Hugging Face's logo

jonhpark
/
roco_smolvla

Safetensors
Model card Files
xet
 Community
jonhpark commited on
Commit
7835c46
·
verified ·
1 Parent(s): 3d2ce97

Add SmolVLA fine-tuned on roco_2 gearbox assembly

Browse files
Files changed (2) hide show
  1. README.md +70 -44
  2. train.log +219 -0
README.md CHANGED
@@ -1,63 +1,89 @@
1
- ---
2
- base_model: lerobot/smolvla_base
3
- datasets: yjsm1203/roco_2
4
- library_name: lerobot
5
- license: apache-2.0
6
- model_name: smolvla
7
- pipeline_tag: robotics
8
- tags:
9
- - lerobot
10
- - smolvla
11
- - robotics
12
- ---
13
 
14
- # Model Card for smolvla
15
 
16
- <!-- Provide a quick summary of what the model is/does. -->
17
-
18
-
19
- [SmolVLA](https://huggingface.co/papers/2506.01844) is a compact, efficient vision-language-action model that achieves competitive performance at reduced computational costs and can be deployed on consumer-grade hardware.
20
-
21
-
22
- This policy has been trained and pushed to the Hub using [LeRobot](https://github.com/huggingface/lerobot).
23
- See the full documentation at [LeRobot Docs](https://huggingface.co/docs/lerobot/index).
24
 
25
- ---
 
 
 
 
 
 
 
26
 
27
- ## How to Get Started with the Model
28
 
29
- For a complete walkthrough, see the [training guide](https://huggingface.co/docs/lerobot/il_robots#train-a-policy).
30
- Below is the short version on how to train and run inference/eval:
 
 
31
 
32
- ### Train from scratch
33
 
34
  ```bash
35
  lerobot-train \
36
- --dataset.repo_id=${HF_USER}/<dataset> \
37
- --policy.type=act \
38
- --output_dir=outputs/train/<desired_policy_repo_id> \
39
- --job_name=lerobot_training \
 
 
 
40
  --policy.device=cuda \
41
- --policy.repo_id=${HF_USER}/<desired_policy_repo_id>
42
- --wandb.enable=true
 
 
 
43
  ```
44
 
45
- _Writes checkpoints to `outputs/train/<desired_policy_repo_id>/checkpoints/`._
46
 
47
- ### Evaluate the policy/run inference
 
 
 
48
 
49
- ```bash
50
- lerobot-record \
51
- --robot.type=so100_follower \
52
- --dataset.repo_id=<hf_user>/eval_<dataset> \
53
- --policy.path=<hf_user>/<desired_policy_repo_id> \
54
- --episodes=10
 
 
 
 
 
 
 
 
 
 
 
55
  ```
56
 
57
- Prefix the dataset repo with **eval\_** and supply `--policy.path` pointing to a local or hub checkpoint.
58
 
59
- ---
 
 
 
 
 
 
 
60
 
61
- ## Model Details
 
 
 
 
 
 
 
 
62
 
63
- - **License:** apache-2.0
 
1
+ # SmolVLA Fine-tuned on roco_2 (Gearbox Assembly)
 
 
 
 
 
 
 
 
 
 
 
2
 
3
+ This model is a fine-tuned version of `lerobot/smolvla_base` on the `yjsm1203/roco_2` dataset for gearbox assembly task.
4
 
5
+ ## Model Details
 
 
 
 
 
 
 
6
 
7
+ | Item | Value |
8
+ |------|-------|
9
+ | **Base Model** | `lerobot/smolvla_base` |
10
+ | **Dataset** | `yjsm1203/roco_2` |
11
+ | **Task** | Gearbox Assembly |
12
+ | **Parameters** | ~450M (100M trainable) |
13
+ | **Training Steps** | 20,000 |
14
+ | **Final Loss** | 0.006 |
15
 
16
+ ## Training Configuration
17
 
18
+ - **Batch Size**: 64
19
+ - **Learning Rate**: 1e-4 (with warmup and cosine decay)
20
+ - **GPU**: RTX 5090 (32GB VRAM)
21
+ - **Training Time**: ~6.5 hours
22
 
23
+ ## Training Command
24
 
25
  ```bash
26
  lerobot-train \
27
+ --policy.path=lerobot/smolvla_base \
28
+ --dataset.repo_id=yjsm1203/roco_2 \
29
+ --dataset.root=./data/roco_2 \
30
+ --batch_size=64 \
31
+ --steps=20000 \
32
+ --output_dir=outputs/train/roco_smolvla \
33
+ --job_name=roco_smolvla_gearbox \
34
  --policy.device=cuda \
35
+ --save_freq=5000 \
36
+ --log_freq=100 \
37
+ --wandb.enable=false \
38
+ --policy.repo_id=jonhpark/roco_smolvla \
39
+ --rename_map='{"observation.images.head": "observation.images.camera1", "observation.images.left_hand": "observation.images.camera2", "observation.images.right_hand": "observation.images.camera3"}'
40
  ```
41
 
42
+ ## Task Description
43
 
44
+ ```
45
+ Assemble the gearbox by placing each small gear onto the tree pins one at a time.
46
+ Install the last small gear in the center, then put the cover on.
47
+ ```
48
 
49
+ ## Dataset Features
50
+
51
+ - **observation.images.head**: (3, 240, 320) - Head camera
52
+ - **observation.images.left_hand**: (3, 240, 320) - Left hand camera
53
+ - **observation.images.right_hand**: (3, 240, 320) - Right hand camera
54
+ - **observation.state**: (28,) - Robot state
55
+ - **action**: (14,) - Robot action
56
+
57
+ ## Usage
58
+
59
+ ```python
60
+ from lerobot.policies.smolvla.modeling_smolvla import SmolVLAPolicy
61
+
62
+ policy = SmolVLAPolicy.from_pretrained("jonhpark/roco_smolvla")
63
+ policy.to("cuda")
64
+
65
+ # Use with your robot or evaluation environment
66
  ```
67
 
68
+ ## Training Metrics
69
 
70
+ | Step | Loss | Gradient Norm |
71
+ |------|------|---------------|
72
+ | 100 | 1.149 | 4.229 |
73
+ | 1,000 | 0.067 | 0.565 |
74
+ | 5,000 | 0.019 | 0.334 |
75
+ | 10,000 | 0.010 | 0.279 |
76
+ | 15,000 | 0.006 | 0.231 |
77
+ | 20,000 | 0.006 | 0.185 |
78
 
79
+ ## Checkpoints
80
+
81
+ Available checkpoints in this repository:
82
+ - `checkpoints/005000/` - Step 5,000
83
+ - `checkpoints/010000/` - Step 10,000
84
+ - `checkpoints/015000/` - Step 15,000
85
+ - `checkpoints/020000/` - Step 20,000 (Final)
86
+
87
+ ## License
88
 
