Hugging Face's logo

ojaffe
/
world-model

Model card Files
xet
 Community
ojaffe commited on
Commit
47a3fc6
·
verified ·
1 Parent(s): 1bb1337

Upload folder using huggingface_hub

Browse files
Files changed (4) hide show
  1. __pycache__/predict.cpython-311.pyc +0 -0
  2. model_pong.pt +1 -1
  3. predict.py +2 -2
  4. train.log +77 -152
__pycache__/predict.cpython-311.pyc CHANGED
Binary files a/__pycache__/predict.cpython-311.pyc and b/__pycache__/predict.cpython-311.pyc differ
 
model_pong.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:d6c8b9235347bea94e7e5f5f0f225d4c1dbd13a749d5e28920c75c91902ecb11
3
  size 2435368
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:96883ed257d0d3f18b836e7b7592bbb3cb72e802de400a89da98733667c88c90
3
  size 2435368
predict.py CHANGED
@@ -1,4 +1,4 @@
1
- """Blend sweep: Pong AR weight 0.9->0.7, Sonic unchanged 0.7->0.3."""
2
  import sys
3
  import os
4
  import numpy as np
@@ -167,7 +167,7 @@ def predict_next_frame(ens, context_frames: np.ndarray) -> np.ndarray:
167
 
168
  predicted = torch.zeros_like(direct_pred)
169
  for step in range(PRED_FRAMES):
170
- ar_weight = 0.9 - (step / (PRED_FRAMES - 1)) * 0.2
171
  direct_weight = 1.0 - ar_weight
172
  predicted[:, step] = ar_weight * ar_pred[:, step] + direct_weight * direct_pred[:, step]
173
 
 
1
+ """Optimized blend: Pong AR weight 0.85->0.65, Sonic unchanged 0.7->0.3."""
2
  import sys
3
  import os
4
  import numpy as np
 
167
 
168
  predicted = torch.zeros_like(direct_pred)
169
  for step in range(PRED_FRAMES):
170
+ ar_weight = 0.85 - (step / (PRED_FRAMES - 1)) * 0.2
171
  direct_weight = 1.0 - ar_weight
172
  predicted[:, step] = ar_weight * ar_pred[:, step] + direct_weight * direct_pred[:, step]
173
 
train.log CHANGED
@@ -1,152 +1,77 @@
1
- [2026-04-11 20:37:53] Starting improved direct 8-frame training for 2026-04-11-210000-direct-improved
2
- [2026-04-11 20:37:53] Device: cuda
3
- [2026-04-11 20:37:53] === pong ===
4
- [2026-04-11 20:37:53] pong: 1,199,224 params (2.3 MB fp16)
5
- [2026-04-11 20:37:54] pong train: 8194 seqs (len=16)
6
- [2026-04-11 20:37:54] pong val: 964 seqs (len=16)
7
- [2026-04-11 20:38:15] pong E1/150 | T:0.245151(S:0.6311) V:0.225469(S:0.6589) LR:2.00e-04
8
- [2026-04-11 20:38:35] pong E2/150 | T:0.207902(S:0.6846) V:0.207636(S:0.6836) LR:2.00e-04
9
- [2026-04-11 20:38:55] pong E3/150 | T:0.183334(S:0.7217) V:0.184517(S:0.7193) LR:2.00e-04
10
- [2026-04-11 20:39:15] pong E4/150 | T:0.162890(S:0.7525) V:0.170139(S:0.7405) LR:2.00e-04
11
- [2026-04-11 20:39:35] pong E5/150 | T:0.148245(S:0.7747) V:0.168455(S:0.7455) LR:1.99e-04
12
- [2026-04-11 20:39:54] pong E6/150 | T:0.134784(S:0.7951) V:0.155488(S:0.7639) LR:1.99e-04
13
- [2026-04-11 20:40:15] pong E7/150 | T:0.125337(S:0.8093) V:0.153139(S:0.7670) LR:1.99e-04
14
