1.2e-6 \\\n",
"--lr_scheduler constant \\\n",
"--save_full_precision\n"
]
},
{
"attachments": {},
"cell_type": "markdown",
"id": "ed464c6b-1a8d-48e4-9787-265e8acaac43",
"metadata": {},
"source": [
"### Optionally you can chain trainings together using multiple configurations combined with `resume_ckpt: findlast`"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "492350d4-9b2f-4d2a-9641-1f723125b296",
"metadata": {},
"outputs": [],
"source": [
"%run train.py --config chain0.json --project_name \"sd1_chain_a\" --data_root \"input\" --resume_ckpt \"panopstor/EveryDream\"\n",
"%run train.py --config chain1.json --project_name \"sd1_chain_b\" --data_root \"input\" --resume_ckpt findlast\n",
"%run train.py --config chain2.json --project_name \"sd1_chain_c\" --data_root \"input\" --resume_ckpt findlast"
]
},
{
"attachments": {},
"cell_type": "markdown",
"id": "3c506e79-bf03-4e34-bf06-9371963d4d7d",
"metadata": {},
"source": [
"# HuggingFace Upload/Download (Optional)\n",
"Run the cell below and paste your token into the prompt. You can get your token from your [huggingface account page](https://huggingface.co/settings/tokens).\n",
"\n",
"The token will not show on the screen, just press enter after you paste it.\n",
"\n",
"### Make sure to run this login cell for any Huggingface uploads or private downloads."
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "138b7776-8783-4e1d-920d-cf358809b802",
"metadata": {
"tags": [
"hflogin"
]
},
"outputs": [],
"source": [
"from huggingface_hub import notebook_login, hf_hub_download\n",
"import os\n",
"notebook_login()"
]
},
{
"attachments": {},
"cell_type": "markdown",
"id": "b252a308-49cf-443f-abbb-d08b471411fb",
"metadata": {},
"source": [
"### HuggingFace Download a checkpoint\n",
"Make sure you are logged in using the above login cell first. "
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "86b66fe4-c2ca-46fa-813c-8fe390813add",
"metadata": {
"scrolled": true,
"tags": []
},
"outputs": [],
"source": [
"repo=\"panopstor/EveryDream\"\n",
"ckpt_file=\"sd_v1-5_vae.ckpt\"\n",
"\n",
"print(f\"Downloading {ckpt_file} from {repo}\")\n",
"downloaded_model_path = hf_hub_download(repo, ckpt_file, cache_dir=\"/workspace/hfcache\")\n",
"ckpt_name = os.path.splitext(os.path.basename(downloaded_model_path))[0]\n",
"print(f\"Downloaded {ckpt_name} to {downloaded_model_path}\")\n",
"\n",
"if not os.path.exists(f\"ckpt_cache/{ckpt_name}\"):\n",
" print(f\"Converting {ckpt_name} to Diffusers format\")\n",
" %run utils/convert_original_stable_diffusion_to_diffusers.py --scheduler_type ddim \\\n",
" --original_config_file v1-inference.yaml \\\n",
" --image_size 512 \\\n",
" --checkpoint_path \"{downloaded_model_path}\" \\\n",
" --prediction_type epsilon \\\n",
" --upcast_attn False \\\n",
" --dump_path \"ckpt_cache/{ckpt_name}\"\n",
"\n",
"print(\"DONE\")"
]
},
{
"attachments": {},
"cell_type": "markdown",
"id": "f24eee3d-f5df-45f3-9acc-ee0206cfe6b1",
"metadata": {},
"source": [
"### HuggingFace upload all your checkpoints\n",
"Make sure you are **logged in** using the above login cell first. \n",
"\n",
"Use the cell below to upload one or more checkpoints to your personal HuggingFace account, if you want, instead of manually downloading. You should already be authorized to Huggingface by token if you used the download/token cells above.\n",
"\n",
"* You can get your account name from your [HuggingFace account page](https://huggingface.co/settings/account). Look for your \"username\" field and paste it below.\n",
"\n",
"* You only need to setup a repository one time. You can create it here: [Create New HF Model](https://huggingface.co/new) Make sure you write down the repo name you make for future use. You can reuse it later.\n",
"\n",
"When you run this, **CLICK** to select which .ckpt files are marked for upload. This allows you to select which ones to upload. If you don't click of the ckpts, nothing will happen."