89
+ Apache 2.0
train.log ADDED
@@ -0,0 +1,219 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ INFO 2026-01-06 01:04:54 ot_train.py:282 Logs will be saved locally.
2
+ INFO 2026-01-06 01:04:54 ot_train.py:294 Creating dataset
3
+ INFO 2026-01-06 01:04:54 ot_train.py:313 Creating policy
4
+ Loading HuggingFaceTB/SmolVLM2-500M-Video-Instruct weights ...
5
+ `torch_dtype` is deprecated! Use `dtype` instead!
6
+ Reducing the number of VLM layers to 16 ...
7
+ INFO 2026-01-06 01:05:03 ot_train.py:366 Creating optimizer and scheduler
8
+ INFO 2026-01-06 01:05:03 hedulers.py:105 Auto-scaling LR scheduler: num_training_steps (20000) < num_decay_steps (30000). Scaling warmup: 1000 → 666, decay: 30000 → 20000 (scale factor: 0.667)
9
+ INFO 2026-01-06 01:05:03 ot_train.py:401 Output dir: outputs/train/roco_smolvla
10
+ INFO 2026-01-06 01:05:03 ot_train.py:408 cfg.steps=20000 (20K)
11
+ INFO 2026-01-06 01:05:03 ot_train.py:409 dataset.num_frames=112690 (113K)
12
+ INFO 2026-01-06 01:05:03 ot_train.py:410 dataset.num_episodes=191
13
+ INFO 2026-01-06 01:05:03 ot_train.py:413 Effective batch size: 64 x 1 = 64
14
+ INFO 2026-01-06 01:05:03 ot_train.py:414 num_learnable_params=99880992 (100M)
15
+ INFO 2026-01-06 01:05:03 ot_train.py:415 num_total_params=450046176 (450M)
16
+ INFO 2026-01-06 01:05:03 ot_train.py:471 Start offline training on a fixed dataset, with effective batch size: 64
17
+ INFO 2026-01-06 01:07:03 ot_train.py:501 step:100 smpl:6K ep:11 epch:0.06 loss:1.149 grdn:4.229 lr:7.7e-06 updt_s:1.191 data_s:0.012
18
+ INFO 2026-01-06 01:09:03 ot_train.py:501 step:200 smpl:13K ep:22 epch:0.11 loss:0.339 grdn:0.796 lr:2.3e-05 updt_s:1.187 data_s:0.005
19
+ INFO 2026-01-06 01:11:02 ot_train.py:501 step:300 smpl:19K ep:33 epch:0.17 loss:0.180 grdn:0.723 lr:3.8e-05 updt_s:1.188 data_s:0.005
20
+ INFO 2026-01-06 01:13:01 ot_train.py:501 step:400 smpl:26K ep:43 epch:0.23 loss:0.133 grdn:0.739 lr:5.3e-05 updt_s:1.187 data_s:0.005
21
+ INFO 2026-01-06 01:15:01 ot_train.py:501 step:500 smpl:32K ep:54 epch:0.28 loss:0.112 grdn:0.891 lr:6.8e-05 updt_s:1.187 data_s:0.005
22
+ INFO 2026-01-06 01:17:00 ot_train.py:501 step:600 smpl:38K ep:65 epch:0.34 loss:0.099 grdn:0.798 lr:8.3e-05 updt_s:1.187 data_s:0.005
23
+ INFO 2026-01-06 01:18:59 ot_train.py:501 step:700 smpl:45K ep:76 epch:0.40 loss:0.091 grdn:0.740 lr:9.7e-05 updt_s:1.187 data_s:0.005
24
+ INFO 2026-01-06 01:20:59 ot_train.py:501 step:800 smpl:51K ep:87 epch:0.45 loss:0.082 grdn:0.649 lr:1.0e-04 updt_s:1.188 data_s:0.005
25
+ INFO 2026-01-06 01:22:58 ot_train.py:501 step:900 smpl:58K ep:98 epch:0.51 loss:0.073 grdn:0.594 lr:1.0e-04 updt_s:1.187 data_s:0.005
26
+ INFO 2026-01-06 01:24:57 ot_train.py:501 step:1K smpl:64K ep:108 epch:0.57 loss:0.067 grdn:0.565 lr:9.9e-05 updt_s:1.187 data_s:0.005
27
+ INFO 2026-01-06 01:26:57 ot_train.py:501 step:1K smpl:70K ep:119 epch:0.62 loss:0.059 grdn:0.496 lr:9.9e-05 updt_s:1.187 data_s:0.005
28
+ INFO 2026-01-06 01:28:56 ot_train.py:501 step:1K smpl:77K ep:130 epch:0.68 loss:0.055 grdn:0.499 lr:9.9e-05 updt_s:1.187 data_s:0.005
29
+ INFO 2026-01-06 01:30:55 ot_train.py:501 step:1K smpl:83K ep:141 epch:0.74 loss:0.051 grdn:0.469 lr:9.9e-05 updt_s:1.187 data_s:0.005
30
+ INFO 2026-01-06 01:32:55 ot_train.py:501 step:1K smpl:90K ep:152 epch:0.80 loss:0.047 grdn:0.449 lr:9.9e-05 updt_s:1.187 data_s:0.005
31
+ INFO 2026-01-06 01:34:54 ot_train.py:501 step:2K smpl:96K ep:163 epch:0.85 loss:0.044 grdn:0.435 lr:9.9e-05 updt_s:1.187 data_s:0.005
32
+ INFO 2026-01-06 01:36:53 ot_train.py:501 step:2K smpl:102K ep:174 epch:0.91 loss:0.044 grdn:0.447 lr:9.9e-05 updt_s:1.188 data_s:0.005
33
+ INFO 2026-01-06 01:38:53 ot_train.py:501 step:2K smpl:109K ep:184 epch:0.97 loss:0.040 grdn:0.412 lr:9.8e-05 updt_s:1.187 data_s:0.005
34
+ INFO 2026-01-06 01:40:53 ot_train.py:501 step:2K smpl:115K ep:195 epch:1.02 loss:0.039 grdn:0.418 lr:9.8e-05 updt_s:1.186 data_s:0.011
35
+ INFO 2026-01-06 01:42:52 ot_train.py:501 step:2K smpl:122K ep:206 epch:1.08 loss:0.038 grdn:0.429 lr:9.8e-05 updt_s:1.188 data_s:0.005
36
+ INFO 2026-01-06 01:44:51 ot_train.py:501 step:2K smpl:128K ep:217 epch:1.14 loss:0.036 grdn:0.384 lr:9.8e-05 updt_s:1.187 data_s:0.005
37