- [2026-04-11 20:40:35] pong E8/150 | T:0.117548(S:0.8213) V:0.140387(S:0.7865) LR:1.99e-04
15
- [2026-04-11 20:40:55] pong E9/150 | T:0.110727(S:0.8316) V:0.139567(S:0.7879) LR:1.98e-04
16
- [2026-04-11 20:41:15] pong E10/150 | T:0.105040(S:0.8404) V:0.135133(S:0.7943) LR:1.98e-04
17
- [2026-04-11 20:41:55] pong E12/150 | T:0.096706(S:0.8530) V:0.133117(S:0.7976) LR:1.97e-04
18
- [2026-04-11 20:42:15] pong E13/150 | T:0.093069(S:0.8586) V:0.128657(S:0.8049) LR:1.96e-04
19
- [2026-04-11 20:42:35] pong E14/150 | T:0.089584(S:0.8641) V:0.127223(S:0.8071) LR:1.96e-04
20
- [2026-04-11 20:42:55] pong E15/150 | T:0.086566(S:0.8686) V:0.123897(S:0.8117) LR:1.95e-04
21
- [2026-04-11 20:43:35] pong E17/150 | T:0.081673(S:0.8762) V:0.123559(S:0.8119) LR:1.94e-04
22
- [2026-04-11 20:44:15] pong E19/150 | T:0.076816(S:0.8836) V:0.120578(S:0.8168) LR:1.92e-04
23
- [2026-04-11 20:44:35] pong E20/150 | T:0.075138(S:0.8862) V:0.123831(S:0.8129) LR:1.91e-04
24
- [2026-04-11 20:44:54] pong E21/150 | T:0.072953(S:0.8897) V:0.120207(S:0.8171) LR:1.91e-04
25
- [2026-04-11 20:45:14] pong E22/150 | T:0.071339(S:0.8922) V:0.118498(S:0.8202) LR:1.90e-04
26
- [2026-04-11 20:45:35] pong E23/150 | T:0.069781(S:0.8947) V:0.117458(S:0.8223) LR:1.89e-04
27
- [2026-04-11 20:46:12] pong E25/150 | T:0.067305(S:0.8984) V:0.116407(S:0.8233) LR:1.87e-04
28
- [2026-04-11 20:46:32] pong E26/150 | T:0.065223(S:0.9018) V:0.115602(S:0.8242) LR:1.86e-04
29
- [2026-04-11 20:47:12] pong E28/150 | T:0.062886(S:0.9054) V:0.113464(S:0.8277) LR:1.83e-04
30
- [2026-04-11 20:47:32] pong E29/150 | T:0.061637(S:0.9073) V:0.112811(S:0.8285) LR:1.82e-04
31
- [2026-04-11 20:47:51] pong E30/150 | T:0.060839(S:0.9086) V:0.112722(S:0.8292) LR:1.81e-04
32
- [2026-04-11 20:48:12] pong E31/150 | T:0.059712(S:0.9104) V:0.111062(S:0.8316) LR:1.80e-04
33
- [2026-04-11 20:49:12] pong E34/150 | T:0.056734(S:0.9149) V:0.111023(S:0.8316) LR:1.76e-04
34
- [2026-04-11 20:49:32] pong E35/150 | T:0.055712(S:0.9165) V:0.110123(S:0.8330) LR:1.74e-04
35
- [2026-04-11 20:50:10] pong E37/150 | T:0.054400(S:0.9186) V:0.109606(S:0.8340) LR:1.72e-04
36
- [2026-04-11 20:51:10] pong E40/150 | T:0.052566(S:0.9215) V:0.109744(S:0.8340) LR:1.67e-04
37
- [2026-04-11 20:51:30] pong E41/150 | T:0.051662(S:0.9229) V:0.108173(S:0.8359) LR:1.66e-04
38
- [2026-04-11 20:52:10] pong E43/150 | T:0.050829(S:0.9242) V:0.107573(S:0.8367) LR:1.62e-04
39
- [2026-04-11 20:52:30] pong E44/150 | T:0.049584(S:0.9261) V:0.107344(S:0.8375) LR:1.61e-04
40
- [2026-04-11 20:53:30] pong E47/150 | T:0.047870(S:0.9288) V:0.107087(S:0.8377) LR:1.56e-04
41
- [2026-04-11 20:53:49] pong E48/150 | T:0.047106(S:0.9299) V:0.105825(S:0.8395) LR:1.54e-04
42
- [2026-04-11 20:54:29] pong E50/150 | T:0.046352(S:0.9312) V:0.108601(S:0.8357) LR:1.50e-04
43
- [2026-04-11 20:55:09] pong E52/150 | T:0.045529(S:0.9325) V:0.105138(S:0.8406) LR:1.47e-04
