From eee8df53b51e1e9bdc2e910b7c9faef9d65b06d4 Mon Sep 17 00:00:00 2001 From: Patrick Esser Date: Fri, 22 Jul 2022 09:56:22 +0000 Subject: [PATCH] test gpu scripts and other launchers and configs --- configs/stable-diffusion/v1_laionhr.yaml | 135 ++++++++++++++++++ .../launcher.sh | 3 +- scripts/slurm/v1_laionhr_torch111/launcher.sh | 36 +++++ scripts/slurm/v1_laionhr_torch111/sbatch.sh | 42 ++++++ scripts/test_gpu.py | 30 ++++ scripts/test_gpu.sh | 5 + 6 files changed, 250 insertions(+), 1 deletion(-) create mode 100644 configs/stable-diffusion/v1_laionhr.yaml create mode 100755 scripts/slurm/v1_laionhr_torch111/launcher.sh create mode 100755 scripts/slurm/v1_laionhr_torch111/sbatch.sh create mode 100644 scripts/test_gpu.py create mode 100644 scripts/test_gpu.sh diff --git a/configs/stable-diffusion/v1_laionhr.yaml b/configs/stable-diffusion/v1_laionhr.yaml new file mode 100644 index 0000000..15f7739 --- /dev/null +++ b/configs/stable-diffusion/v1_laionhr.yaml @@ -0,0 +1,135 @@ +model: + base_learning_rate: 1.0e-04 + target: ldm.models.diffusion.ddpm.LatentDiffusion + params: + linear_start: 0.00085 + linear_end: 0.0120 + num_timesteps_cond: 1 + log_every_t: 200 + timesteps: 1000 + first_stage_key: "jpg" + cond_stage_key: "txt" + image_size: 64 + channels: 4 + cond_stage_trainable: false # Note: different from the one we trained before + conditioning_key: crossattn + monitor: val/loss_simple_ema + scale_factor: 0.18215 + + scheduler_config: # 10000 warmup steps + target: ldm.lr_scheduler.LambdaLinearScheduler + params: + warm_up_steps: [ 1 ] # NOTE for resuming. use 10000 if starting from scratch + cycle_lengths: [ 10000000000000 ] # incredibly large number to prevent corner cases + f_start: [ 1.e-6 ] + f_max: [ 1. ] + f_min: [ 1. ] + + unet_config: + target: ldm.modules.diffusionmodules.openaimodel.UNetModel + params: + image_size: 32 # unused + in_channels: 4 + out_channels: 4 + model_channels: 320 + attention_resolutions: [ 4, 2, 1 ] + num_res_blocks: 2 + channel_mult: [ 1, 2, 4, 4 ] + num_heads: 8 + use_spatial_transformer: True + transformer_depth: 1 + context_dim: 768 + use_checkpoint: True + legacy: False + + first_stage_config: + target: ldm.models.autoencoder.AutoencoderKL + params: + embed_dim: 4 + monitor: val/rec_loss + ddconfig: + double_z: true + z_channels: 4 + resolution: 256 + in_channels: 3 + out_ch: 3 + ch: 128 + ch_mult: + - 1 + - 2 + - 4 + - 4 + num_res_blocks: 2 + attn_resolutions: [] + dropout: 0.0 + lossconfig: + target: torch.nn.Identity + + cond_stage_config: + target: ldm.modules.encoders.modules.FrozenCLIPEmbedder + + +data: + target: ldm.data.laion.WebDataModuleFromConfig + params: + tar_base: "pipe:aws s3 cp s3://s-datasets/laion-high-resolution/" + batch_size: 4 + num_workers: 4 + multinode: True + train: + shards: '{00000..17279}.tar -' + shuffle: 10000 + image_key: jpg + image_transforms: + - target: torchvision.transforms.Resize + params: + size: 512 + interpolation: 3 + - target: torchvision.transforms.RandomCrop + params: + size: 512 + + # NOTE use enough shards to avoid empty validation loops in workers + validation: + shards: '{17280..17535}.tar -' + shuffle: 0 + image_key: jpg + image_transforms: + - target: torchvision.transforms.Resize + params: + size: 512 + interpolation: 3 + - target: torchvision.transforms.CenterCrop + params: + size: 512 + + +lightning: + find_unused_parameters: False + + modelcheckpoint: + params: + every_n_train_steps: 5000 + + callbacks: + image_logger: + target: main.ImageLogger + params: + batch_frequency: 5000 + max_images: 4 + increase_log_steps: False + log_first_step: False + log_images_kwargs: + use_ema_scope: False + inpaint: False + plot_progressive_rows: False + plot_diffusion_rows: False + N: 4 + unconditional_guidance_scale: 3.0 + unconditional_guidance_label: [""] + + trainer: + benchmark: True + val_check_interval: 5000000 # really sorry + num_sanity_val_steps: 0 + accumulate_grad_batches: 2 diff --git a/scripts/slurm/v1_improvedaesthetics_torch111/launcher.sh b/scripts/slurm/v1_improvedaesthetics_torch111/launcher.sh index 4cd8a81..51c339e 100755 --- a/scripts/slurm/v1_improvedaesthetics_torch111/launcher.sh +++ b/scripts/slurm/v1_improvedaesthetics_torch111/launcher.sh @@ -24,7 +24,8 @@ CONFIG="/fsx/stable-diffusion/stable-diffusion/configs/stable-diffusion/v1_impro # resume and set new seed to reshuffle data #EXTRA="--seed 714 model.params.ckpt_path=/fsx/stable-diffusion/stable-diffusion/logs/2022-07-11T20-16-11_txt2img-1p4B-multinode-clip-encoder-high-res-512_improvedaesthetic/checkpoints/last.ckpt" #EXTRA="--seed 715 model.params.ckpt_path=/fsx/stable-diffusion/stable-diffusion/logs/2022-07-14T23-26-13_v1_improvedaesthetics/checkpoints/last.ckpt" -EXTRA="--seed 716 --resume_from_checkpoint /fsx/stable-diffusion/stable-diffusion/logs/2022-07-18T23-11-11_v1_improvedaesthetics/checkpoints/last.ckpt" +#EXTRA="--seed 716 --resume_from_checkpoint /fsx/stable-diffusion/stable-diffusion/logs/2022-07-18T23-11-11_v1_improvedaesthetics/checkpoints/last.ckpt" +EXTRA="--seed 717 --resume_from_checkpoint /fsx/stable-diffusion/stable-diffusion/logs/2022-07-19T19-03-04_v1_improvedaesthetics/checkpoints/last.ckpt" # time to decay EXTRA="${EXTRA} model.params.scheduler_config.params.cycle_lengths=[50000] model.params.scheduler_config.params.f_min=[1e-6]" diff --git a/scripts/slurm/v1_laionhr_torch111/launcher.sh b/scripts/slurm/v1_laionhr_torch111/launcher.sh new file mode 100755 index 0000000..48d45e2 --- /dev/null +++ b/scripts/slurm/v1_laionhr_torch111/launcher.sh @@ -0,0 +1,36 @@ +#!/bin/bash + +# mpi version for node rank +H=`hostname` +THEID=`echo -e $HOSTNAMES | python3 -c "import sys;[sys.stdout.write(str(i)) for i,line in enumerate(next(sys.stdin).split(' ')) if line.strip() == '$H'.strip()]"` +export NODE_RANK=${THEID} +echo THEID=$THEID + +echo "##########################################" +echo MASTER_ADDR=${MASTER_ADDR} +echo MASTER_PORT=${MASTER_PORT} +echo NODE_RANK=${NODE_RANK} +echo WORLD_SIZE=${WORLD_SIZE} +echo "##########################################" +# debug environment worked great so we stick with it +# no magic there, just a miniconda python=3.9, pytorch=1.12, cudatoolkit=11.3 +# env with pip dependencies from stable diffusion's requirements.txt +eval "$(/fsx/stable-diffusion/debug/miniconda3/bin/conda shell.bash hook)" +conda activate torch111 +cd /fsx/stable-diffusion/stable-diffusion + +CONFIG="/fsx/stable-diffusion/stable-diffusion/configs/stable-diffusion/v1_laionhr.yaml" + +# resume and set new seed to reshuffle data +EXTRA="--seed 718 model.params.ckpt_path=/fsx/stable-diffusion/stable-diffusion/checkpoints2/v1pp/v1pp-flatline.ckpt" + +# time to decay +#EXTRA="${EXTRA} model.params.scheduler_config.params.cycle_lengths=[50000] model.params.scheduler_config.params.f_min=[1e-6]" + +# custom logdir +#EXTRA="${EXTRA} --logdir rlogs" + +# debugging +#EXTRA="${EXTRA} -d True lightning.callbacks.image_logger.params.batch_frequency=50" + +python main.py --base $CONFIG --gpus 0,1,2,3,4,5,6,7 -t --num_nodes ${WORLD_SIZE} --scale_lr False $EXTRA diff --git a/scripts/slurm/v1_laionhr_torch111/sbatch.sh b/scripts/slurm/v1_laionhr_torch111/sbatch.sh new file mode 100755 index 0000000..9eef7ce --- /dev/null +++ b/scripts/slurm/v1_laionhr_torch111/sbatch.sh @@ -0,0 +1,42 @@ +#!