+ INFO 2026-01-06 01:46:51 ot_train.py:501 step:2K smpl:134K ep:228 epch:1.19 loss:0.035 grdn:0.413 lr:9.7e-05 updt_s:1.187 data_s:0.005
38
+ INFO 2026-01-06 01:48:50 ot_train.py:501 step:2K smpl:141K ep:239 epch:1.25 loss:0.034 grdn:0.405 lr:9.7e-05 updt_s:1.187 data_s:0.005
39
+ INFO 2026-01-06 01:50:49 ot_train.py:501 step:2K smpl:147K ep:249 epch:1.31 loss:0.033 grdn:0.400 lr:9.7e-05 updt_s:1.187 data_s:0.005
40
+ INFO 2026-01-06 01:52:49 ot_train.py:501 step:2K smpl:154K ep:260 epch:1.36 loss:0.031 grdn:0.389 lr:9.7e-05 updt_s:1.187 data_s:0.005
41
+ INFO 2026-01-06 01:54:48 ot_train.py:501 step:2K smpl:160K ep:271 epch:1.42 loss:0.031 grdn:0.378 lr:9.6e-05 updt_s:1.188 data_s:0.005
42
+ INFO 2026-01-06 01:56:47 ot_train.py:501 step:3K smpl:166K ep:282 epch:1.48 loss:0.029 grdn:0.361 lr:9.6e-05 updt_s:1.187 data_s:0.005
43
+ INFO 2026-01-06 01:58:47 ot_train.py:501 step:3K smpl:173K ep:293 epch:1.53 loss:0.030 grdn:0.380 lr:9.6e-05 updt_s:1.187 data_s:0.005
44
+ INFO 2026-01-06 02:00:46 ot_train.py:501 step:3K smpl:179K ep:304 epch:1.59 loss:0.028 grdn:0.362 lr:9.6e-05 updt_s:1.187 data_s:0.005
45
+ INFO 2026-01-06 02:02:45 ot_train.py:501 step:3K smpl:186K ep:315 epch:1.65 loss:0.028 grdn:0.379 lr:9.5e-05 updt_s:1.187 data_s:0.005
46
+ INFO 2026-01-06 02:04:45 ot_train.py:501 step:3K smpl:192K ep:325 epch:1.70 loss:0.029 grdn:0.397 lr:9.5e-05 updt_s:1.187 data_s:0.005
47
+ INFO 2026-01-06 02:06:44 ot_train.py:501 step:3K smpl:198K ep:336 epch:1.76 loss:0.026 grdn:0.364 lr:9.5e-05 updt_s:1.187 data_s:0.005
48
+ INFO 2026-01-06 02:08:43 ot_train.py:501 step:3K smpl:205K ep:347 epch:1.82 loss:0.025 grdn:0.341 lr:9.4e-05 updt_s:1.187 data_s:0.005
49
+ INFO 2026-01-06 02:10:43 ot_train.py:501 step:3K smpl:211K ep:358 epch:1.87 loss:0.027 grdn:0.371 lr:9.4e-05 updt_s:1.187 data_s:0.005
50
+ INFO 2026-01-06 02:12:42 ot_train.py:501 step:3K smpl:218K ep:369 epch:1.93 loss:0.025 grdn:0.335 lr:9.3e-05 updt_s:1.187 data_s:0.005
51
+ INFO 2026-01-06 02:14:41 ot_train.py:501 step:4K smpl:224K ep:380 epch:1.99 loss:0.026 grdn:0.367 lr:9.3e-05 updt_s:1.187 data_s:0.005
52
+ INFO 2026-01-06 02:16:41 ot_train.py:501 step:4K smpl:230K ep:391 epch:2.04 loss:0.025 grdn:0.346 lr:9.3e-05 updt_s:1.185 data_s:0.012
53
+ INFO 2026-01-06 02:18:40 ot_train.py:501 step:4K smpl:237K ep:401 epch:2.10 loss:0.024 grdn:0.347 lr:9.2e-05 updt_s:1.187 data_s:0.005
54
+ INFO 2026-01-06 02:20:40 ot_train.py:501 step:4K smpl:243K ep:412 epch:2.16 loss:0.022 grdn:0.307 lr:9.2e-05 updt_s:1.187 data_s:0.005
55
+ INFO 2026-01-06 02:22:39 ot_train.py:501 step:4K smpl:250K ep:423 epch:2.21 loss:0.023 grdn:0.362 lr:9.1e-05 updt_s:1.187 data_s:0.005
56
+ INFO 2026-01-06 02:24:38 ot_train.py:501 step:4K smpl:256K ep:434 epch:2.27 loss:0.022 grdn:0.361 lr:9.1e-05 updt_s:1.187 data_s:0.005
57
+ INFO 2026-01-06 02:26:38 ot_train.py:501 step:4K smpl:262K ep:445 epch:2.33 loss:0.022 grdn:0.343 lr:9.0e-05 updt_s:1.187 data_s:0.005
58
+ INFO 2026-01-06 02:28:37 ot_train.py:501 step:4K smpl:269K ep:456 epch:2.39 loss:0.022 grdn:0.337 lr:9.0e-05 updt_s:1.187 data_s:0.005
59
+ INFO 2026-01-06 02:30:36 ot_train.py:501 step:4K smpl:275K ep:466 epch:2.44 loss:0.022 grdn:0.344 lr:9.0e-05 updt_s:1.187 data_s:0.005
60
+ INFO 2026-01-06 02:32:35 ot_train.py:501 step:4K smpl:282K ep:477 epch:2.50 loss:0.020 grdn:0.348 lr:8.9e-05 updt_s:1.187 data_s:0.005
61
+ INFO 2026-01-06 02:34:35 ot_train.py:501 step:4K smpl:288K ep:488 epch:2.56 loss:0.022 grdn:0.352 lr:8.9e-05 updt_s:1.187 data_s:0.005
62
+ INFO 2026-01-06 02:36:34 ot_train.py:501 step:5K smpl:294K ep:499 epch:2.61 loss:0.021 grdn:0.337 lr:8.8e-05 updt_s:1.187 data_s:0.005
63
+ INFO 2026-01-06 02:38:33 ot_train.py:501 step:5K smpl:301K ep:510 epch:2.67 loss:0.021 grdn:0.311 lr:8.8e-05 updt_s:1.187 data_s:0.005
64
+ INFO 2026-01-06 02:40:33 ot_train.py:501 step:5K smpl:307K ep:521 epch:2.73 loss:0.019 grdn:0.336 lr:8.7e-05 updt_s:1.187 data_s:0.005
65
+ INFO 2026-01-06 02:42:32 ot_train.py:501 step:5K smpl:314K ep:532 epch:2.78 loss:0.020 grdn:0.331 lr:8.7e-05 updt_s:1.187 data_s:0.005
66
+ INFO 2026-01-06 02:44:31 ot_train.py:501 step:5K smpl:320K ep:542 epch:2.84 loss:0.019 grdn:0.334 lr:8.6e-05 updt_s:1.187 data_s:0.005
67
+ INFO 2026-01-06 02:44:31 ot_train.py:521 Checkpoint policy after step 5000
