add scripts

scripts/eval_all.sh

@@ -0,0 +1,34 @@
+#!/bin/bash
+#for model in ddgan_sd_v10 ddgan_laion2b_v2 ddgan_ddb_v1 ddgan_ddb_v2 ddgan_ddb_v3 ddgan_ddb_v4;do 
+#for model in ddgan_ddb_v2 ddgan_ddb_v3 ddgan_ddb_v4 ddgan_ddb_v5;do 
+#for model in ddgan_ddb_v4 ddgan_ddb_v6 ddgan_ddb_v7 ddgan_laion_aesthetic_v15;do 
+#for model in ddgan_ddb_v6;do 
+#for model in ddgan_laion_aesthetic_v15;do 
+#for model in ddgan_ddb_v3 ddgan_ddb_v11 ddgan_laion_aesthetic_v15;do 
+#for model in ddgan_ddb_v3 ddgan_ddb_v11 ddgan_ddb_v2;do 
+#for model in ddgan_ddb_v6 ddgan_ddb_v4 ddgan_ddb_v10 ddgan_ddb_v9;do 
+#for model in ddgan_ddb_v3 ddgan_ddb_v11;do 
+for model in ddgan_ddb_v11;do 
+#for model in ddgan_ddb_v3;do 
+    if [ "$model" == "ddgan_ddb_v3" ]; then
+        bs=32
+    elif [ "$model" == "ddgan_laion_aesthetic_v15" ]; then
+        bs=32
+    elif [ "$model" == "ddgan_ddb_v6" ]; then
+        bs=32
+    elif [ "$model" == "ddgan_ddb_v4" ]; then
+        bs=16
+    elif [ "$model" == "ddgan_ddb_v9" ]; then
+        bs=16
+    elif [ "$model" == "ddgan_ddb_v10" ]; then
+        bs=16
+    elif [ "$model" == "ddgan_ddb_v11" ]; then
+        bs=16
+    else
+        bs=64
+    fi
+    sbatch --partition dc-gpu -t 360 -N 1 -n1 scripts/run_jurecadc_ddp.sh test_ddgan.py --name $model --cond-text=parti_prompts.txt  --batch-size=$bs --epoch=-1 --compute-image-reward --eval-name=parti_image_reward
+    #sbatch --partition dc-gpu -t 360 -N 1 -n1 scripts/run_jurecadc_ddp.sh test_ddgan.py --name $model --cond-text=parti_prompts.txt  --batch-size=$bs --epoch=-1 --compute-clip-score --eval-name=parti;
+    #sbatch --partition dc-gpu -t 360 -N 1 -n1 scripts/run_jurecadc_ddp.sh test_ddgan.py --name $model --fid --real-img-dir inception_statistics_coco_val2014_256x256.npz --cond-text coco_val2014_captions.txt  --batch-size=$bs --epoch=-1 --nb-images-for-fid=30000 --eval-name=coco --compute-clip-score;
+    #sbatch --partition dc-gpu -t 360 -N 1 -n1 scripts/run_jurecadc_ddp.sh test_ddgan.py --name $model --cond-text=drawbench.txt  --batch-size=$bs --epoch=-1 --compute-clip-score --eval-name=drawbench;
+done

scripts/init.sh

@@ -0,0 +1,14 @@
+ml purge
+ml use $OTHERSTAGES
+ml Stages/2022
+ml GCC/11.2.0
+ml OpenMPI/4.1.2
+ml CUDA/11.5
+ml cuDNN/8.3.1.22-CUDA-11.5
+ml NCCL/2.12.7-1-CUDA-11.5
+ml PyTorch/1.11-CUDA-11.5
+ml Horovod/0.24
+ml torchvision/0.12.0
+source /p/home/jusers/cherti1/jureca/ccstdl/code/feed_forward_vqgan_clip/envs/jureca_2022/bin/activate
+export HOROVOD_CACHE_CAPACITY=4096
+export CUDA_VISIBLE_DEVICES=0,1,2,3

scripts/run_hdfml.sh

@@ -0,0 +1,25 @@
+#!/bin/bash -x
+#SBATCH --account=cstdl
+#SBATCH --nodes=8
+#SBATCH --ntasks-per-node=4
+#SBATCH --cpus-per-task=8
+#SBATCH --time=06:00:00
+#SBATCH --gres=gpu
+#SBATCH --partition=batch
+ml purge
+ml use $OTHERSTAGES
+ml Stages/2022
+ml GCC/11.2.0
+ml OpenMPI/4.1.2
+ml CUDA/11.5
+ml cuDNN/8.3.1.22-CUDA-11.5
+ml NCCL/2.12.7-1-CUDA-11.5
+ml PyTorch/1.11-CUDA-11.5
+ml Horovod/0.24
+ml torchvision/0.12.0
+source envs/hdfml/bin/activate
+export CUDA_VISIBLE_DEVICES=0,1,2,3
+echo "Job id: $SLURM_JOB_ID"
+export TOKENIZERS_PARALLELISM=false
+export NCCL_ASYNC_ERROR_HANDLING=1
+srun python -u $*

