add model checkpoints
Browse files- v1t_franke2022/args.yaml +71 -0
- v1t_franke2022/ckpt/model_state.pt +3 -0
- v1t_sensorium/args.yaml +62 -0
- v1t_sensorium/ckpt/model_state.pt +3 -0
v1t_franke2022/args.yaml
ADDED
|
@@ -0,0 +1,71 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
adam_beta1: 0.9
|
| 2 |
+
adam_beta2: 0.9999
|
| 3 |
+
adam_eps: 1e-08
|
| 4 |
+
amp: false
|
| 5 |
+
backend: null
|
| 6 |
+
batch_size: 16
|
| 7 |
+
behavior_mode: 3
|
| 8 |
+
bias_mode: 0
|
| 9 |
+
center_crop: 1.0
|
| 10 |
+
clear_output_dir: true
|
| 11 |
+
core: vit
|
| 12 |
+
core_lr: 0.001647
|
| 13 |
+
core_reg_scale: 0.5379
|
| 14 |
+
criterion: poisson
|
| 15 |
+
cropper_reg_scale: 0.0
|
| 16 |
+
dataset: data/franke2022
|
| 17 |
+
deterministic: false
|
| 18 |
+
device: cuda
|
| 19 |
+
disable_bias: false
|
| 20 |
+
disable_grid_predictor: false
|
| 21 |
+
dpi: 120
|
| 22 |
+
drop_path: 0.0
|
| 23 |
+
ds_name: franke2022
|
| 24 |
+
ds_scale: 1
|
| 25 |
+
emb_dim: 155
|
| 26 |
+
epochs: 400
|
| 27 |
+
format: svg
|
| 28 |
+
git_hash: dd1cb29
|
| 29 |
+
grad_checkpointing: 0
|
| 30 |
+
gray_scale: false
|
| 31 |
+
grid_predictor_dim: 2
|
| 32 |
+
hostname: gpu-q-39
|
| 33 |
+
input_shape: [2, 36, 64]
|
| 34 |
+
limit_data: null
|
| 35 |
+
lr: 0.001647
|
| 36 |
+
micro_batch_size: 16
|
| 37 |
+
mlp_dim: 488
|
| 38 |
+
mouse_ids: [F, G, H, I, J, K, L, M, N, O]
|
| 39 |
+
num_blocks: 4
|
| 40 |
+
num_heads: 4
|
| 41 |
+
num_workers: 2
|
| 42 |
+
output_dir: runs/franke2022/013_v1t
|
| 43 |
+
output_shapes:
|
| 44 |
+
F: [867]
|
| 45 |
+
G: [922]
|
| 46 |
+
H: [773]
|
| 47 |
+
I: [1023]
|
| 48 |
+
J: [910]
|
| 49 |
+
K: [1121]
|
| 50 |
+
L: [1125]
|
| 51 |
+
M: [1160]
|
| 52 |
+
N: [824]
|
| 53 |
+
O: [1109]
|
| 54 |
+
p_dropout: 0.0229
|
| 55 |
+
patch_mode: 0
|
| 56 |
+
patch_size: 8
|
| 57 |
+
patch_stride: 1
|
| 58 |
+
pretrain_core: ''
|
| 59 |
+
readout: gaussian2d
|
| 60 |
+
readout_reg_scale: 0.0076
|
| 61 |
+
resize_image: 1
|
| 62 |
+
save_plots: false
|
| 63 |
+
seed: 1234
|
| 64 |
+
shift_mode: 2
|
| 65 |
+
shifter_reg_scale: 0.0
|
| 66 |
+
t_dropout: 0.2544
|
| 67 |
+
trainable_params: 4050650
|
| 68 |
+
use_lsa: false
|
| 69 |
+
use_wandb: true
|
| 70 |
+
verbose: 1
|
| 71 |
+
wandb_group: franke2022
|
v1t_franke2022/ckpt/model_state.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d74fecf88eec5b1b7b60bca6c1923011b3e82c8f2c1f1c2d36f5fbe9766aebd3
|
| 3 |
+
size 48940197
|
v1t_sensorium/args.yaml
ADDED
|
@@ -0,0 +1,62 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
adam_beta1: 0.9
|
| 2 |
+
adam_beta2: 0.9999
|
| 3 |
+
adam_eps: 1e-08
|
| 4 |
+
amp: false
|
| 5 |
+
batch_size: 16
|
| 6 |
+
behavior_mode: 3
|
| 7 |
+
bias_mode: 0
|
| 8 |
+
center_crop: 1.0
|
| 9 |
+
clear_output_dir: true
|
| 10 |
+
compile: false
|
| 11 |
+
core: vit
|
| 12 |
+
core_lr_scale: 1
|
| 13 |
+
core_reg_scale: 0.5379
|
| 14 |
+
criterion: poisson
|
| 15 |
+
cropper_reg_scale: 0.0
|
| 16 |
+
dataset: data/sensorium
|
| 17 |
+
deterministic: false
|
| 18 |
+
device: cuda
|
| 19 |
+
disable_bias: false
|
| 20 |
+
disable_grid_predictor: false
|
| 21 |
+
dpi: 120
|
| 22 |
+
drop_path: 0.0
|
| 23 |
+
ds_name: sensorium
|
| 24 |
+
ds_scale: 1
|
| 25 |
+
emb_dim: 155
|
| 26 |
+
epochs: 400
|
| 27 |
+
format: svg
|
| 28 |
+
grad_checkpointing: 1
|
| 29 |
+
gray_scale: false
|
| 30 |
+
grid_predictor_dim: 2
|
| 31 |
+
input_shape: [1, 144, 256]
|
| 32 |
+
limit_data: null
|
| 33 |
+
lr: 0.001647
|
| 34 |
+
micro_batch_size: 16
|
| 35 |
+
mlp_dim: 488
|
| 36 |
+
mouse_ids: [S1, A, B, C, D, E]
|
| 37 |
+
num_blocks: 4
|
| 38 |
+
num_heads: 4
|
| 39 |
+
num_workers: 2
|
| 40 |
+
output_dir: runs/v1t_sensorium
|
| 41 |
+
output_shapes:
|
| 42 |
+
A: [8372]
|
| 43 |
+
B: [7344]
|
| 44 |
+
C: [7334]
|
| 45 |
+
D: [8107]
|
| 46 |
+
E: [8098]
|
| 47 |
+
S1: [7538]
|
| 48 |
+
p_dropout: 0.0229
|
| 49 |
+
patch_mode: 0
|
| 50 |
+
patch_size: 8
|
| 51 |
+
patch_stride: 1
|
| 52 |
+
pretrain_core: ''
|
| 53 |
+
readout: gaussian2d
|
| 54 |
+
readout_reg_scale: 0.0076
|
| 55 |
+
resize_image: 1
|
| 56 |
+
save_plots: false
|
| 57 |
+
seed: 1234
|
| 58 |
+
shift_mode: 2
|
| 59 |
+
shifter_reg_scale: 0.0
|
| 60 |
+
t_dropout: 0.2544
|
| 61 |
+
use_lsa: false
|
| 62 |
+
verbose: 0
|
v1t_sensorium/ckpt/model_state.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:56996a1e79c606d4d2ae9e7b24c2d1840d624e0868c98a742f01916bfc2a3255
|
| 3 |
+
size 120266389
|