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "b9df5e1a-3c68-41c0-a4ed-ea0abcd19858",
"metadata": {},
"outputs": [],
"source": [
"import glob\n",
"import os\n",
"from huggingface_hub import HfApi\n",
"from ipywidgets import *\n",
"\n",
"all_ckpts = [f for f in glob.glob(\"*.safetensors\")]\n",
" \n",
"ckpt_picker = SelectMultiple(options=all_ckpts, layout=Layout(width=\"600px\")) \n",
"hfuser = Text(placeholder='Your HF user name')\n",
"hfrepo = Text(placeholder='Your HF repo name')\n",
"\n",
"api = HfApi()\n",
"upload_btn = Button(description='Upload')\n",
"out = Output()\n",
"\n",
"def upload_ckpts(_):\n",
" repo_id=f\"{hfuser.value or hfuser.placeholder}/{hfrepo.value or hfrepo.placeholder}\"\n",
" with out:\n",
" if ckpt_picker is None or len(ckpt_picker.value) < 1:\n",
" print(\"Nothing selected for upload, make sure to click one of the ckpt files in the list, or, you have no ckpt files in the current directory.\")\n",
" for ckpt in ckpt_picker.value:\n",
" print(f\"Uploading to HF: huggingface.co/{repo_id}/{ckpt}\")\n",
" response = api.upload_file(\n",
" path_or_fileobj=ckpt,\n",
" path_in_repo=ckpt,\n",
" repo_id=repo_id,\n",
" repo_type=None,\n",
" create_pr=1,\n",
" )\n",
" display(response)\n",
" print(\"DONE\")\n",
" print(\"Go to your repo and accept the PRs this created to see your files\")\n",
"\n",
"upload_btn.on_click(upload_ckpts)\n",
"box = VBox([ckpt_picker, HBox([hfuser, hfrepo]), upload_btn, out])\n",
"\n",
"display(box)"
]
},
{
"attachments": {},
"cell_type": "markdown",
"id": "c1a00d16-9b84-492f-8e6a-defe71e82b43",
"metadata": {},
"source": [
"# Test inference on your checkpoints"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "efb1a8cd-6a04-44e5-a770-c23ee247ce82",
"metadata": {},
"outputs": [],
"source": [
"from ipywidgets import *\n",
"from IPython.display import display, clear_output\n",
"import os\n",
"import gc\n",
"import random\n",
"import torch\n",
"import inspect\n",
"\n",
"from torch import autocast\n",
"from diffusers import StableDiffusionPipeline, AutoencoderKL, UNet2DConditionModel, DDIMScheduler, DDPMScheduler, PNDMScheduler, EulerAncestralDiscreteScheduler\n",
"from transformers import CLIPTextModel, CLIPTokenizer\n",
"\n",
"\n",
"checkpoints_ts = []\n",
"for root, dirs, files in os.walk(\".\"):\n",
" for file in files:\n",
" if os.path.basename(file) == \"model_index.json\":\n",
" ts = os.path.getmtime(os.path.join(root,file))\n",
" ckpt = root\n",
" checkpoints_ts.append((ts, root))\n",
"\n",
"checkpoints = [ckpt for (_, ckpt) in sorted(checkpoints_ts, reverse=True)]\n",
"full_width = Layout(width='600px')\n",
"half_width = Layout(width='300px')\n",
"\n",
"checkpoint = Dropdown(options=checkpoints, description='Checkpoint:', layout=full_width)\n",
"prompt = Textarea(value='a photo of ', description='Prompt:', layout=full_width)\n",
"height = IntSlider(value=512, min=256, max=768, step=32, description='Height:', layout=half_width)\n",
"width = IntSlider(value=512, min=256, max=768, step=32, description='Width:', layout=half_width)\n",
"cfg = FloatSlider(value=7.0, min=0.0, max=14.0, step=0.2, description='CFG Scale:', layout=half_width)\n",
"steps = IntSlider(value=30, min=10, max=100, description='Steps:', layout=half_width)\n",
"seed = IntText(value=-1, description='Seed:', layout=half_width)\n",
"generate_btn = Button(description='Generate', layout=full_width)\n",
"out = Output()\n",
"\n",
"def generate(_):\n",
" with out:\n",
" clear_output()\n",
" display(f\"Loading model {checkpoint.value}\")\n",
" actual_seed = seed.value if seed.value != -1 else random.randint(0, 2**30)\n",
"\n",