44
- [2026-04-11 20:56:08] pong E55/150 | T:0.043897(S:0.9350) V:0.104684(S:0.8416) LR:1.41e-04
45
- [2026-04-11 20:56:27] pong E56/150 | T:0.043453(S:0.9357) V:0.104668(S:0.8417) LR:1.39e-04
46
- [2026-04-11 20:57:45] pong E60/150 | T:0.041694(S:0.9384) V:0.103771(S:0.8432) LR:1.31e-04
47
- [2026-04-11 20:58:23] pong E62/150 | T:0.040995(S:0.9395) V:0.103378(S:0.8433) LR:1.27e-04
48
- [2026-04-11 20:59:03] pong E64/150 | T:0.040430(S:0.9403) V:0.102376(S:0.8449) LR:1.23e-04
49
- [2026-04-11 21:01:00] pong E70/150 | T:0.038175(S:0.9439) V:0.101971(S:0.8454) LR:1.11e-04
50
- [2026-04-11 21:01:20] pong E71/150 | T:0.038048(S:0.9441) V:0.101782(S:0.8458) LR:1.09e-04
51
- [2026-04-11 21:02:01] pong E73/150 | T:0.037274(S:0.9453) V:0.101483(S:0.8464) LR:1.05e-04
52
- [2026-04-11 21:02:21] pong E74/150 | T:0.037082(S:0.9456) V:0.101132(S:0.8470) LR:1.03e-04
53
- [2026-04-11 21:03:35] pong E78/150 | T:0.036125(S:0.9471) V:0.100967(S:0.8472) LR:9.43e-05
54
- [2026-04-11 21:04:14] pong E80/150 | T:0.035617(S:0.9479) V:0.101149(S:0.8473) LR:9.01e-05
55
- [2026-04-11 21:04:34] pong E81/150 | T:0.035287(S:0.9484) V:0.100887(S:0.8473) LR:8.80e-05
56
- [2026-04-11 21:05:14] pong E83/150 | T:0.034890(S:0.9490) V:0.100673(S:0.8476) LR:8.39e-05
57
- [2026-04-11 21:05:34] pong E84/150 | T:0.034546(S:0.9495) V:0.100503(S:0.8481) LR:8.19e-05
58
- [2026-04-11 21:05:53] pong E85/150 | T:0.034332(S:0.9499) V:0.100085(S:0.8486) LR:7.98e-05
59
- [2026-04-11 21:06:51] pong E88/150 | T:0.033905(S:0.9505) V:0.099703(S:0.8491) LR:7.37e-05
60
- [2026-04-11 21:07:32] pong E90/150 | T:0.033352(S:0.9514) V:0.100154(S:0.8485) LR:6.98e-05
61
- [2026-04-11 21:07:49] pong E91/150 | T:0.033275(S:0.9515) V:0.099463(S:0.8496) LR:6.78e-05
62
- [2026-04-11 21:08:28] pong E93/150 | T:0.032808(S:0.9522) V:0.099426(S:0.8497) LR:6.39e-05
63
- [2026-04-11 21:09:07] pong E95/150 | T:0.032401(S:0.9529) V:0.099274(S:0.8498) LR:6.00e-05
64
- [2026-04-11 21:10:47] pong E100/150 | T:0.031693(S:0.9539) V:0.099102(S:0.8502) LR:5.07e-05
65
- [2026-04-11 21:11:07] pong E101/150 | T:0.031556(S:0.9542) V:0.098989(S:0.8504) LR:4.90e-05
66
- [2026-04-11 21:12:05] pong E104/150 | T:0.031188(S:0.9547) V:0.098818(S:0.8506) LR:4.37e-05
67
- [2026-04-11 21:12:26] pong E105/150 | T:0.031013(S:0.9550) V:0.098630(S:0.8508) LR:4.20e-05
68
- [2026-04-11 21:12:45] pong E106/150 | T:0.030918(S:0.9552) V:0.098413(S:0.8511) LR:4.03e-05
69
- [2026-04-11 21:14:05] pong E110/150 | T:0.030445(S:0.9559) V:0.099033(S:0.8506) LR:3.39e-05
70
- [2026-04-11 21:15:25] pong E114/150 | T:0.030015(S:0.9566) V:0.098366(S:0.8513) LR:2.80e-05
71
- [2026-04-11 21:16:25] pong E117/150 | T:0.029792(S:0.9569) V:0.098095(S:0.8517) LR:2.38e-05
72
- [2026-04-11 21:17:24] pong E120/150 | T:0.029565(S:0.9573) V:0.098241(S:0.8515) LR:2.00e-05
73