68
+ INFO 2026-01-06 02:46:31 ot_train.py:501 step:5K smpl:326K ep:553 epch:2.90 loss:0.019 grdn:0.328 lr:8.5e-05 updt_s:1.187 data_s:0.005
69
+ INFO 2026-01-06 02:48:31 ot_train.py:501 step:5K smpl:333K ep:564 epch:2.95 loss:0.019 grdn:0.332 lr:8.5e-05 updt_s:1.187 data_s:0.005
70
+ INFO 2026-01-06 02:50:30 ot_train.py:501 step:5K smpl:339K ep:575 epch:3.01 loss:0.019 grdn:0.332 lr:8.4e-05 updt_s:1.185 data_s:0.012
71
+ INFO 2026-01-06 02:52:30 ot_train.py:501 step:5K smpl:346K ep:586 epch:3.07 loss:0.019 grdn:0.332 lr:8.4e-05 updt_s:1.187 data_s:0.005
72
+ INFO 2026-01-06 02:54:29 ot_train.py:501 step:6K smpl:352K ep:597 epch:3.12 loss:0.019 grdn:0.328 lr:8.3e-05 updt_s:1.187 data_s:0.005
73
+ INFO 2026-01-06 02:56:28 ot_train.py:501 step:6K smpl:358K ep:607 epch:3.18 loss:0.018 grdn:0.333 lr:8.3e-05 updt_s:1.187 data_s:0.005
74
+ INFO 2026-01-06 02:58:28 ot_train.py:501 step:6K smpl:365K ep:618 epch:3.24 loss:0.017 grdn:0.312 lr:8.2e-05 updt_s:1.187 data_s:0.005
75
+ INFO 2026-01-06 03:00:27 ot_train.py:501 step:6K smpl:371K ep:629 epch:3.29 loss:0.016 grdn:0.313 lr:8.1e-05 updt_s:1.187 data_s:0.005
76
+ INFO 2026-01-06 03:02:26 ot_train.py:501 step:6K smpl:378K ep:640 epch:3.35 loss:0.016 grdn:0.301 lr:8.1e-05 updt_s:1.187 data_s:0.005
77
+ INFO 2026-01-06 03:04:26 ot_train.py:501 step:6K smpl:384K ep:651 epch:3.41 loss:0.017 grdn:0.313 lr:8.0e-05 updt_s:1.187 data_s:0.005
78
+ INFO 2026-01-06 03:06:25 ot_train.py:501 step:6K smpl:390K ep:662 epch:3.46 loss:0.016 grdn:0.317 lr:8.0e-05 updt_s:1.187 data_s:0.005
79
+ INFO 2026-01-06 03:08:24 ot_train.py:501 step:6K smpl:397K ep:673 epch:3.52 loss:0.016 grdn:0.321 lr:7.9e-05 updt_s:1.187 data_s:0.005
80
+ INFO 2026-01-06 03:10:24 ot_train.py:501 step:6K smpl:403K ep:683 epch:3.58 loss:0.015 grdn:0.293 lr:7.8e-05 updt_s:1.187 data_s:0.005
81
+ INFO 2026-01-06 03:12:23 ot_train.py:501 step:6K smpl:410K ep:694 epch:3.63 loss:0.016 grdn:0.312 lr:7.8e-05 updt_s:1.187 data_s:0.005
82
+ INFO 2026-01-06 03:14:22 ot_train.py:501 step:6K smpl:416K ep:705 epch:3.69 loss:0.015 grdn:0.318 lr:7.7e-05 updt_s:1.187 data_s:0.005
83
+ INFO 2026-01-06 03:16:22 ot_train.py:501 step:7K smpl:422K ep:716 epch:3.75 loss:0.016 grdn:0.304 lr:7.6e-05 updt_s:1.187 data_s:0.005
84
+ INFO 2026-01-06 03:18:21 ot_train.py:501 step:7K smpl:429K ep:727 epch:3.81 loss:0.015 grdn:0.326 lr:7.6e-05 updt_s:1.187 data_s:0.005
85
+ INFO 2026-01-06 03:20:20 ot_train.py:501 step:7K smpl:435K ep:738 epch:3.86 loss:0.014 grdn:0.322 lr:7.5e-05 updt_s:1.187 data_s:0.005
86
+ INFO 2026-01-06 03:22:20 ot_train.py:501 step:7K smpl:442K ep:748 epch:3.92 loss:0.014 grdn:0.300 lr:7.4e-05 updt_s:1.187 data_s:0.005
87
+ INFO 2026-01-06 03:24:19 ot_train.py:501 step:7K smpl:448K ep:759 epch:3.98 loss:0.014 grdn:0.303 lr:7.4e-05 updt_s:1.187 data_s:0.005
88
+ INFO 2026-01-06 03:26:19 ot_train.py:501 step:7K smpl:454K ep:770 epch:4.03 loss:0.016 grdn:0.331 lr:7.3e-05 updt_s:1.185 data_s:0.012
89
+ INFO 2026-01-06 03:28:18 ot_train.py:501 step:7K smpl:461K ep:781 epch:4.09 loss:0.015 grdn:0.322 lr:7.2e-05 updt_s:1.187 data_s:0.005
90
+ INFO 2026-01-06 03:30:17 ot_train.py:501 step:7K smpl:467K ep:792 epch:4.15 loss:0.014 grdn:0.286 lr:7.2e-05 updt_s:1.187 data_s:0.005
91
+ INFO 2026-01-06 03:32:17 ot_train.py:501 step:7K smpl:474K ep:803 epch:4.20 loss:0.013 grdn:0.286 lr:7.1e-05 updt_s:1.187 data_s:0.005
92
+ INFO 2026-01-06 03:34:16 ot_train.py:501 step:8K smpl:480K ep:814 epch:4.26 loss:0.015 grdn:0.357 lr:7.0e-05 updt_s:1.187 data_s:0.005
93
+ INFO 2026-01-06 03:36:15 ot_train.py:501 step:8K smpl:486K ep:824 epch:4.32 loss:0.013 grdn:0.302 lr:7.0e-05 updt_s:1.187 data_s:0.005
94
+ INFO 2026-01-06 03:38:15 ot_train.py:501 step:8K smpl:493K ep:835 epch:4.37 loss:0.013 grdn:0.305 lr:6.9e-05 updt_s:1.187 data_s:0.005
95
+ INFO 2026-01-06 03:40:14 ot_train.py:501 step:8K smpl:499K ep:846 epch:4.43 loss:0.013 grdn:0.302 lr:6.8e-05 updt_s:1.187 data_s:0.005
96
+ INFO 2026-01-06 03:42:13 ot_train.py:501 step:8K smpl:506K ep:857 epch:4.49 loss:0.012 grdn:0.278 lr:6.7e-05 updt_s:1.187 data_s:0.005
97
+ INFO 2026-01-06 03:44:13 ot_train.py:501 step:8K smpl:512K ep:868 epch:4.54 loss:0.013 grdn:0.301 lr:6.7e-05 updt_s:1.187 data_s:0.005
98
