Add files using upload-large-folder tool
Browse files- rubik_r1_20261106011120/MANIFEST.json +0 -0
- rubik_r1_20261106011120/README.md +10 -0
- rubik_r1_20261106011120/data/dataset_state.json +46 -0
- rubik_r1_20261106011120/data/info.json +62 -0
- rubik_r1_20261106011120/logs/base_20261103_231337.log +0 -0
- rubik_r1_20261106011120/logs/base_20261103_233353.log +0 -0
- rubik_r1_20261106011120/logs/data_gpu0b.log +89 -0
- rubik_r1_20261106011120/logs/data_gpu1a.log +61 -0
- rubik_r1_20261106011120/logs/data_gpu1b.log +70 -0
- rubik_r1_20261106011120/logs/data_gpu2a.log +37 -0
- rubik_r1_20261106011120/logs/data_gpu3a.log +73 -0
- rubik_r1_20261106011120/logs/data_gpu3b.log +36 -0
- rubik_r1_20261106011120/logs/data_gpu5b.log +94 -0
- rubik_r1_20261106011120/logs/data_gpu6b.log +51 -0
- rubik_r1_20261106011120/logs/infer_rk_ffn_ckpt_004000.log +9 -0
- rubik_r1_20261106011120/logs/infer_rk_ffn_ckpt_008000.log +9 -0
- rubik_r1_20261106011120/logs/infer_rk_ffn_ckpt_012000.log +9 -0
- rubik_r1_20261106011120/logs/infer_rk_k_ckpt_004000.log +9 -0
- rubik_r1_20261106011120/logs/infer_rk_k_ckpt_008000.log +9 -0
- rubik_r1_20261106011120/logs/infer_rk_k_ckpt_012000.log +9 -0
- rubik_r1_20261106011120/logs/infer_rk_q_ckpt_004000.log +9 -0
- rubik_r1_20261106011120/logs/infer_rk_q_ckpt_012000.log +9 -0
- rubik_r1_20261106011120/logs/infer_rk_qkv_ckpt_004000.log +9 -0
- rubik_r1_20261106011120/logs/infer_rk_qkv_ckpt_008000.log +9 -0
- rubik_r1_20261106011120/logs/infer_rk_qkv_ckpt_012000.log +9 -0
- rubik_r1_20261106011120/logs/infer_rk_qkv_norm_ckpt_004000.log +9 -0
- rubik_r1_20261106011120/logs/infer_rk_qkv_norm_ckpt_008000.log +9 -0
- rubik_r1_20261106011120/logs/infer_rk_qkv_norm_ckpt_012000.log +9 -0
- rubik_r1_20261106011120/logs/infer_rk_v_ckpt_004000.log +9 -0
- rubik_r1_20261106011120/logs/infer_rk_v_norm_ckpt_012000.log +9 -0
- rubik_r1_20261106011120/logs/infer_rubik_base_ema_040000_ctx0.log +9 -0
- rubik_r1_20261106011120/logs/rk_q.log +271 -0
- rubik_r1_20261106011120/logs/rk_qkv.log +286 -0
- rubik_r1_20261106011120/logs/rk_qkv_norm.log +290 -0
- rubik_r1_20261106011120/logs/rk_v.log +274 -0
- rubik_r1_20261106011120/rk_ctrl/info.json +62 -0
- rubik_r1_20261106011120/rk_ctrl/log.jsonl +0 -0
- rubik_r1_20261106011120/rk_ffn/info.json +62 -0
- rubik_r1_20261106011120/rk_ffn/log.jsonl +0 -0
- rubik_r1_20261106011120/rk_q/info.json +62 -0
- rubik_r1_20261106011120/rk_q/log.jsonl +0 -0
- rubik_r1_20261106011120/rk_qkv/info.json +62 -0
- rubik_r1_20261106011120/rk_qkv/log.jsonl +0 -0
- rubik_r1_20261106011120/rk_qkv_norm/info.json +62 -0
- rubik_r1_20261106011120/rk_qkv_norm/log.jsonl +0 -0
- rubik_r1_20261106011120/rk_v/info.json +62 -0
- rubik_r1_20261106011120/rk_v/log.jsonl +0 -0
- rubik_r1_20261106011120/rk_v_norm/info.json +62 -0
- rubik_r1_20261106011120/rk_v_norm/log.jsonl +0 -0
- rubik_r1_20261106011120/rubik_base/log.jsonl +0 -0
rubik_r1_20261106011120/MANIFEST.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
rubik_r1_20261106011120/README.md
ADDED
|
@@ -0,0 +1,10 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# rubik_r1_20261106011120
|
| 2 |
+
|
| 3 |
+
Hand-back of the rubik_memory_remote package (2x2 Rubik's cube, visual memory in native fast weights): checkpoints, training logs and the generated latents of the test splits (no evaluation).