- [2026-04-11 21:19:24] pong E126/150 | T:0.029172(S:0.9579) V:0.098066(S:0.8518) LR:1.33e-05
74
- [2026-04-11 21:20:03] pong E128/150 | T:0.029068(S:0.9580) V:0.098040(S:0.8518) LR:1.14e-05
75
- [2026-04-11 21:20:43] pong E130/150 | T:0.029001(S:0.9581) V:0.098031(S:0.8518) LR:9.60e-06
76
- [2026-04-11 21:21:03] pong E131/150 | T:0.028939(S:0.9582) V:0.097892(S:0.8521) LR:8.77e-06
77
- [2026-04-11 21:24:01] pong E140/150 | T:0.028630(S:0.9587) V:0.097958(S:0.8520) LR:3.17e-06
78
- [2026-04-11 21:25:19] pong E144/150 | T:0.028567(S:0.9588) V:0.097818(S:0.8522) LR:1.78e-06
79
- [2026-04-11 21:27:18] pong E150/150 | T:0.028536(S:0.9589) V:0.097904(S:0.8521) LR:1.00e-06
80
- [2026-04-11 21:27:18] pong done. Best: 0.097818
81
- [2026-04-11 21:27:18] === pole_position ===
82
- [2026-04-11 21:27:18] pole_position: 1,465,848 params (2.8 MB fp16)
83
- [2026-04-11 21:27:19] pole_position train: 4097 seqs (len=16)
84
- [2026-04-11 21:27:19] pole_position val: 482 seqs (len=16)
85
- [2026-04-11 21:27:29] pole_position E1/150 | T:0.072483(S:0.9013) V:0.055104(S:0.9272) LR:2.00e-04
86
- [2026-04-11 21:27:39] pole_position E2/150 | T:0.060683(S:0.9178) V:0.049558(S:0.9337) LR:2.00e-04
87
- [2026-04-11 21:27:49] pole_position E3/150 | T:0.056482(S:0.9234) V:0.049681(S:0.9330) LR:2.00e-04
88
- [2026-04-11 21:28:00] pole_position E4/150 | T:0.053962(S:0.9268) V:0.047040(S:0.9371) LR:2.00e-04
89
- [2026-04-11 21:28:09] pole_position E5/150 | T:0.051783(S:0.9298) V:0.043891(S:0.9410) LR:1.99e-04
90
- [2026-04-11 21:28:39] pole_position E8/150 | T:0.047291(S:0.9360) V:0.042535(S:0.9431) LR:1.99e-04
91
- [2026-04-11 21:28:48] pole_position E9/150 | T:0.046385(S:0.9372) V:0.041829(S:0.9440) LR:1.98e-04
92
- [2026-04-11 21:28:58] pole_position E10/150 | T:0.045514(S:0.9385) V:0.042079(S:0.9438) LR:1.98e-04
93
- [2026-04-11 21:29:09] pole_position E11/150 | T:0.044486(S:0.9399) V:0.040944(S:0.9449) LR:1.97e-04
94
- [2026-04-11 21:29:47] pole_position E15/150 | T:0.041111(S:0.9447) V:0.039725(S:0.9465) LR:1.95e-04
95
- [2026-04-11 21:30:37] pole_position E20/150 | T:0.038348(S:0.9487) V:0.038742(S:0.9477) LR:1.91e-04
96
- [2026-04-11 21:30:46] pole_position E21/150 | T:0.037787(S:0.9495) V:0.038655(S:0.9481) LR:1.91e-04
97
- [2026-04-11 21:31:18] pole_position E24/150 | T:0.036649(S:0.9511) V:0.038322(S:0.9481) LR:1.88e-04
98
- [2026-04-11 21:32:18] pole_position E30/150 | T:0.034550(S:0.9541) V:0.038329(S:0.9484) LR:1.81e-04
99
- [2026-04-11 21:33:18] pole_position E36/150 | T:0.033149(S:0.9561) V:0.038183(S:0.9480) LR:1.73e-04
100
- [2026-04-11 21:33:59] pole_position E40/150 | T:0.032073(S:0.9577) V:0.038579(S:0.9475) LR:1.67e-04
101
- [2026-04-11 21:34:09] pole_position E41/150 | T:0.031761(S:0.9581) V:0.037873(S:0.9485) LR:1.66e-04
102
- [2026-04-11 21:34:19] pole_position E42/150 | T:0.031713(S:0.9582) V:0.036978(S:0.9498) LR:1.64e-04