+ INFO 2026-01-06 03:46:12 ot_train.py:501 step:8K smpl:518K ep:879 epch:4.60 loss:0.013 grdn:0.294 lr:6.6e-05 updt_s:1.187 data_s:0.005
99
+ INFO 2026-01-06 03:48:11 ot_train.py:501 step:8K smpl:525K ep:889 epch:4.66 loss:0.013 grdn:0.296 lr:6.5e-05 updt_s:1.187 data_s:0.005
100
+ INFO 2026-01-06 03:50:11 ot_train.py:501 step:8K smpl:531K ep:900 epch:4.71 loss:0.012 grdn:0.291 lr:6.4e-05 updt_s:1.187 data_s:0.005
101
+ INFO 2026-01-06 03:52:10 ot_train.py:501 step:8K smpl:538K ep:911 epch:4.77 loss:0.012 grdn:0.293 lr:6.4e-05 updt_s:1.187 data_s:0.005
102
+ INFO 2026-01-06 03:54:09 ot_train.py:501 step:8K smpl:544K ep:922 epch:4.83 loss:0.011 grdn:0.265 lr:6.3e-05 updt_s:1.188 data_s:0.005
103
+ INFO 2026-01-06 03:56:09 ot_train.py:501 step:9K smpl:550K ep:933 epch:4.88 loss:0.012 grdn:0.290 lr:6.2e-05 updt_s:1.187 data_s:0.005
104
+ INFO 2026-01-06 03:58:08 ot_train.py:501 step:9K smpl:557K ep:944 epch:4.94 loss:0.012 grdn:0.279 lr:6.2e-05 updt_s:1.187 data_s:0.005
105
+ INFO 2026-01-06 04:00:07 ot_train.py:501 step:9K smpl:563K ep:955 epch:5.00 loss:0.011 grdn:0.306 lr:6.1e-05 updt_s:1.187 data_s:0.005
106
+ INFO 2026-01-06 04:02:07 ot_train.py:501 step:9K smpl:570K ep:965 epch:5.05 loss:0.011 grdn:0.290 lr:6.0e-05 updt_s:1.185 data_s:0.012
107
+ INFO 2026-01-06 04:04:06 ot_train.py:501 step:9K smpl:576K ep:976 epch:5.11 loss:0.011 grdn:0.276 lr:5.9e-05 updt_s:1.187 data_s:0.005
108
+ INFO 2026-01-06 04:06:06 ot_train.py:501 step:9K smpl:582K ep:987 epch:5.17 loss:0.012 grdn:0.299 lr:5.8e-05 updt_s:1.187 data_s:0.005
109
+ INFO 2026-01-06 04:08:05 ot_train.py:501 step:9K smpl:589K ep:998 epch:5.22 loss:0.010 grdn:0.254 lr:5.8e-05 updt_s:1.187 data_s:0.005
110
+ INFO 2026-01-06 04:10:04 ot_train.py:501 step:9K smpl:595K ep:1K epch:5.28 loss:0.011 grdn:0.312 lr:5.7e-05 updt_s:1.187 data_s:0.005
111
+ INFO 2026-01-06 04:12:04 ot_train.py:501 step:9K smpl:602K ep:1K epch:5.34 loss:0.010 grdn:0.274 lr:5.6e-05 updt_s:1.187 data_s:0.005
112
+ INFO 2026-01-06 04:14:03 ot_train.py:501 step:10K smpl:608K ep:1K epch:5.40 loss:0.010 grdn:0.310 lr:5.5e-05 updt_s:1.187 data_s:0.005
113
+ INFO 2026-01-06 04:16:02 ot_train.py:501 step:10K smpl:614K ep:1K epch:5.45 loss:0.011 grdn:0.293 lr:5.5e-05 updt_s:1.187 data_s:0.005
114
+ INFO 2026-01-06 04:18:02 ot_train.py:501 step:10K smpl:621K ep:1K epch:5.51 loss:0.010 grdn:0.286 lr:5.4e-05 updt_s:1.187 data_s:0.005
115
+ INFO 2026-01-06 04:20:01 ot_train.py:501 step:10K smpl:627K ep:1K epch:5.57 loss:0.010 grdn:0.256 lr:5.3e-05 updt_s:1.187 data_s:0.005
116
+ INFO 2026-01-06 04:22:00 ot_train.py:501 step:10K smpl:634K ep:1K epch:5.62 loss:0.010 grdn:0.270 lr:5.2e-05 updt_s:1.187 data_s:0.005
117
+ INFO 2026-01-06 04:24:00 ot_train.py:501 step:10K smpl:640K ep:1K epch:5.68 loss:0.010 grdn:0.279 lr:5.2e-05 updt_s:1.187 data_s:0.005
118
+ INFO 2026-01-06 04:24:00 ot_train.py:521 Checkpoint policy after step 10000
119
+ INFO 2026-01-06 04:26:00 ot_train.py:501 step:10K smpl:646K ep:1K epch:5.74 loss:0.010 grdn:0.256 lr:5.1e-05 updt_s:1.187 data_s:0.005
120
+ INFO 2026-01-06 04:27:59 ot_train.py:501 step:10K smpl:653K ep:1K epch:5.79 loss:0.010 grdn:0.262 lr:5.0e-05 updt_s:1.187 data_s:0.005
121
+ INFO 2026-01-06 04:29:58 ot_train.py:501 step:10K smpl:659K ep:1K epch:5.85 loss:0.010 grdn:0.248 lr:4.9e-05 updt_s:1.187 data_s:0.005
122
+ INFO 2026-01-06 04:31:58 ot_train.py:501 step:10K smpl:666K ep:1K epch:5.91 loss:0.009 grdn:0.279 lr:4.9e-05 updt_s:1.187 data_s:0.005
123
+ INFO 2026-01-06 04:33:57 ot_train.py:501 step:10K smpl:672K ep:1K epch:5.96 loss:0.010 grdn:0.258 lr:4.8e-05 updt_s:1.187 data_s:0.005
124
+ INFO 2026-01-06 04:35:57 ot_train.py:501 step:11K smpl:678K ep:1K epch:6.02 loss:0.010 grdn:0.298 lr:4.7e-05 updt_s:1.185 data_s:0.011
125
+ INFO 2026-01-06 04:37:56 ot_train.py:501 step:11K smpl:685K ep:1K epch:6.08 loss:0.009 grdn:0.241 lr:4.6e-05 updt_s:1.187 data_s:0.005
126
+ INFO 2026-01-06 04:39:55 ot_train.py:501 step:11K smpl:691K ep:1K epch:6.13 loss:0.009 grdn:0.307 lr:4.6e-05 updt_s:1.187 data_s:0.005
127
+ INFO 2026-01-06 04:41:55 ot_train.py:501 step:11K smpl:698K ep:1K epch:6.19 loss:0.009 grdn:0.267 lr:4.5e-05 updt_s:1.188 data_s:0.005
128
+ INFO 2026-01-06 04:43:54 ot_train.py:501 step:11K smpl:704K ep:1K epch:6.25 loss:0.009 grdn:0.246 lr:4.4e-05 updt_s:1.187 data_s:0.005
