Upload folder using huggingface_hub
Browse files- config.yaml +91 -0
- videomae.pt +3 -0
config.yaml
ADDED
|
@@ -0,0 +1,91 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
algorithm:
|
| 2 |
+
_target_: gengaze.algorithms.GRPO
|
| 3 |
+
group_size: 1
|
| 4 |
+
std_normalize: false
|
| 5 |
+
discount_factor: 1
|
| 6 |
+
has_loss_on_eos_tokens: false
|
| 7 |
+
optimize_task_loss_prediction: false
|
| 8 |
+
dataset:
|
| 9 |
+
_target_: gengaze.datasets.video_folder.VideoFolder
|
| 10 |
+
root: /home/baifengs/baifengs/data/gengaze/100DoH_res448_250K,/home/baifengs/baifengs/data/gengaze/Ego4D_res448_250K,/home/baifengs/baifengs/data/gengaze/InternVid_res448_250K,/home/baifengs/baifengs/data/gengaze/scanning_SAM_res448_50K,/home/baifengs/baifengs/data/gengaze/scanning_idl_res448_50K
|
| 11 |
+
clip_len: 16
|
| 12 |
+
frame_sample_rate: 1
|
| 13 |
+
gt_gazing_pos_paths:
|
| 14 |
+
train: null
|
| 15 |
+
val: null
|
| 16 |
+
random_sample_frame: false
|
| 17 |
+
model:
|
| 18 |
+
_target_: gengaze.models.video_random_gaze.VideoRandomGaze
|
| 19 |
+
gazing_ratio_config:
|
| 20 |
+
sample_strategy: exponential
|
| 21 |
+
fixed:
|
| 22 |
+
gazing_ratio: 0.5
|
| 23 |
+
uniform:
|
| 24 |
+
gazing_ratio_min: 0
|
| 25 |
+
gazing_ratio_max: 1
|
| 26 |
+
exponential:
|
| 27 |
+
gazing_ratio_min: 0.02
|
| 28 |
+
gazing_ratio_max: 0.15
|
| 29 |
+
lambda: 10
|
| 30 |
+
gazing_ratio_each_frame_config:
|
| 31 |
+
sample_strategy: dirichlet
|
| 32 |
+
dirichlet:
|
| 33 |
+
alpha: 10,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3
|
| 34 |
+
gazing_prob_each_scale_config:
|
| 35 |
+
sample_strategy: dirichlet
|
| 36 |
+
dirichlet:
|
| 37 |
+
alpha: 0.5
|
| 38 |
+
scales: 32+64+112+224
|
| 39 |
+
num_vision_tokens_each_frame: 265
|
| 40 |
+
frame_sampling_rate: 1
|
| 41 |
+
task:
|
| 42 |
+
_target_: gengaze.tasks.video_mae_reconstruction.VideoMAEReconstruction
|
| 43 |
+
recon_model: facebook/vit-mae-large
|
| 44 |
+
recon_sample_rate: 0.125
|
| 45 |
+
attn_mode: flash_attention_2
|
| 46 |
+
recon_model_config:
|
| 47 |
+
scale_embed: true
|
| 48 |
+
max_num_frames: 256
|
| 49 |
+
time_embed: true
|
| 50 |
+
causal: true
|
| 51 |
+
loss_type: l1+dinov2_reg+siglip2
|
| 52 |
+
loss_weights: 1+0.3+0.3
|
| 53 |
+
l1_loss_config: null
|
| 54 |
+
dinov2_reg_loss_config:
|
| 55 |
+
model: facebook/dinov2-with-registers-base
|
| 56 |
+
siglip2_loss_config:
|
| 57 |
+
model: google/siglip2-base-patch16-224
|
| 58 |
+
scales: 32+64+112+224
|
| 59 |
+
trainer:
|
| 60 |
+
_target_: gengaze.trainer.Trainer
|
| 61 |
+
batch_size: 512
|
| 62 |
+
per_gpu_max_batch_size: 8
|
| 63 |
+
lr: 0.0002
|
| 64 |
+
min_lr: 1.0e-05
|
| 65 |
+
lr_schedule: linear_w_warmup
|
| 66 |
+
optimizer: adam
|
| 67 |
+
train_gaze: false
|
| 68 |
+
train_task: true
|
| 69 |
+
train_w_ntp: false
|
| 70 |
+
val_nsteps: 2000
|
| 71 |
+
n_epochs: 50
|
| 72 |
+
logdir: exps/
|
| 73 |
+
exp_name_prefix: ''
|
| 74 |
+
exp_name: '250819_1751'
|
| 75 |
+
exp_name_suffix: ''
|
| 76 |
+
resume: auto
|
| 77 |
+
gaze_weights: null
|
| 78 |
+
task_weights: null
|
| 79 |
+
seed: 666
|
| 80 |
+
val_only: false
|
| 81 |
+
temp_schedule_args:
|
| 82 |
+
mode: exp
|
| 83 |
+
exp:
|
| 84 |
+
temp_start: 10000.0
|
| 85 |
+
temp_end: 1.0
|
| 86 |
+
neg_cosine:
|
| 87 |
+
temp_min: 1.0
|
| 88 |
+
temp_max: 10000.0
|
| 89 |
+
num_period: 1
|
| 90 |
+
val_args:
|
| 91 |
+
sample_gaze_for_reconstruction_oracle: 0
|
videomae.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ef08ef3468e0409460ad5505fd8208cd1cccb124691448aaf06ded24345f7e3e
|
| 3 |
+
size 2038058888
|