ostris/ai-toolkit
1
0
Fork 0
mirror of https://github.com/ostris/ai-toolkit.git synced 2026-01-26 16:39:47 +00:00
Code Issues Packages Projects Releases Wiki Activity

Separate dependencies for DGX OS devices (#610)

Browse Source
This commit is contained in:
raziel2001au
2025-12-27 01:32:26 +10:00
committed by GitHub
parent a767b82b60
commit 90fc99f486
2 changed files with 50 additions and 13 deletions
Download Patch File Download Diff File Expand all files Collapse all files

22
dgx_instructions.md
View File

@@ -12,39 +12,39 @@ This guide will assume you have a fresh installation of DGX OS, and will guide y
Install the latest version of miniconda: Install the latest version of miniconda:
``` ```
$ wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-aarch64.sh wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-aarch64.sh
$ chmod u+x Miniconda3-latest-Linux-aarch64.sh chmod u+x Miniconda3-latest-Linux-aarch64.sh
$ ./Miniconda3-latest-Linux-aarch64.sh ./Miniconda3-latest-Linux-aarch64.sh
``` ```
Restart your bash or ssh session. If miniconda was installed successfully, it will automatically load the 'base' environment by default. If you want to disable this behaviour, run: Restart your bash or ssh session. If miniconda was installed successfully, it will automatically load the 'base' environment by default. If you want to disable this behaviour, run:
``` ```
$ conda config --set auto_activate_base false conda config --set auto_activate_base false
``` ```
Now you can create a Python 3.11 environment for ai-toolkit: Now you can create a Python 3.11 environment for ai-toolkit:
``` ```
$ conda create --name ai-toolkit python=3.11 conda create --name ai-toolkit python=3.11
``` ```
Then activate the environment with: Then activate the environment with:
``` ```
$ conda activate ai-toolkit conda activate ai-toolkit
``` ```
**2) Install PyTorch** **2) Install PyTorch**
``` ```
$ pip3 install torch==2.9.1 torchvision==0.24.1 torchaudio==2.9.1 --index-url https://download.pytorch.org/whl/cu130 pip3 install torch==2.9.1 torchvision==0.24.1 torchaudio==2.9.1 --index-url https://download.pytorch.org/whl/cu130
``` ```
**3) Install the remaining requirements (dgx_requirements.txt)** **3) Install the remaining requirements (dgx_requirements.txt)**
``` ```
$ pip3 install -r dgx_requirements.txt pip3 install -r dgx_requirements.txt
``` ```
### Running the UI on DGX OS: ### Running the UI on DGX OS:
@@ -66,8 +66,8 @@ export PATH=“/opt/node-v24.11.1-linux-arm64/bin:$PATH”
Change to the ui directory, then build and run the UI: Change to the ui directory, then build and run the UI:
``` ```
$ cd ui cd ui
$ npm run build_and_start npm run build_and_start
``` ```
If all went well, you’ll be able to access the UI on port 8675 and start training. If all went well, you’ll be able to access the UI on port 8675 and start training.
@@ -78,7 +78,7 @@ If all went well, you’ll be able to access the UI on port 8675 and start train
If you’re not getting any output when starting a training job from the UI, it’s probably crashing before the process started, the best way to debug these issues is to run the python training script directly (which is normally started by the UI). To do this, set up a training job in the UI, go to the advanced config screen, copy and paste the configuration into a file like train.yaml, then run the training script like this with the conda virtual environment active: If you’re not getting any output when starting a training job from the UI, it’s probably crashing before the process started, the best way to debug these issues is to run the python training script directly (which is normally started by the UI). To do this, set up a training job in the UI, go to the advanced config screen, copy and paste the configuration into a file like train.yaml, then run the training script like this with the conda virtual environment active:
``` ```
$ python run.py path/to/train.yaml python run.py path/to/train.yaml
``` ```
</details> </details>
<br> <br>

41
dgx_requirements.txt
View File

@@ -10,5 +10,42 @@ pywavelets==1.9.0
contourpy==1.3.3 contourpy==1.3.3
opencv_python_headless==4.11.0.86 opencv_python_headless==4.11.0.86
# we include the base requirements.txt for the remaining dependencies: # additional dependencies matching requirements.txt:
-r requirements.txt torchao==0.10.0
safetensors
git+https://github.com/huggingface/diffusers@6bf668c4d217ebc96065e673d8a257fd79950d34
transformers==4.57.3
lycoris-lora==1.8.3
flatten_json
pyyaml
oyaml
tensorboard
kornia
invisible-watermark
einops
accelerate
toml
albumentations==1.4.15
albucore==0.0.16
pydantic
omegaconf
k-diffusion
open_clip_torch
timm
prodigyopt
controlnet_aux==0.0.10
python-dotenv
bitsandbytes
hf_transfer
lpips
pytorch_fid
optimum-quanto==0.2.4
sentencepiece
huggingface_hub
peft
gradio
python-slugify
opencv-python
pytorch-wavelets==1.3.0
matplotlib==3.10.1
setuptools==69.5.1