update multi_noiides

README.md

@@ -178,10 +178,10 @@ python -m pip install -e detectron2
 ```
 
 ## 训练
-下载预训练模型 [MAE ViT-Large model ](https://dl.fbaipublicfiles.com/mae/pretrain/mae_pretrain_vit_large.pth), 修改 `$Painter_ROOT/train_painter_vit_large.sh` 中finetune参数地址.
+下载预训练模型 [MAE ViT-Large model ](https://dl.fbaipublicfiles.com/mae/pretrain/mae_pretrain_vit_large.pth), 修改 `$Painter_ROOT/train.sh` 中finetune参数地址.
 
 ### 单机多卡
-本项目默认参数是单机4卡 (total_bsz = 1x4x32 = 128), 如需使用其他的卡数, 请修改 train_painter_vit_large.sh 中对应参数.
+本项目默认参数是单机4卡 (total_bsz = 1x4x32 = 128), 如需使用其他的卡数, 请修改 train.sh 中对应参数.
 ```bash
 bash train.sh
 ```

run_train_multi.sh

@@ -8,13 +8,13 @@ np=$(($np*4))
 echo $np
 
 nodename=$(cat $hostfile |sed -n "1p")
-dist_url=`echo $nodename | awk '{print $1}'`
+export dist_url=`echo $nodename | awk '{print $1}'`
 which mpirun
 
 # 添加pythonlib环境, 用户需修改为自己的环境变量地址
 export LD_LIBRARY_PATH=${LD_LIBRARY_PATH}:/path/of/conda/envs/{env_name}/lib
 export PYTHON=python3
-
+export NPROC_PER_NODE=4
 
 # -np 显卡数量
 # -x 将变量传递到single_process.sh脚本中

single_process.sh

@@ -16,7 +16,7 @@ export NCCL_CROSS_NIC=1
 export RCCL_NCHANNELS=4
 export MASTER_ADDR=$dist_url
 export MASTER_PORT=4321
-export RANK=$OMPI_COMM_WORLD_RANK
+export LOCAL_RANK=$OMPI_COMM_WORLD_LOCAL_RANK
 
 lrank=$OMPI_COMM_WORLD_LOCAL_RANK
 

train.sh

@@ -6,6 +6,8 @@ export USE_MIOPEN_BATCHNORM=1
 DATA_PATH=/home/datasets
 name=painter_vit_large
 python -m torch.distributed.launch --nproc_per_node=4 \
+	--nnodes=${WORLD_SIZE} --node_rank=$RANK \
+	--master_addr=$MASTER_ADDR --master_port=12358 \
 	--use_env main_train.py  \
     --batch_size 2 \
     --accum_iter 16  \

train_painter_vit_large.sh

-#!/bin/bash
-export HSA_FORCE_FINE_GRAIN_PCIE=1
-export USE_MIOPEN_BATCHNORM=1
-
-DATA_PATH=/home/datasets
-name=painter_vit_large
-python -m torch.distributed.launch --nproc_per_node=8 \
-	--nnodes=${WORLD_SIZE} --node_rank=$RANK \
-	--master_addr=$MASTER_ADDR --master_port=12358 \
-	--use_env main_train.py  \
-    --batch_size 2 \
-    --accum_iter 16  \
-    --model painter_vit_large_patch16_input896x448_win_dec64_8glb_sl1 \
-    --num_mask_patches 784 \
-    --max_mask_patches_per_block 392 \
-    --epochs 15 \
-    --warmup_epochs 1 \
-    --lr 1e-3 \
-    --clip_grad 3 \
-    --layer_decay 0.8 \
-    --drop_path 0.1 \
-    --input_size 896 448 \
-    --save_freq 1 \
-    --data_path $DATA_PATH/ \
-    --json_path  \
-    $DATA_PATH/nyu_depth_v2/nyuv2_sync_image_depth.json \
-    $DATA_PATH/ade20k/ade20k_training_image_semantic.json \
-    $DATA_PATH/coco/pano_ca_inst/coco_train_image_panoptic_inst.json \
-    $DATA_PATH/coco/pano_sem_seg/coco_train2017_image_panoptic_sem_seg.json \
-    $DATA_PATH/coco_pose/coco_pose_256x192_train.json \
-    $DATA_PATH/denoise/denoise_ssid_train.json \
-    $DATA_PATH/derain/derain_train.json \
-    $DATA_PATH/light_enhance/enhance_lol_train.json \
-    --val_json_path \
-    $DATA_PATH/nyu_depth_v2/nyuv2_test_image_depth.json \
-    $DATA_PATH/ade20k/ade20k_validation_image_semantic.json \
-    $DATA_PATH/coco/pano_ca_inst/coco_val_image_panoptic_inst.json \
-    $DATA_PATH/coco/pano_sem_seg/coco_val2017_image_panoptic_sem_seg.json \
-    $DATA_PATH/coco_pose/coco_pose_256x192_val.json \
-    $DATA_PATH/denoise/denoise_ssid_val.json \
-    $DATA_PATH/derain/derain_test_rain100h.json \
-    $DATA_PATH/light_enhance/enhance_lol_val.json \
-    --output_dir models/$name \
-    --log_dir models/$name/logs \
-    --finetune path/to/mae_pretrain_vit_large.pth \
-    # --log_wandb \
-