103
- [2026-04-11 21:35:39] pole_position E50/150 | T:0.030141(S:0.9604) V:0.038688(S:0.9470) LR:1.50e-04
104
- [2026-04-11 21:37:19] pole_position E60/150 | T:0.028820(S:0.9623) V:0.038451(S:0.9473) LR:1.31e-04
105
- [2026-04-11 21:39:00] pole_position E70/150 | T:0.027896(S:0.9636) V:0.038069(S:0.9478) LR:1.11e-04
106
- [2026-04-11 21:40:41] pole_position E80/150 | T:0.026692(S:0.9653) V:0.038111(S:0.9476) LR:9.01e-05
107
- [2026-04-11 21:42:22] pole_position E90/150 | T:0.025885(S:0.9665) V:0.037698(S:0.9482) LR:6.98e-05
108
- [2026-04-11 21:44:01] pole_position E100/150 | T:0.025252(S:0.9674) V:0.038106(S:0.9475) LR:5.07e-05
109
- [2026-04-11 21:45:42] pole_position E110/150 | T:0.024739(S:0.9681) V:0.037902(S:0.9479) LR:3.39e-05
110
- [2026-04-11 21:47:24] pole_position E120/150 | T:0.024433(S:0.9685) V:0.037644(S:0.9482) LR:2.00e-05
111
- [2026-04-11 21:49:05] pole_position E130/150 | T:0.024134(S:0.9690) V:0.037839(S:0.9479) LR:9.60e-06
112
- [2026-04-11 21:50:45] pole_position E140/150 | T:0.024041(S:0.9691) V:0.037827(S:0.9479) LR:3.17e-06
113
- [2026-04-11 21:52:25] pole_position E150/150 | T:0.023978(S:0.9692) V:0.037890(S:0.9478) LR:1.00e-06
114
- [2026-04-11 21:52:25] pole_position done. Best: 0.036978
115
- [2026-04-11 21:52:25] === sonic ===
116
- [2026-04-11 21:52:25] sonic: 3,071,016 params (5.9 MB fp16)
117
- [2026-04-11 21:52:29] sonic train: 30848 seqs (len=16)
118
- [2026-04-11 21:52:30] sonic val: 3856 seqs (len=16)
119
- [2026-04-11 21:53:45] sonic E1/150 | T:0.162348(S:0.7775) V:0.144778(S:0.8026) LR:2.00e-04
120
- [2026-04-11 21:54:58] sonic E2/150 | T:0.146694(S:0.7991) V:0.136146(S:0.8135) LR:2.00e-04
121
- [2026-04-11 21:56:13] sonic E3/150 | T:0.140699(S:0.8074) V:0.134994(S:0.8151) LR:2.00e-04
122
- [2026-04-11 21:57:28] sonic E4/150 | T:0.136883(S:0.8127) V:0.131966(S:0.8195) LR:2.00e-04
123
- [2026-04-11 21:58:43] sonic E5/150 | T:0.133509(S:0.8175) V:0.129399(S:0.8230) LR:1.99e-04
124
- [2026-04-11 21:59:57] sonic E6/150 | T:0.130435(S:0.8218) V:0.128735(S:0.8233) LR:1.99e-04
125
- [2026-04-11 22:01:14] sonic E7/150 | T:0.127966(S:0.8254) V:0.126909(S:0.8259) LR:1.99e-04
126
- [2026-04-11 22:04:57] sonic E10/150 | T:0.121944(S:0.8340) V:0.124331(S:0.8295) LR:1.98e-04
127
- [2026-04-11 22:06:11] sonic E11/150 | T:0.120293(S:0.8364) V:0.123743(S:0.8304) LR:1.97e-04
128
- [2026-04-11 22:08:39] sonic E13/150 | T:0.117218(S:0.8409) V:0.122889(S:0.8311) LR:1.97e-04
129
- [2026-04-11 22:14:48] sonic E18/150 | T:0.111090(S:0.8498) V:0.122588(S:0.8317) LR:1.93e-04
130
- [2026-04-11 22:17:15] sonic E20/150 | T:0.109268(S:0.8525) V:0.125544(S:0.8273) LR:1.92e-04
131
- [2026-04-11 22:20:54] sonic E23/150 | T:0.106768(S:0.8561) V:0.121705(S:0.8325) LR:1.89e-04
132
- [2026-04-11 22:23:25] sonic E25/150 | T:0.105207(S:0.8584) V:0.121577(S:0.8323) LR:1.87e-04
