Hugging Face's logo

Spaces:
liruiw
/
hma
Running on Zero

App Files Community
1
hma
File size: 1,242 Bytes 
246c106
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
 
#!/bin/bash
set -x
ulimit -c 0


# check if NUM_GPU is unset, if so set it to 8
NUM_GPU=${NUM_GPU:-8}
# SLURM_JOB_NUM_NODES to 1
SLURM_JOB_NUM_NODES=${SLURM_JOB_NUM_NODES:-1}
# MASTER_ADDR to localhost
MASTER_ADDR=${MASTER_ADDR:-localhost}
# MASTER_PORT to 29500
MASTER_PORT=${MASTER_PORT:-29500}
script_name=final2_40dataset_waction_traj10000_gpu_${NUM_GPU}_nodes_${SLURM_JOB_NUM_NODES}_68536steps

source venv/bin/activate
torchrun   --nnodes=${SLURM_JOB_NUM_NODES} --nproc_per_node=${NUM_GPU} \
    --rdzv-id=${SLURM_JOB_ID} --rdzv-backend=c10d --rdzv-endpoint=${MASTER_ADDR}:${MASTER_PORT} \
     train_multi.py --genie_config genie/configs/magvit_n32_h8_d256_action_concat.json \
    --output_dir data/$script_name \
    --max_eval_steps 10 \
    --num_episodes_per_dataset 10000 \
    --max_train_steps 68536 \
    --save_second_epoch \
    --per_device_train_batch_size 8  \
    --gradient_accumulation_steps 1 \
    --run_name $script_name \
    --resume_from_checkpoint data/$script_name/  \
    --train_split experiments/datasplit/dataset40.yaml #&&

#chmod +x experiments/scripts/eval_action_scripts/run_evaluation_multidataset.sh
#bash experiments/scripts/eval_action_scripts/run_evaluation_multidataset.sh  $script_name dataset40