|
| 4 |
+
Host pluto-prod-lehany-dev-3dvwkl-38-0, GPUs -, gathered 2026-11-06 01:13:47.
|
| 5 |
+
|
| 6 |
+
Runs: rk_ctrl, rk_ffn, rk_k, rk_q, rk_qkv, rk_qkv_norm, rk_v, rk_v_norm, rubik_base
|
| 7 |
+
|
| 8 |
+
Inference (27): rk_ctrl/ckpt_004000, rk_ctrl/ckpt_008000, rk_ctrl/ckpt_012000, rk_ffn/ckpt_004000, rk_ffn/ckpt_008000, rk_ffn/ckpt_012000, rk_k/ckpt_004000, rk_k/ckpt_008000, rk_k/ckpt_012000, rk_q/ckpt_004000, rk_q/ckpt_008000, rk_q/ckpt_012000, rk_qkv/ckpt_004000, rk_qkv/ckpt_008000, rk_qkv/ckpt_012000, rk_qkv_norm/ckpt_004000, rk_qkv_norm/ckpt_008000, rk_qkv_norm/ckpt_012000, rk_v/ckpt_004000, rk_v/ckpt_008000, rk_v/ckpt_012000, rk_v_norm/ckpt_004000, rk_v_norm/ckpt_008000, rk_v_norm/ckpt_012000, rubik_base/ema_020000_ctx-1, rubik_base/ema_040000_ctx-1, rubik_base/ema_040000_ctx0
|
| 9 |
+
|
| 10 |
+
Every file with size and sha256: MANIFEST.json.
|
rubik_r1_20261106011120/data/dataset_state.json
ADDED
|
@@ -0,0 +1,46 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"counts": {
|
| 3 |
+
"test/n00": {
|
| 4 |
+
"meta": 64,
|
| 5 |
+
"latents": 64
|
| 6 |
+
},
|
| 7 |
+
"test/n02": {
|
| 8 |
+
"meta": 64,
|
| 9 |
+
"latents": 64
|
| 10 |
+
},
|
| 11 |
+
"test/n04": {
|
| 12 |
+
"meta": 64,
|
| 13 |
+
"latents": 64
|
| 14 |
+
},
|
| 15 |
+
"test/n08": {
|
| 16 |
+
"meta": 64,
|
| 17 |
+
"latents": 64
|
| 18 |
+
},
|
| 19 |
+
"test/n12": {
|
| 20 |
+
"meta": 64,
|
| 21 |
+
"latents": 64
|
| 22 |
+
},
|
| 23 |
+
"test/n16": {
|
| 24 |
+
"meta": 64,
|
| 25 |
+
"latents": 64
|
| 26 |
+
},
|
| 27 |
+
"test/n24": {
|
| 28 |
+
"meta": 64,
|
| 29 |
+
"latents": 64
|
| 30 |
+
},
|
| 31 |
+
"train": {
|
| 32 |
+
"meta": 60000,
|
| 33 |
+
"latents": 60000
|
| 34 |
+
}
|
| 35 |
+
},
|
| 36 |
+
"meta_sha256": {
|
| 37 |
+
"test/n00": "3dbd9f26dc0809c0d63ca66c562a013a5ecfa704bea449ef73bdaefd13018005",
|
| 38 |
+
"test/n02": "1b7fe79a7da6e4454d2460baffaf391ac576ac3e8ae7d20b4c3238c315646fdf",
|
| 39 |
+
"test/n04": "1a6bf54c7a27c40d956a2b06aae258d7d19b4e8a643fbb382cd4c0f5cac23fe1",
|
| 40 |
+