133
- [2026-04-11 22:27:00] sonic E28/150 | T:0.103225(S:0.8613) V:0.120523(S:0.8339) LR:1.84e-04
134
- [2026-04-11 22:29:26] sonic E30/150 | T:0.102327(S:0.8626) V:0.123279(S:0.8299) LR:1.82e-04
135
- [2026-04-11 22:41:41] sonic E40/150 | T:0.097330(S:0.8698) V:0.123097(S:0.8299) LR:1.69e-04
136
- [2026-04-11 22:53:59] sonic E50/150 | T:0.093990(S:0.8746) V:0.124820(S:0.8272) LR:1.53e-04
137
- [2026-04-11 23:06:25] sonic E60/150 | T:0.091507(S:0.8782) V:0.124273(S:0.8274) LR:1.34e-04
138
- [2026-04-11 23:18:48] sonic E70/150 | T:0.089327(S:0.8813) V:0.123787(S:0.8284) LR:1.15e-04
139
- [2026-04-11 23:31:12] sonic E80/150 | T:0.087458(S:0.8839) V:0.122982(S:0.8295) LR:9.51e-05
140
- [2026-04-11 23:44:30] sonic E90/150 | T:0.085960(S:0.8861) V:0.123807(S:0.8280) LR:7.56e-05
141
- [2026-04-11 23:57:54] sonic E100/150 | T:0.084800(S:0.8877) V:0.124093(S:0.8276) LR:5.75e-05
142
- [2026-04-12 00:11:21] sonic E110/150 | T:0.083786(S:0.8892) V:0.124339(S:0.8272) LR:4.14e-05
143
- [2026-04-12 00:25:17] sonic E120/150 | T:0.083098(S:0.8902) V:0.125028(S:0.8261) LR:2.81e-05
144
- [2026-04-12 00:39:13] sonic E130/150 | T:0.082543(S:0.8910) V:0.124983(S:0.8261) LR:1.82e-05
145
- [2026-04-12 00:53:09] sonic E140/150 | T:0.082184(S:0.8915) V:0.125446(S:0.8255) LR:1.21e-05
146
- [2026-04-12 01:06:48] sonic E150/150 | T:0.082048(S:0.8917) V:0.124580(S:0.8267) LR:1.00e-05
147
- [2026-04-12 01:06:48] sonic done. Best: 0.120523
148
- [2026-04-12 01:06:48] pong: 2.3 MB
149
- [2026-04-12 01:06:48] pole_position: 2.8 MB
150
- [2026-04-12 01:06:48] sonic: 5.9 MB
151
- [2026-04-12 01:06:48] Total: 11.1 MB
152
- [2026-04-12 01:06:48] Training complete!
 
1
+ [2026-04-12 05:14:21] Starting augmented Pong AR training for 2026-04-12-110000-pong-augmented
2
+ [2026-04-12 05:14:21] Device: cuda
3
+ [2026-04-12 05:14:21] Pong params: 1,198,531 (2.3 MB fp16)
4
+ [2026-04-12 05:14:21] Phase 1: single-step + hflip aug, 100 epochs, lr=0.0003
5
+ [2026-04-12 05:14:22] pong train: 8432 sequences (seq_len=9)
6
+ [2026-04-12 05:14:22] pong val: 992 sequences (seq_len=9)
7
+ [2026-04-12 05:14:31] P1 E1/100 | T:0.127401(S:0.8196) V:0.110707(S:0.8434) LR:3.00e-04
8
+ [2026-04-12 05:14:38] P1 E2/100 | T:0.101264(S:0.8567) V:0.099352(S:0.8593) LR:3.00e-04
9
+ [2026-04-12 05:14:46] P1 E3/100 | T:0.093089(S:0.8683) V:0.101696(S:0.8561) LR:2.99e-04
10
+ [2026-04-12 05:15:01] P1 E5/100 | T:0.082154(S:0.8838) V:0.084095(S:0.8810) LR:2.98e-04
11
+ [2026-04-12 05:15:07] P1 E6/100 | T:0.077151(S:0.8908) V:0.080748(S:0.8857) LR:2.97e-04
12
+ [2026-04-12 05:15:22] P1 E8/100 | T:0.066223(S:0.9063) V:0.075570(S:0.8930) LR:2.95e-04
13
+ [2026-04-12 05:15:37] P1 E10/100 | T:0.059430(S:0.9160) V:0.077257(S:0.8907) LR:2.93e-04
14
+ [2026-04-12 05:15:44] P1 E11/100 | T:0.055401(S:0.9217) V:0.071990(S:0.8981) LR:2.91e-04
15