129
+ INFO 2026-01-06 04:45:53 ot_train.py:501 step:11K smpl:710K ep:1K epch:6.30 loss:0.009 grdn:0.275 lr:4.3e-05 updt_s:1.187 data_s:0.005
130
+ INFO 2026-01-06 04:47:53 ot_train.py:501 step:11K smpl:717K ep:1K epch:6.36 loss:0.009 grdn:0.240 lr:4.2e-05 updt_s:1.187 data_s:0.005
131
+ INFO 2026-01-06 04:49:52 ot_train.py:501 step:11K smpl:723K ep:1K epch:6.42 loss:0.009 grdn:0.315 lr:4.2e-05 updt_s:1.187 data_s:0.005
132
+ INFO 2026-01-06 04:51:51 ot_train.py:501 step:11K smpl:730K ep:1K epch:6.47 loss:0.009 grdn:0.230 lr:4.1e-05 updt_s:1.187 data_s:0.005
133
+ INFO 2026-01-06 04:53:51 ot_train.py:501 step:12K smpl:736K ep:1K epch:6.53 loss:0.009 grdn:0.254 lr:4.0e-05 updt_s:1.187 data_s:0.005
134
+ INFO 2026-01-06 04:55:50 ot_train.py:501 step:12K smpl:742K ep:1K epch:6.59 loss:0.008 grdn:0.286 lr:3.9e-05 updt_s:1.187 data_s:0.005
135
+ INFO 2026-01-06 04:57:49 ot_train.py:501 step:12K smpl:749K ep:1K epch:6.64 loss:0.008 grdn:0.247 lr:3.9e-05 updt_s:1.187 data_s:0.005
136
+ INFO 2026-01-06 04:59:49 ot_train.py:501 step:12K smpl:755K ep:1K epch:6.70 loss:0.008 grdn:0.248 lr:3.8e-05 updt_s:1.187 data_s:0.005
137
+ INFO 2026-01-06 05:01:48 ot_train.py:501 step:12K smpl:762K ep:1K epch:6.76 loss:0.009 grdn:0.257 lr:3.7e-05 updt_s:1.187 data_s:0.005
138
+ INFO 2026-01-06 05:03:47 ot_train.py:501 step:12K smpl:768K ep:1K epch:6.82 loss:0.008 grdn:0.274 lr:3.7e-05 updt_s:1.187 data_s:0.005
139
+ INFO 2026-01-06 05:05:47 ot_train.py:501 step:12K smpl:774K ep:1K epch:6.87 loss:0.008 grdn:0.243 lr:3.6e-05 updt_s:1.187 data_s:0.005
140
+ INFO 2026-01-06 05:07:46 ot_train.py:501 step:12K smpl:781K ep:1K epch:6.93 loss:0.008 grdn:0.241 lr:3.5e-05 updt_s:1.187 data_s:0.005
141
+ INFO 2026-01-06 05:09:45 ot_train.py:501 step:12K smpl:787K ep:1K epch:6.99 loss:0.008 grdn:0.242 lr:3.4e-05 updt_s:1.188 data_s:0.005
142
+ INFO 2026-01-06 05:11:45 ot_train.py:501 step:12K smpl:794K ep:1K epch:7.04 loss:0.008 grdn:0.256 lr:3.4e-05 updt_s:1.185 data_s:0.012
143
+ INFO 2026-01-06 05:13:44 ot_train.py:501 step:12K smpl:800K ep:1K epch:7.10 loss:0.008 grdn:0.233 lr:3.3e-05 updt_s:1.187 data_s:0.005
144
+ INFO 2026-01-06 05:15:44 ot_train.py:501 step:13K smpl:806K ep:1K epch:7.16 loss:0.008 grdn:0.253 lr:3.2e-05 updt_s:1.187 data_s:0.005
145
+ INFO 2026-01-06 05:17:43 ot_train.py:501 step:13K smpl:813K ep:1K epch:7.21 loss:0.008 grdn:0.247 lr:3.2e-05 updt_s:1.187 data_s:0.005
146
+ INFO 2026-01-06 05:19:42 ot_train.py:501 step:13K smpl:819K ep:1K epch:7.27 loss:0.008 grdn:0.235 lr:3.1e-05 updt_s:1.187 data_s:0.005
147
+ INFO 2026-01-06 05:21:42 ot_train.py:501 step:13K smpl:826K ep:1K epch:7.33 loss:0.007 grdn:0.235 lr:3.0e-05 updt_s:1.187 data_s:0.005
148
+ INFO 2026-01-06 05:23:41 ot_train.py:501 step:13K smpl:832K ep:1K epch:7.38 loss:0.008 grdn:0.257 lr:2.9e-05 updt_s:1.187 data_s:0.005
149
+ INFO 2026-01-06 05:25:40 ot_train.py:501 step:13K smpl:838K ep:1K epch:7.44 loss:0.007 grdn:0.246 lr:2.9e-05 updt_s:1.187 data_s:0.005
150
+ INFO 2026-01-06 05:27:40 ot_train.py:501 step:13K smpl:845K ep:1K epch:7.50 loss:0.007 grdn:0.238 lr:2.8e-05 updt_s:1.187 data_s:0.005
151
+ INFO 2026-01-06 05:29:39 ot_train.py:501 step:13K smpl:851K ep:1K epch:7.55 loss:0.007 grdn:0.251 lr:2.7e-05 updt_s:1.187 data_s:0.005
152
+ INFO 2026-01-06 05:31:38 ot_train.py:501 step:13K smpl:858K ep:1K epch:7.61 loss:0.007 grdn:0.265 lr:2.7e-05 updt_s:1.187 data_s:0.005
153
+ INFO 2026-01-06 05:33:38 ot_train.py:501 step:14K smpl:864K ep:1K epch:7.67 loss:0.007 grdn:0.229 lr:2.6e-05 updt_s:1.187 data_s:0.005
154
+ INFO 2026-01-06 05:35:37 ot_train.py:501 step:14K smpl:870K ep:1K epch:7.72 loss:0.007 grdn:0.236 lr:2.5e-05 updt_s:1.187 data_s:0.005
155
+ INFO 2026-01-06 05:37:36 ot_train.py:501 step:14K smpl:877K ep:1K epch:7.78 loss:0.007 grdn:0.221 lr:2.5e-05 updt_s:1.187 data_s:0.005
156
+ INFO 2026-01-06 05:39:36 ot_train.py:501 step:14K smpl:883K ep:1K epch:7.84 loss:0.007 grdn:0.264 lr:2.4e-05 updt_s:1.187 data_s:0.005
157
+ INFO 2026-01-06 05:41:35 ot_train.py:501 step:14K smpl:890K ep:2K epch:7.89 loss:0.007 grdn:0.211 lr:2.4e-05 updt_s:1.187 data_s:0.005
158
+ INFO 2026-01-06 05:43:34 ot_train.py:501 step:14K smpl:896K ep:2K epch:7.95 loss:0.007 grdn:0.227 lr:2.3e-05 updt_s:1.187 data_s:0.005
