-
Notifications
You must be signed in to change notification settings - Fork 2
/
rerun2.sh
56 lines (49 loc) · 3.2 KB
/
rerun2.sh
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
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
#!/bin/bash
## pre-training and linear-probing evaluation with Transformer backbone.
LanguageArray=(
# "pretrain_PointCAE_clean" ## No_corruptions
# ###add noise
# "pretrain_PointCAE_add_global"
# "pretrain_PointCAE_add_local"
# "pretrain_PointCAE_jitter"
# # drop points
# "pretrain_PointCAE_transformer_affine_r3"
# "pretrain_PointCAE_transformer_dropout_patch"
# "pretrain_PointCAE_transformer_dropout_patch_affine_r3_maskpatch_p0005"
"pretrain_PointCAE_transformer_dropout_patch_affine_r3_maskpatch_p0005_whole" ## verified, other yamls may require tiny modification before use.
# "pretrain_PointCAE_transformer_dropout_patch_affine_r3_maskpatch_p0005_whole_4xlonger"
# "pretrain_PointCAE_nonuniform_density" ## Scan
# ##
# "pretrain_PointCAE_rotate_z"
# "pretrain_PointCAE_rotate"
# "pretrain_PointCAE_reflection"
# "pretrain_PointCAE_scale_nonorm"
# "pretrain_PointCAE_shear"
# "pretrain_PointCAE_translate"
# "pretrain_PointCAE_affine_r3" ## Affine
# # affine transformation combination
# "pretrain_PointCAE_affine_r3_dropout_local"
# "pretrain_PointCAE_affine_r3_dropout_patch"
)
# !!!!!!!!!!! IF you want to do testing only, please update the "./experiments/${YAML}${ModelName_method}/cfgs/tslog/ckpt-last.pth" to the ckpt-last.pth downloaded from:
# https://drive.google.com/drive/folders/1vAktAxgSNPNTNwLG89xcfrmRDWIYlhEH?usp=sharing ,which provide the checkpoint for 'pretrain_PointCAE_transformer_dropout_patch_affine_r3_maskpatch_p0005_whole' setup.
for random in $(seq 1 1); do
for YAML in ${LanguageArray[*]}; do
ModelName_method=PointCAE_transformer_fc_global_folding_local
total_bs=256
### pretraining
CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 python main.py --config cfgs/${YAML}.yaml --exp_name tslog --model_name ${ModelName_method} --total_bs ${total_bs} --num_workers 4
ModelName=PointTransformerNoClassTokenSVMFeature
total_bs=16
# training a SVM classifier on pre-extracted features.
# IF you want to do testing only, please update the "./experiments/${YAML}${ModelName_method}/cfgs/tslog/ckpt-last.pth" to the .pth downloaded here.
CUDA_VISIBLE_DEVICES=0 python main.py --config cfgs/finetune_scan_hardest_svm_classification_clean.yaml \
--finetune_model --svm_classification --exp_name ${YAML} --ckpts ./experiments/${YAML}${ModelName_method}/cfgs/tslog/ckpt-last.pth --model_name ${ModelName} --total_bs ${total_bs}
CUDA_VISIBLE_DEVICES=0 python main.py --config cfgs/finetune_scan_objbg_svm_classification_clean.yaml \
--finetune_model --svm_classification --exp_name ${YAML} --ckpts ./experiments/${YAML}${ModelName_method}/cfgs/tslog/ckpt-last.pth --model_name ${ModelName} --total_bs ${total_bs}
# CUDA_VISIBLE_DEVICES=1 python main.py --config cfgs/finetune_scan_objonly_svm_classification_clean.yaml \
# --finetune_model --svm_classification --exp_name ${YAML} --ckpts ./experiments/${YAML}${ModelName_method}/cfgs/log/ckpt-last.pth --model_name ${ModelName} --total_bs ${total_bs}
# CUDA_VISIBLE_DEVICES=1 python main.py --config cfgs/finetune_modelnet_svm_classification.yaml \
# --finetune_model --svm_classification --exp_name ${YAML} --ckpts ./experiments/${YAML}${ModelName_method}/cfgs/log/ckpt-last.pth --model_name ${ModelName} --total_bs ${total_bs}
done
done