+ [2026-04-12 05:15:51] P1 E12/100 | T:0.052138(S:0.9263) V:0.069664(S:0.9014) LR:2.90e-04
16
+ [2026-04-12 05:15:59] P1 E13/100 | T:0.049025(S:0.9307) V:0.069328(S:0.9019) LR:2.88e-04
17
+ [2026-04-12 05:16:06] P1 E14/100 | T:0.047183(S:0.9333) V:0.063733(S:0.9098) LR:2.86e-04
18
+ [2026-04-12 05:16:12] P1 E15/100 | T:0.044935(S:0.9365) V:0.063349(S:0.9104) LR:2.84e-04
19
+ [2026-04-12 05:16:27] P1 E17/100 | T:0.042284(S:0.9403) V:0.061114(S:0.9135) LR:2.79e-04
20
+ [2026-04-12 05:16:35] P1 E18/100 | T:0.040933(S:0.9422) V:0.058447(S:0.9173) LR:2.77e-04
21
+ [2026-04-12 05:16:49] P1 E20/100 | T:0.038315(S:0.9459) V:0.061833(S:0.9125) LR:2.71e-04
22
+ [2026-04-12 05:16:57] P1 E21/100 | T:0.037595(S:0.9469) V:0.057463(S:0.9187) LR:2.69e-04
23
+ [2026-04-12 05:17:04] P1 E22/100 | T:0.036860(S:0.9479) V:0.055401(S:0.9216) LR:2.66e-04
24
+ [2026-04-12 05:17:18] P1 E24/100 | T:0.033665(S:0.9525) V:0.055152(S:0.9220) LR:2.59e-04
25
+ [2026-04-12 05:17:25] P1 E25/100 | T:0.032287(S:0.9544) V:0.054026(S:0.9235) LR:2.56e-04
26
+ [2026-04-12 05:17:40] P1 E27/100 | T:0.030628(S:0.9568) V:0.053318(S:0.9245) LR:2.49e-04
27
+ [2026-04-12 05:17:47] P1 E28/100 | T:0.029062(S:0.9590) V:0.051204(S:0.9275) LR:2.46e-04
28
+ [2026-04-12 05:18:02] P1 E30/100 | T:0.027471(S:0.9612) V:0.049118(S:0.9305) LR:2.38e-04
29
+ [2026-04-12 05:18:31] P1 E34/100 | T:0.024918(S:0.9649) V:0.048185(S:0.9318) LR:2.23e-04
30
+ [2026-04-12 05:18:53] P1 E37/100 | T:0.022570(S:0.9682) V:0.046684(S:0.9339) LR:2.10e-04
31
+ [2026-04-12 05:19:14] P1 E40/100 | T:0.020725(S:0.9708) V:0.045685(S:0.9353) LR:1.97e-04
32
+ [2026-04-12 05:19:29] P1 E42/100 | T:0.019517(S:0.9725) V:0.045561(S:0.9355) LR:1.88e-04
33
+ [2026-04-12 05:19:51] P1 E45/100 | T:0.018041(S:0.9746) V:0.045091(S:0.9362) LR:1.74e-04
34
+ [2026-04-12 05:20:05] P1 E47/100 | T:0.017219(S:0.9757) V:0.044821(S:0.9365) LR:1.65e-04
35
+ [2026-04-12 05:20:13] P1 E48/100 | T:0.016920(S:0.9762) V:0.044686(S:0.9367) LR:1.60e-04
36
+ [2026-04-12 05:20:28] P1 E50/100 | T:0.016031(S:0.9774) V:0.045932(S:0.9350) LR:1.50e-04
37
+ [2026-04-12 05:20:43] P1 E52/100 | T:0.015575(S:0.9781) V:0.044620(S:0.9368) LR:1.41e-04
38
+ [2026-04-12 05:20:58] P1 E54/100 | T:0.014855(S:0.9791) V:0.044487(S:0.9370) LR:1.32e-04
39
+ [2026-04-12 05:21:04] P1 E55/100 | T:0.014467(S:0.9796) V:0.043267(S:0.9387) LR:1.27e-04
40
+ [2026-04-12 05:21:11] P1 E56/100 | T:0.013987(S:0.9803) V:0.043114(S:0.9389) LR:1.22e-04
41
+ [2026-04-12 05:21:41] P1 E60/100 | T:0.012814(S:0.9820) V:0.042679(S:0.9396) LR:1.04e-04
42
+ [2026-04-12 05:21:55] P1 E62/100 | T:0.012138(S:0.9829) V:0.041535(S:0.9412) LR:9.55e-05
43
+ [2026-04-12 05:22:09] P1 E64/100 | T:0.011468(S:0.9839) V:0.041223(S:0.9416) LR:8.68e-05
44
+ [2026-04-12 05:22:42] P1 E69/100 | T:0.010313(S:0.9855) V:0.040714(S:0.9423) LR:6.65e-05
45
