From b6e250370b54baf65a8a6decd0dda3087d5cd238 Mon Sep 17 00:00:00 2001 From: Mike Walmsley Date: Sat, 4 Nov 2023 20:54:48 -0400 Subject: [PATCH] typo --- only_for_me/narval/train.py | 4 ++-- only_for_me/narval/train.sh | 6 +++--- 2 files changed, 5 insertions(+), 5 deletions(-) diff --git a/only_for_me/narval/train.py b/only_for_me/narval/train.py index a0faaebb..25c755e2 100644 --- a/only_for_me/narval/train.py +++ b/only_for_me/narval/train.py @@ -126,8 +126,8 @@ num_workers=6, random_state=random_state, learning_rate=1e-3, - cache_dir=os.environ['SLURM_TMPDIR'] + '/cache' - # cache_dir='/tmp/cache' + # cache_dir=os.environ['SLURM_TMPDIR'] + '/cache' + cache_dir='/tmp/cache' # /tmp for ramdisk (400GB total, vs 4TB total for nvme) ) diff --git a/only_for_me/narval/train.sh b/only_for_me/narval/train.sh index 3fff67f9..f5801794 100644 --- a/only_for_me/narval/train.sh +++ b/only_for_me/narval/train.sh @@ -10,8 +10,8 @@ nvidia-smi PYTHON=/home/walml/envs/zoobot39_dev/bin/python -mkdir $SLURM_TMPDIR/cache -# mkdir /tmp/cache +# mkdir $SLURM_TMPDIR/cache +mkdir /tmp/cache export NCCL_BLOCKING_WAIT=1 #Set this environment variable if you wish to use the NCCL backend for inter-GPU communication. # export MASTER_ADDR=$(hostname) #Store the master node’s IP address in the MASTER_ADDR environment variable. @@ -22,6 +22,6 @@ REPO_DIR=/project/def-bovy/walml/zoobot/ srun $PYTHON $REPO_DIR/only_for_me/narval/train.py \ --save-dir $REPO_DIR/only_for_me/narval/debug_models \ --batch-size 128 \ - --architecture maxvit_tiny_224 \ + --architecture maxvit_tiny_tf_224 \ --color --wandb --mixed-precision # srun python $SLURM_TMPDIR/zoobot/only_for_me/narval/finetune.py