Add sam2act_plus_all_v4 (SAM2Act+, full 10 epochs)
Browse files- sam2act_plus_all_v4/args_plus.yaml +9 -0
- sam2act_plus_all_v4/exp_cfg_plus.yaml +40 -0
- sam2act_plus_all_v4/model_last.pth +3 -0
- sam2act_plus_all_v4/model_plus_0.pth +3 -0
- sam2act_plus_all_v4/model_plus_1.pth +3 -0
- sam2act_plus_all_v4/model_plus_2.pth +3 -0
- sam2act_plus_all_v4/model_plus_3.pth +3 -0
- sam2act_plus_all_v4/model_plus_4.pth +3 -0
- sam2act_plus_all_v4/model_plus_5.pth +3 -0
- sam2act_plus_all_v4/model_plus_6.pth +3 -0
- sam2act_plus_all_v4/model_plus_7.pth +3 -0
- sam2act_plus_all_v4/model_plus_8.pth +3 -0
- sam2act_plus_all_v4/model_plus_9.pth +3 -0
- sam2act_plus_all_v4/model_plus_last.pth +3 -0
- sam2act_plus_all_v4/mvt_cfg_plus.yaml +50 -0
sam2act_plus_all_v4/args_plus.yaml
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
device: '0'
|
| 2 |
+
exp_cfg_opts: tasks PickXtimes,StopCube,SwingXtimes,BinFill,VideoUnmaskSwap,VideoUnmask,ButtonUnmaskSwap,ButtonUnmask,VideoRepick,VideoPlaceButton,VideoPlaceOrder,PickHighlight,InsertPeg,MoveCube,PatternLock,RouteStick
|
| 3 |
+
train_iter 800000 epochs 10 bs 20 exp_id sam2act exp_name plus_all_v4
|
| 4 |
+
exp_cfg_path: configs/sam2act_plus.yaml
|
| 5 |
+
log_dir: runs
|
| 6 |
+
mvt_cfg_opts: use_memory True
|
| 7 |
+
mvt_cfg_path: mvt/configs/sam2act_plus.yaml
|
| 8 |
+
refresh_replay: false
|
| 9 |
+
with_eval: false
|
sam2act_plus_all_v4/exp_cfg_plus.yaml
ADDED
|
@@ -0,0 +1,40 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
agent: our
|
| 2 |
+
bs: 20
|
| 3 |
+
demo: 100
|
| 4 |
+
epochs: 10
|
| 5 |
+
exp_id: sam2act
|
| 6 |
+
exp_name: plus_all_v4
|
| 7 |
+
num_workers: 5
|
| 8 |
+
peract:
|
| 9 |
+
add_rgc_loss: true
|
| 10 |
+
amp: true
|
| 11 |
+
bnb: true
|
| 12 |
+
lambda_weight_l2: 0.0001
|
| 13 |
+
lr: 1.25e-06
|
| 14 |
+
lr_cos_dec: true
|
| 15 |
+
num_rotation_classes: 72
|
| 16 |
+
optimizer_type: lamb
|
| 17 |
+
same_trans_aug_per_seq: false
|
| 18 |
+
transform_augmentation: true
|
| 19 |
+
transform_augmentation_rpy:
|
| 20 |
+
- 0.0
|
| 21 |
+
- 0.0
|
| 22 |
+
- 45.0
|
| 23 |
+
transform_augmentation_xyz:
|
| 24 |
+
- 0.125
|
| 25 |
+
- 0.125
|
| 26 |
+
- 0.125
|
| 27 |
+
warmup_steps: 2000
|
| 28 |
+
peract_official:
|
| 29 |
+
cfg_path: configs/peract_official_config.yaml
|
| 30 |
+
resume: ''
|
| 31 |
+
rvt:
|
| 32 |
+
gt_hm_sigma: 1.5
|
| 33 |
+
img_aug: 0.0
|
| 34 |
+
move_pc_in_bound: true
|
| 35 |
+
place_with_mean: false
|
| 36 |
+
sample_distribution_mode: demo_uniform_temporal
|
| 37 |
+
tasks: PickXtimes,StopCube,SwingXtimes,BinFill,VideoUnmaskSwap,VideoUnmask,ButtonUnmaskSwap,ButtonUnmask,VideoRepick,VideoPlaceButton,VideoPlaceOrder,PickHighlight,InsertPeg,MoveCube,PatternLock,RouteStick
|
| 38 |
+
train_iter: 800000
|
| 39 |
+
wandb: true
|
| 40 |
+
|
sam2act_plus_all_v4/model_last.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1c3d4bc62bd2b238337fe21de95d14d78c77e42f9751a888cd9404d1fca8b73b
|
| 3 |
+
size 1185067109
|
sam2act_plus_all_v4/model_plus_0.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7ab90b7a64db6cb5ae4bb16aaafeef8a523b58b9da2f5f5e56278f7a1f1c612b
|
| 3 |
+
size 638654249
|
sam2act_plus_all_v4/model_plus_1.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4f53ee5779ac4f128d8e469d915fba4e523df805a5a8fcb43cd3639e6d8a1600