+ [2026-04-12 05:22:49] P1 E70/100 | T:0.010118(S:0.9858) V:0.041125(S:0.9418) LR:6.26e-05
46
+ [2026-04-12 05:23:22] P1 E75/100 | T:0.009125(S:0.9872) V:0.040657(S:0.9424) LR:4.48e-05
47
+ [2026-04-12 05:23:52] P1 E79/100 | T:0.008602(S:0.9879) V:0.040464(S:0.9427) LR:3.24e-05
48
+ [2026-04-12 05:23:59] P1 E80/100 | T:0.008478(S:0.9881) V:0.040606(S:0.9425) LR:2.96e-05
49
+ [2026-04-12 05:24:19] P1 E83/100 | T:0.008091(S:0.9886) V:0.040422(S:0.9427) LR:2.18e-05
50
+ [2026-04-12 05:24:34] P1 E85/100 | T:0.008020(S:0.9887) V:0.040420(S:0.9427) LR:1.73e-05
51
+ [2026-04-12 05:24:49] P1 E87/100 | T:0.007915(S:0.9889) V:0.040373(S:0.9428) LR:1.33e-05
52
+ [2026-04-12 05:24:56] P1 E88/100 | T:0.007870(S:0.9890) V:0.040319(S:0.9429) LR:1.15e-05
53
+ [2026-04-12 05:25:10] P1 E90/100 | T:0.007718(S:0.9892) V:0.040199(S:0.9431) LR:8.32e-06
54
+ [2026-04-12 05:25:17] P1 E91/100 | T:0.007692(S:0.9892) V:0.040185(S:0.9431) LR:6.94e-06
55
+ [2026-04-12 05:25:24] P1 E92/100 | T:0.007651(S:0.9893) V:0.040164(S:0.9431) LR:5.70e-06
56
+ [2026-04-12 05:25:32] P1 E93/100 | T:0.007631(S:0.9893) V:0.040117(S:0.9432) LR:4.60e-06
57
+ [2026-04-12 05:26:21] P1 E100/100 | T:0.007589(S:0.9893) V:0.040109(S:0.9432) LR:1.00e-06
58
+ [2026-04-12 05:26:21] Phase 1 done. Best val loss: 0.040109
59
+ [2026-04-12 05:26:21] Phase 2: 2-step AR + hflip aug, 50 epochs, lr=5e-05
60
+ [2026-04-12 05:26:22] pong train: 8398 sequences (seq_len=10)
61
+ [2026-04-12 05:26:22] pong val: 988 sequences (seq_len=10)
62
+ [2026-04-12 05:26:44] P2 E1/50 | T:0.014117(S:0.9801) V:0.056007(S:0.9206) LR:5.00e-05
63
+ [2026-04-12 05:27:05] P2 E2/50 | T:0.013481(S:0.9810) V:0.054888(S:0.9222) LR:4.98e-05
64
+ [2026-04-12 05:27:26] P2 E3/50 | T:0.013067(S:0.9816) V:0.054390(S:0.9229) LR:4.96e-05
65
+ [2026-04-12 05:28:52] P2 E7/50 | T:0.011823(S:0.9834) V:0.053736(S:0.9238) LR:4.77e-05
66
+ [2026-04-12 05:30:01] P2 E10/50 | T:0.011506(S:0.9838) V:0.054305(S:0.9230) LR:4.53e-05
67
+ [2026-04-12 05:33:34] P2 E20/50 | T:0.010052(S:0.9859) V:0.053792(S:0.9237) LR:3.31e-05
68
+ [2026-04-12 05:34:17] P2 E22/50 | T:0.009761(S:0.9863) V:0.053448(S:0.9242) LR:3.01e-05
69
+ [2026-04-12 05:34:39] P2 E23/50 | T:0.009488(S:0.9867) V:0.053285(S:0.9245) LR:2.86e-05
70
+ [2026-04-12 05:35:22] P2 E25/50 | T:0.009336(S:0.9869) V:0.052888(S:0.9250) LR:2.55e-05
71
+ [2026-04-12 05:37:11] P2 E30/50 | T:0.008903(S:0.9875) V:0.053298(S:0.9244) LR:1.79e-05
72
+ [2026-04-12 05:40:56] P2 E40/50 | T:0.008178(S:0.9885) V:0.053040(S:0.9248) LR:5.68e-06
73
+ [2026-04-12 05:42:25] P2 E44/50 | T:0.008011(S:0.9888) V:0.052805(S:0.9251) LR:2.72e-06
74
+ [2026-04-12 05:44:41] P2 E50/50 | T:0.007865(S:0.9890) V:0.052993(S:0.9249) LR:1.00e-06
75
+ [2026-04-12 05:44:41] Phase 2 done. Best val loss: 0.052805
76
+ [2026-04-12 05:44:41] Pong model: 2.3 MB
77
+ [2026-04-12 05:44:41] Training complete!