scripts/run_jurecadc_conda.sh

@@ -0,0 +1,23 @@
+#!/bin/bash -x
+#SBATCH --account=zam
+#SBATCH --nodes=1
+#SBATCH --ntasks-per-node=4
+#SBATCH --cpus-per-task=24
+#SBATCH --time=06:00:00
+#SBATCH --gres=gpu:4
+#SBATCH --partition=dc-gpu
+ml CUDA
+source /p/project/laionize/miniconda/bin/activate
+conda activate ddgan
+#source scripts/init_2022.sh
+#source scripts/init_2020.sh
+#source scripts/init.sh
+export CUDA_VISIBLE_DEVICES=0,1,2,3
+echo "Job id: $SLURM_JOB_ID"
+export TOKENIZERS_PARALLELISM=false
+#export NCCL_ASYNC_ERROR_HANDLING=1
+export NCCL_IB_TIMEOUT=50
+export UCX_RC_TIMEOUT=4s
+export NCCL_IB_RETRY_CNT=10
+export TORCH_DISTRIBUTED_DEBUG=INFO
+srun python -u $*

scripts/run_jurecadc_ddp.sh

@@ -0,0 +1,21 @@
+#!/bin/bash -x
+#SBATCH --account=zam
+#SBATCH --nodes=1
+#SBATCH --ntasks-per-node=4
+#SBATCH --cpus-per-task=24
+#SBATCH --time=06:00:00
+#SBATCH --gres=gpu:4
+#SBATCH --partition=dc-gpu
+source set_torch_distributed_vars.sh
+#source scripts/init_2022.sh
+#source scripts/init_2020.sh
+source scripts/init.sh
+export CUDA_VISIBLE_DEVICES=0,1,2,3
+echo "Job id: $SLURM_JOB_ID"
+export TOKENIZERS_PARALLELISM=false
+#export NCCL_ASYNC_ERROR_HANDLING=1
+export NCCL_IB_TIMEOUT=50
+export UCX_RC_TIMEOUT=4s
+export NCCL_IB_RETRY_CNT=10
+export TRANSFORMERS_CACHE=cache
+srun python -u $*

scripts/run_jusuf_ddp.sh

@@ -0,0 +1,14 @@
+#!/bin/bash -x
+#SBATCH --account=zam
+#SBATCH --nodes=1
+#SBATCH --ntasks-per-node=1
+#SBATCH --cpus-per-task=24
+#SBATCH --time=06:00:00
+#SBATCH --gres=gpu:1
+#SBATCH --partition=gpus
+source set_torch_distributed_vars.sh
+source scripts/init.sh
+export CUDA_VISIBLE_DEVICES=0,1,2,3
+echo "Job id: $SLURM_JOB_ID"
+export TOKENIZERS_PARALLELISM=false
+srun python -u $*

scripts/run_juwelsbooster_conda.sh

@@ -0,0 +1,19 @@
+#!/bin/bash -x
+#SBATCH --account=laionize
+#SBATCH --nodes=1
+#SBATCH --ntasks-per-node=4
+#SBATCH --cpus-per-task=24
+#SBATCH --time=06:00:00
+#SBATCH --gres=gpu:4
+#SBATCH --partition=booster
+ml CUDA
+source /p/project/laionize/miniconda/bin/activate
+conda activate ddgan
+export CUDA_VISIBLE_DEVICES=0,1,2,3
+echo "Job id: $SLURM_JOB_ID"
+export TOKENIZERS_PARALLELISM=false
+#export NCCL_ASYNC_ERROR_HANDLING=1
+export NCCL_IB_TIMEOUT=50
+export UCX_RC_TIMEOUT=4s
+export NCCL_IB_RETRY_CNT=10
+srun python -u $*

scripts/run_juwelsbooster_ddp.sh

@@ -0,0 +1,17 @@
+#!/bin/bash -x
+#SBATCH --account=covidnetx
+#SBATCH --nodes=4
+#SBATCH --ntasks-per-node=4
+#SBATCH --cpus-per-task=24
+#SBATCH --time=06:00:00
+#SBATCH --gres=gpu:4
+#SBATCH --partition=booster
+source set_torch_distributed_vars.sh
+#source scripts/init_2022.sh
+#source scripts/init_2020.sh
+source scripts/init.sh
+export CUDA_VISIBLE_DEVICES=0,1,2,3
+echo "Job id: $SLURM_JOB_ID"
+export TOKENIZERS_PARALLELISM=false
+export NCCL_ASYNC_ERROR_HANDLING=1
+srun python -u $*