/bin/bash +#SBATCH --partition=compute-od-gpu +#SBATCH --job-name=stable-diffusion-v1-laionhr-torch111 +#SBATCH --nodes 20 +#SBATCH --ntasks-per-node 1 +#SBATCH --cpus-per-gpu=4 +#SBATCH --gres=gpu:8 +#SBATCH --exclusive +#SBATCH --output=%x_%j.out +#SBATCH --comment "Key=Monitoring,Value=ON" + +module load intelmpi +source /opt/intel/mpi/latest/env/vars.sh +export LD_LIBRARY_PATH=/opt/aws-ofi-nccl/lib:/opt/amazon/efa/lib64:/usr/local/cuda-11.0/efa/lib:/usr/local/cuda-11.0/lib:/usr/local/cuda-11.0/lib64:/usr/local/cuda-11.0:/opt/nccl/build/lib:/opt/aws-ofi-nccl-inst +all/lib:/opt/aws-ofi-nccl/lib:$LD_LIBRARY_PATH +export NCCL_PROTO=simple +export PATH=/opt/amazon/efa/bin:$PATH +export LD_PRELOAD="/opt/nccl/build/lib/libnccl.so" +export FI_EFA_FORK_SAFE=1 +export FI_LOG_LEVEL=1 +export FI_EFA_USE_DEVICE_RDMA=1 # use for p4dn +export NCCL_DEBUG=info +export PYTHONFAULTHANDLER=1 +export CUDA_LAUNCH_BLOCKING=0 +export OMPI_MCA_mtl_base_verbose=1 +export FI_EFA_ENABLE_SHM_TRANSFER=0 +export FI_PROVIDER=efa +export FI_EFA_TX_MIN_CREDITS=64 +export NCCL_TREE_THRESHOLD=0 + +# sent to sub script +export HOSTNAMES=`scontrol show hostnames "$SLURM_JOB_NODELIST"` +export MASTER_ADDR=$(scontrol show hostnames "$SLURM_JOB_NODELIST" | head -n 1) +export MASTER_PORT=12802 +export COUNT_NODE=`scontrol show hostnames "$SLURM_JOB_NODELIST" | wc -l` +export WORLD_SIZE=$COUNT_NODE + +echo go $COUNT_NODE +echo $HOSTNAMES +echo $WORLD_SIZE + +mpirun -n $COUNT_NODE -perhost 1 /fsx/stable-diffusion/stable-diffusion/scripts/slurm/v1_laionhr_torch111/launcher.sh diff --git a/scripts/test_gpu.py b/scripts/test_gpu.py new file mode 100644 index 0000000..36d4e78 --- /dev/null +++ b/scripts/test_gpu.py @@ -0,0 +1,30 @@ +import socket +try: + import torch + n_gpus = torch.cuda.device_count() + print(f"checking {n_gpus} gpus.") + for i_gpu in range(n_gpus): + print(i_gpu) + device = torch.device(f"cuda:{i_gpu}") + torch.backends.cuda.matmul.allow_tf32 = False + torch.backends.cudnn.benchmark = True + torch.backends.cudnn.deterministic = False + torch.backends.cudnn.allow_tf32 = True + data = torch.randn([4, 640, 32, 32], dtype=torch.float, device=device, requires_grad=True) + net = torch.nn.Conv2d(640, 640, kernel_size=[3, 3], padding=[1, 1], stride=[1, 1], dilation=[1, 1], groups=1) + net = net.to(device=device).float() + out = net(data) + out.backward(torch.randn_like(out)) + torch.cuda.synchronize() +except RuntimeError as err: + import requests + import datetime + import os + device = os.environ.get("CUDA_VISIBLE_DEVICES", "?") + hostname = socket.gethostname() + ts = datetime.datetime.utcnow().strftime('%Y-%m-%d %H:%M:%S') + resp = requests.get('http://169.254.169.254/latest/meta-data/instance-id') + print(f'ERROR at {ts} on {hostname}/{resp.text} (CUDA_VISIBLE_DEVICES={device}): {type(err).__name__}: {err}', flush=True) + raise err +else: + print(f"checked {socket.gethostname()}") diff --git a/scripts/test_gpu.sh b/scripts/test_gpu.sh new file mode 100644 index 0000000..7a95237 --- /dev/null +++ b/scripts/test_gpu.sh @@ -0,0 +1,5 @@ +#!/bin/bash +eval "$(/fsx/stable-diffusion/debug/miniconda3/bin/conda shell.bash hook)" +conda activate stable +cd /fsx/stable-diffusion/stable-diffusion +python scripts/test_gpu.py