159
+ INFO 2026-01-06 05:45:34 ot_train.py:501 step:14K smpl:902K ep:2K epch:8.01 loss:0.007 grdn:0.237 lr:2.2e-05 updt_s:1.185 data_s:0.012
160
+ INFO 2026-01-06 05:47:33 ot_train.py:501 step:14K smpl:909K ep:2K epch:8.06 loss:0.007 grdn:0.222 lr:2.2e-05 updt_s:1.187 data_s:0.005
161
+ INFO 2026-01-06 05:49:33 ot_train.py:501 step:14K smpl:915K ep:2K epch:8.12 loss:0.007 grdn:0.249 lr:2.1e-05 updt_s:1.187 data_s:0.005
162
+ INFO 2026-01-06 05:51:32 ot_train.py:501 step:14K smpl:922K ep:2K epch:8.18 loss:0.007 grdn:0.227 lr:2.0e-05 updt_s:1.187 data_s:0.005
163
+ INFO 2026-01-06 05:53:31 ot_train.py:501 step:14K smpl:928K ep:2K epch:8.23 loss:0.007 grdn:0.242 lr:2.0e-05 updt_s:1.187 data_s:0.005
164
+ INFO 2026-01-06 05:55:31 ot_train.py:501 step:15K smpl:934K ep:2K epch:8.29 loss:0.007 grdn:0.229 lr:1.9e-05 updt_s:1.187 data_s:0.005
165
+ INFO 2026-01-06 05:57:30 ot_train.py:501 step:15K smpl:941K ep:2K epch:8.35 loss:0.007 grdn:0.228 lr:1.9e-05 updt_s:1.187 data_s:0.005
166
+ INFO 2026-01-06 05:59:29 ot_train.py:501 step:15K smpl:947K ep:2K epch:8.41 loss:0.007 grdn:0.220 lr:1.8e-05 updt_s:1.187 data_s:0.005
167
+ INFO 2026-01-06 06:01:29 ot_train.py:501 step:15K smpl:954K ep:2K epch:8.46 loss:0.006 grdn:0.238 lr:1.8e-05 updt_s:1.187 data_s:0.005
168
+ INFO 2026-01-06 06:03:28 ot_train.py:501 step:15K smpl:960K ep:2K epch:8.52 loss:0.006 grdn:0.231 lr:1.7e-05 updt_s:1.187 data_s:0.005
169
+ INFO 2026-01-06 06:03:28 ot_train.py:521 Checkpoint policy after step 15000
170
+ INFO 2026-01-06 06:05:28 ot_train.py:501 step:15K smpl:966K ep:2K epch:8.58 loss:0.007 grdn:0.209 lr:1.7e-05 updt_s:1.187 data_s:0.005
171
+ INFO 2026-01-06 06:07:27 ot_train.py:501 step:15K smpl:973K ep:2K epch:8.63 loss:0.006 grdn:0.215 lr:1.6e-05 updt_s:1.187 data_s:0.005
172
+ INFO 2026-01-06 06:09:27 ot_train.py:501 step:15K smpl:979K ep:2K epch:8.69 loss:0.007 grdn:0.228 lr:1.5e-05 updt_s:1.187 data_s:0.005
173
+ INFO 2026-01-06 06:11:26 ot_train.py:501 step:15K smpl:986K ep:2K epch:8.75 loss:0.006 grdn:0.191 lr:1.5e-05 updt_s:1.187 data_s:0.005
174
+ INFO 2026-01-06 06:13:25 ot_train.py:501 step:16K smpl:992K ep:2K epch:8.80 loss:0.006 grdn:0.210 lr:1.4e-05 updt_s:1.187 data_s:0.005
175
+ INFO 2026-01-06 06:15:25 ot_train.py:501 step:16K smpl:998K ep:2K epch:8.86 loss:0.006 grdn:0.204 lr:1.4e-05 updt_s:1.187 data_s:0.005
176
+ INFO 2026-01-06 06:17:24 ot_train.py:501 step:16K smpl:1M ep:2K epch:8.92 loss:0.007 grdn:0.223 lr:1.3e-05 updt_s:1.187 data_s:0.005
177
+ INFO 2026-01-06 06:19:23 ot_train.py:501 step:16K smpl:1M ep:2K epch:8.97 loss:0.007 grdn:0.213 lr:1.3e-05 updt_s:1.187 data_s:0.005
178
+ INFO 2026-01-06 06:21:23 ot_train.py:501 step:16K smpl:1M ep:2K epch:9.03 loss:0.006 grdn:0.227 lr:1.2e-05 updt_s:1.185 data_s:0.012
179
+ INFO 2026-01-06 06:23:22 ot_train.py:501 step:16K smpl:1M ep:2K epch:9.09 loss:0.006 grdn:0.210 lr:1.2e-05 updt_s:1.187 data_s:0.005
180
+ INFO 2026-01-06 06:25:22 ot_train.py:501 step:16K smpl:1M ep:2K epch:9.14 loss:0.006 grdn:0.210 lr:1.2e-05 updt_s:1.187 data_s:0.005
181
+ INFO 2026-01-06 06:27:21 ot_train.py:501 step:16K smpl:1M ep:2K epch:9.20 loss:0.006 grdn:0.209 lr:1.1e-05 updt_s:1.187 data_s:0.005
182
+ INFO 2026-01-06 06:29:21 ot_train.py:501 step:16K smpl:1M ep:2K epch:9.26 loss:0.006 grdn:0.202 lr:1.1e-05 updt_s:1.187 data_s:0.005
183
+ INFO 2026-01-06 06:31:20 ot_train.py:501 step:16K smpl:1M ep:2K epch:9.31 loss:0.006 grdn:0.206 lr:1.0e-05 updt_s:1.187 data_s:0.005
184
+ INFO 2026-01-06 06:33:19 ot_train.py:501 step:16K smpl:1M ep:2K epch:9.37 loss:0.006 grdn:0.206 lr:9.9e-06 updt_s:1.187 data_s:0.005
185
+ INFO 2026-01-06 06:35:18 ot_train.py:501 step:17K smpl:1M ep:2K epch:9.43 loss:0.006 grdn:0.205 lr:9.5e-06 updt_s:1.187 data_s:0.005
186
+ INFO 2026-01-06 06:37:18 ot_train.py:501 step:17K smpl:1M ep:2K epch:9.48 loss:0.006 grdn:0.188 lr:9.1e-06 updt_s:1.187 data_s:0.005
187
+ INFO 2026-01-06 06:39:17 ot_train.py:501 step:17K smpl:1M ep:2K epch:9.54 loss:0.006 grdn:0.204 lr:8.7e-06 updt_s:1.187 data_s:0.005
188
+ INFO 2026-01-06 06:41:16 ot_train.py:501 step:17K smpl:1M ep:2K epch:9.60 loss:0.006 grdn:0.184 lr:8.3e-06 updt_s:1.187 data_s:0.005
189