|
| 3 |
+
size 638654249
|
sam2act_plus_all_v4/model_plus_2.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:077d0620137ad13ece56e1e29b5c3f54fce0013062e58447025712d645396c76
|
| 3 |
+
size 638654249
|
sam2act_plus_all_v4/model_plus_3.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:da0dd56a3d2102ac91f35074a07220790cb65aed215f6c5b778fbacc2866bd36
|
| 3 |
+
size 638654249
|
sam2act_plus_all_v4/model_plus_4.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b30e5c5a15273545123e547bebde831d7497c6fa09b2678080f58aee900ce5bd
|
| 3 |
+
size 638654249
|
sam2act_plus_all_v4/model_plus_5.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:156c240cef32843270ba59a6e9027a3c4bf8140dc4757aea980ecd859f54b8ba
|
| 3 |
+
size 638654249
|
sam2act_plus_all_v4/model_plus_6.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0c1842feedadcaa3f6778e754d26e83082d946090f4ba1f986a510c7372b0648
|
| 3 |
+
size 638654633
|
sam2act_plus_all_v4/model_plus_7.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b5d02039bc49215db1ea6e3728bbbcc00b18ee78fa86b98f6d0fbaba9b2f354f
|
| 3 |
+
size 638654633
|
sam2act_plus_all_v4/model_plus_8.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a09d365d5f58b8eef4d81f0f418d534cf6918f6e04e8448b9532f39037f3a556
|
| 3 |
+
size 638654633
|
sam2act_plus_all_v4/model_plus_9.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2f0ad897b2dc37b5f5f84117e078ac54b8991fea0b4be5a5afe916f1391a175a
|
| 3 |
+
size 638654633
|
sam2act_plus_all_v4/model_plus_last.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cb7167b32a5d63d0b8714cacd40a08615f1c5790c36531e89fc61b4098b8a088
|
| 3 |
+
size 638659442
|
sam2act_plus_all_v4/mvt_cfg_plus.yaml
ADDED
|
@@ -0,0 +1,50 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
activation: lrelu
|
| 2 |
+
add_corr: true
|
| 3 |
+
add_depth: true
|
| 4 |
+
add_lang: true
|
| 5 |
+
add_pixel_loc: true
|
| 6 |
+
add_proprio: true
|
| 7 |
+
attn_dim: 512
|
| 8 |
+
attn_dim_head: 64
|
| 9 |
+
attn_dropout: 0.1
|
| 10 |
+
attn_heads: 8
|
| 11 |
+
cvx_up: true
|
| 12 |
+
decoder_dropout: 0.0
|
| 13 |
+
depth: 8
|
| 14 |
+
feat_dim: 220
|
| 15 |
+
feat_ver: 1
|
| 16 |
+
final_dim: 64
|
| 17 |
+
ifSAM2: true
|
| 18 |
+
ifsep: false
|
| 19 |
+
im_channels: 64
|
| 20 |
+
img_aug_2: 0.05
|
| 21 |
+
img_feat_dim: 3
|
| 22 |
+
img_patch_size: 14
|
| 23 |
+
img_size: 224
|
| 24 |
+
inp_pre_con: false
|
| 25 |
+
inp_pre_pro: false
|
| 26 |
+
lang_dim: 512
|
| 27 |
+
lang_len: 77
|
| 28 |
+
lora_finetune: true
|
| 29 |
+
lora_r: 16
|
| 30 |
+
norm_corr: true
|
| 31 |
+
num_maskmem: 19
|
| 32 |
+
num_rot: 72
|
| 33 |
+
pe_fix: true
|
| 34 |
+
proprio_dim: 4
|
| 35 |
+
rend_three_views: true
|
| 36 |
+
resize_rgb: true
|
| 37 |
+
rot_ver: 1
|
| 38 |
+
sam2_ckpt: /nfs/turbo/coe-chaijy-unreplicated/hongzefu/sam2act_historybench/sam2act/mvt/sam2_train/checkpoints/sam2.1_hiera_base_plus.pt
|
| 39 |
+
sam2_config: /configs/sam2.1/sam2.1_hiera_b+
|
| 40 |
+
self_cross_ver: 1
|
| 41 |
+
st_sca: 4
|
| 42 |
+
st_wpt_loc_aug: 0.05
|
| 43 |
+
st_wpt_loc_inp_no_noise: true
|
| 44 |
+
stage_two: true
|
| 45 |
+
use_memory: true
|
| 46 |
+
use_point_renderer: true
|
| 47 |
+
weight_tie_layers: false
|
| 48 |
+
wpt_img_aug: 0.0
|
| 49 |
+
xops: true
|
| 50 |
+
|