" text_encoder = CLIPTextModel.from_pretrained(checkpoint.value, subfolder=\"text_encoder\")\n",
" vae = AutoencoderKL.from_pretrained(checkpoint.value, subfolder=\"vae\")\n",
" unet = UNet2DConditionModel.from_pretrained(checkpoint.value, subfolder=\"unet\")\n",
" tokenizer = CLIPTokenizer.from_pretrained(checkpoint.value, subfolder=\"tokenizer\", use_fast=False)\n",
" scheduler = DDIMScheduler.from_pretrained(checkpoint.value, subfolder=\"scheduler\")\n",
" text_encoder.eval()\n",
" vae.eval()\n",
" unet.eval()\n",
"\n",
" text_encoder.to(\"cuda\")\n",
" vae.to(\"cuda\")\n",
" unet.to(\"cuda\")\n",
"\n",
" pipe = StableDiffusionPipeline(\n",
" vae=vae,\n",
" text_encoder=text_encoder,\n",
" tokenizer=tokenizer,\n",
" unet=unet,\n",
" scheduler=scheduler,\n",
" safety_checker=None, # save vram\n",
" requires_safety_checker=None, # avoid nag\n",
" feature_extractor=None, # must be none of no safety checker\n",
" )\n",
"\n",
" pipe.scheduler = EulerAncestralDiscreteScheduler.from_config(pipe.scheduler.config)\n",
" \n",
" print(inspect.cleandoc(f\"\"\"\n",
" Prompt: {prompt.value}\n",
" Resolution: {width.value}x{height.value}\n",
" CFG: {cfg.value}\n",
" Steps: {steps.value}\n",
" Seed: {actual_seed}\n",
" \"\"\"))\n",
" with autocast(\"cuda\"):\n",
" image = pipe(prompt.value, \n",
" generator=torch.Generator(\"cuda\").manual_seed(actual_seed),\n",
" num_inference_steps=steps.value, \n",
" guidance_scale=cfg.value,\n",
" width=width.value,\n",
" height=height.value\n",
" ).images[0]\n",
" del pipe\n",
" gc.collect()\n",
" with torch.cuda.device(\"cuda\"):\n",
" torch.cuda.empty_cache()\n",
" torch.cuda.ipc_collect()\n",
" display(image)\n",
" \n",
"generate_btn.on_click(generate)\n",
"box = VBox(\n",
" children=[\n",
" checkpoint, prompt, \n",
" HBox([VBox([width, height]), VBox([steps, cfg])]), \n",
" seed, \n",
" generate_btn, \n",
" out]\n",
")\n",
"\n",
"\n",
"display(box)"
]
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3 (ipykernel)",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.10.12"
},
"vscode": {
"interpreter": {
"hash": "2e677f113ff5b533036843965d6e18980b635d0aedc1c5cebd058006c5afc92a"
}
}
},
"nbformat": 4,
"nbformat_minor": 5
}

View file

@ -0,0 +1,18 @@
group "default" {
targets = ["py310-cuda121"]
}
target "py310-cuda121" {
contexts = {
scripts = "../../container-template"
proxy = "../../container-template/proxy"
logo = "../../container-template"
}
dockerfile = "Dockerfile"
args = {
BASE_IMAGE = "madiator2011/better-pytorch:cuda12.1",
TORCH = "torch==2.1.0+cu121 -f https://download.pytorch.org/whl/torch_stable.html",
PYTHON_VERSION = "3.10"
}
tags = ["madiator2011/better-everydream2:cuda12.1"]
}

View file

@ -0,0 +1,31 @@
#!/bin/bash
# Exit immediately if a command exits with a non-zero status
set -e
# Function to print colorized feedback
print_feedback() {
GREEN='\033[0;32m'
NC='\033[0m' # No Color
echo -e "${GREEN}[EveryDream2trainer Startup]:${NC} $1"
}
# Function to run rsync with progress bar
rsync_with_progress() {
stdbuf -i0 -o0 -e0 rsync -au --info=progress2 "$@" | stdbuf -i0 -o0 -e0 tr '\r' '\n' | stdbuf -i0 -o0 -e0 grep -oP '\d+%|\d+.\d+[mMgG]' | tqdm --bar-format='{l_bar}{bar}' --total=100 --unit='%' > /dev/null
}
# Check for required commands
for cmd in stdbuf rsync tr grep tqdm; do
if ! command -v $cmd &> /dev/null; then
echo "$cmd could not be found, please install it."
exit 1
fi
done
print_feedback "Starting EveryDream2trainer setup..."
print_feedback "Syncing EveryDream2trainer files..."
rsync_with_progress --remove-source-files /root/EveryDream2trainer/ /workspace/
print_feedback "Setup complete."