+ INFO 2026-01-06 06:43:16 ot_train.py:501 step:17K smpl:1M ep:2K epch:9.65 loss:0.006 grdn:0.215 lr:8.0e-06 updt_s:1.187 data_s:0.005
190
+ INFO 2026-01-06 06:45:15 ot_train.py:501 step:17K smpl:1M ep:2K epch:9.71 loss:0.006 grdn:0.199 lr:7.6e-06 updt_s:1.187 data_s:0.005
191
+ INFO 2026-01-06 06:47:14 ot_train.py:501 step:17K smpl:1M ep:2K epch:9.77 loss:0.006 grdn:0.196 lr:7.3e-06 updt_s:1.187 data_s:0.005
192
+ INFO 2026-01-06 06:49:14 ot_train.py:501 step:17K smpl:1M ep:2K epch:9.83 loss:0.006 grdn:0.187 lr:7.0e-06 updt_s:1.187 data_s:0.005
193
+ INFO 2026-01-06 06:51:13 ot_train.py:501 step:17K smpl:1M ep:2K epch:9.88 loss:0.006 grdn:0.183 lr:6.7e-06 updt_s:1.187 data_s:0.005
194
+ INFO 2026-01-06 06:53:12 ot_train.py:501 step:18K smpl:1M ep:2K epch:9.94 loss:0.006 grdn:0.202 lr:6.4e-06 updt_s:1.187 data_s:0.005
195
+ INFO 2026-01-06 06:55:12 ot_train.py:501 step:18K smpl:1M ep:2K epch:10.00 loss:0.006 grdn:0.212 lr:6.1e-06 updt_s:1.187 data_s:0.005
196
+ INFO 2026-01-06 06:57:11 ot_train.py:501 step:18K smpl:1M ep:2K epch:10.05 loss:0.006 grdn:0.208 lr:5.8e-06 updt_s:1.185 data_s:0.012
197
+ INFO 2026-01-06 06:59:11 ot_train.py:501 step:18K smpl:1M ep:2K epch:10.11 loss:0.006 grdn:0.183 lr:5.5e-06 updt_s:1.187 data_s:0.005
198
+ INFO 2026-01-06 07:01:10 ot_train.py:501 step:18K smpl:1M ep:2K epch:10.17 loss:0.006 grdn:0.189 lr:5.3e-06 updt_s:1.187 data_s:0.005
199
+ INFO 2026-01-06 07:03:09 ot_train.py:501 step:18K smpl:1M ep:2K epch:10.22 loss:0.006 grdn:0.188 lr:5.0e-06 updt_s:1.187 data_s:0.005
200
+ INFO 2026-01-06 07:05:09 ot_train.py:501 step:18K smpl:1M ep:2K epch:10.28 loss:0.006 grdn:0.196 lr:4.8e-06 updt_s:1.187 data_s:0.005
201
+ INFO 2026-01-06 07:07:08 ot_train.py:501 step:18K smpl:1M ep:2K epch:10.34 loss:0.006 grdn:0.176 lr:4.5e-06 updt_s:1.187 data_s:0.005
202
+ INFO 2026-01-06 07:09:07 ot_train.py:501 step:18K smpl:1M ep:2K epch:10.39 loss:0.006 grdn:0.205 lr:4.3e-06 updt_s:1.187 data_s:0.005
203
+ INFO 2026-01-06 07:11:07 ot_train.py:501 step:18K smpl:1M ep:2K epch:10.45 loss:0.006 grdn:0.198 lr:4.1e-06 updt_s:1.187 data_s:0.005
204
+ INFO 2026-01-06 07:13:06 ot_train.py:501 step:18K smpl:1M ep:2K epch:10.51 loss:0.006 grdn:0.179 lr:3.9e-06 updt_s:1.187 data_s:0.005
205
+ INFO 2026-01-06 07:15:05 ot_train.py:501 step:19K smpl:1M ep:2K epch:10.56 loss:0.006 grdn:0.182 lr:3.8e-06 updt_s:1.187 data_s:0.005
206
+ INFO 2026-01-06 07:17:05 ot_train.py:501 step:19K smpl:1M ep:2K epch:10.62 loss:0.006 grdn:0.186 lr:3.6e-06 updt_s:1.187 data_s:0.005
207
+ INFO 2026-01-06 07:19:04 ot_train.py:501 step:19K smpl:1M ep:2K epch:10.68 loss:0.006 grdn:0.189 lr:3.4e-06 updt_s:1.187 data_s:0.005
208
+ INFO 2026-01-06 07:21:03 ot_train.py:501 step:19K smpl:1M ep:2K epch:10.73 loss:0.006 grdn:0.190 lr:3.3e-06 updt_s:1.187 data_s:0.005
209
+ INFO 2026-01-06 07:23:03 ot_train.py:501 step:19K smpl:1M ep:2K epch:10.79 loss:0.006 grdn:0.188 lr:3.2e-06 updt_s:1.187 data_s:0.005
210
+ INFO 2026-01-06 07:25:02 ot_train.py:501 step:19K smpl:1M ep:2K epch:10.85 loss:0.006 grdn:0.190 lr:3.0e-06 updt_s:1.187 data_s:0.005
211
+ INFO 2026-01-06 07:27:01 ot_train.py:501 step:19K smpl:1M ep:2K epch:10.90 loss:0.006 grdn:0.194 lr:2.9e-06 updt_s:1.187 data_s:0.005
212
+ INFO 2026-01-06 07:29:01 ot_train.py:501 step:19K smpl:1M ep:2K epch:10.96 loss:0.006 grdn:0.186 lr:2.8e-06 updt_s:1.187 data_s:0.005
213
+ INFO 2026-01-06 07:31:00 ot_train.py:501 step:19K smpl:1M ep:2K epch:11.02 loss:0.006 grdn:0.201 lr:2.8e-06 updt_s:1.185 data_s:0.012
214
+ INFO 2026-01-06 07:33:00 ot_train.py:501 step:20K smpl:1M ep:2K epch:11.07 loss:0.006 grdn:0.168 lr:2.7e-06 updt_s:1.187 data_s:0.005
215
+ INFO 2026-01-06 07:34:59 ot_train.py:501 step:20K smpl:1M ep:2K epch:11.13 loss:0.006 grdn:0.196 lr:2.6e-06 updt_s:1.187 data_s:0.005
216
+ INFO 2026-01-06 07:36:58 ot_train.py:501 step:20K smpl:1M ep:2K epch:11.19 loss:0.006 grdn:0.178 lr:2.6e-06 updt_s:1.187 data_s:0.005
217
+ INFO 2026-01-06 07:38:58 ot_train.py:501 step:20K smpl:1M ep:2K epch:11.25 loss:0.006 grdn:0.183 lr:2.5e-06 updt_s:1.187 data_s:0.005
218
+ INFO 2026-01-06 07:40:57 ot_train.py:501 step:20K smpl:1M ep:2K epch:11.30 loss:0.006 grdn:0.185 lr:2.5e-06 updt_s:1.187 data_s:0.005
219
+ INFO 2026-01-06 07:42:56 ot_train.py:501 step:20K smpl:1M ep:2K epch:11.36 los