"test/n08": "8b6c1683545115f403c4134af2817be9777cc9d6d1b0820ea09663d6e3d79216",
|
| 41 |
+
"test/n12": "abd884df096850c81bfb7dddc32b96d90a62dd4cd9770df03978052ac28179e9",
|
| 42 |
+
"test/n16": "3be609a9ffe13da317bae91fe7cca8f1e287e556f50f3d26b19105643b5a27d4",
|
| 43 |
+
"test/n24": "eed03df16a988c54bfce91eb503e212b1c93bacb42458cd4f37c677f021b7f88",
|
| 44 |
+
"train": "358e1d3b65ea52478d8c2103a3b093f81ac196445d3599948e5123070b00ab2d"
|
| 45 |
+
}
|
| 46 |
+
}
|
rubik_r1_20261106011120/data/info.json
ADDED
|
@@ -0,0 +1,62 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"source": "auto_research/visual/datagen/rubik.py + build_rubik.py (torch ray-cast renderer, 4x4 SSAA)",
|
| 3 |
+
"size": [
|
| 4 |
+
256,
|
| 5 |
+
192
|
| 6 |
+
],
|
| 7 |
+
"fps": 16,
|
| 8 |
+
"n": 2,
|
| 9 |
+
"grid": 3,
|
| 10 |
+
"colors": 8,
|
| 11 |
+
"seed": 0,
|
| 12 |
+
"train_moves": [
|
| 13 |
+
0,
|
| 14 |
+
1,
|
| 15 |
+
2,
|
| 16 |
+
3,
|
| 17 |
+
4,
|
| 18 |
+
5,
|
| 19 |
+
6,
|
| 20 |
+
7,
|
| 21 |
+
8
|
| 22 |
+
],
|
| 23 |
+
"test_moves": [
|
| 24 |
+
0,
|
| 25 |
+
2,
|
| 26 |
+
4,
|
| 27 |
+
8,
|
| 28 |
+
12,
|
| 29 |
+
16,
|
| 30 |
+
24
|
| 31 |
+
],
|
| 32 |
+
"train_num": 60000,
|
| 33 |
+
"test_num": 64,
|
| 34 |
+
"actions": [
|
| 35 |
+
"hold",
|
| 36 |
+
"U",
|
| 37 |
+
"U'",
|
| 38 |
+
"D",
|
| 39 |
+
"D'",
|
| 40 |
+
"L",
|
| 41 |
+
"L'",
|
| 42 |
+
"R",
|
| 43 |
+
"R'",
|
| 44 |
+
"F",
|
| 45 |
+
"F'",
|
| 46 |
+
"B",
|
| 47 |
+
"B'",
|
| 48 |
+
"x",
|
| 49 |
+
"x'",
|
| 50 |
+
"y",
|
| 51 |
+
"y'",
|
| 52 |
+
"z",
|
| 53 |
+
"z'"
|
| 54 |
+
],
|
| 55 |
+
"display": [
|
| 56 |
+
"y",
|
| 57 |
+
"y",
|
| 58 |
+
"x"
|
| 59 |
+
],
|
| 60 |
+
"episode": "chunk 0 hold, chunks 1-3 display (y y x), N random quarter face turns (never the inverse of the previous one), final display (y y x); K = 7 + N chunks; evaluate the end frames of the last scramble chunk and the 3 final display chunks",
|
| 61 |
+
"latents": "float16 [16, 5K, 24, 32], Wan2.1 VAE mode, (mu - latents_mean) / latents_std"
|
| 62 |
+
}
|
rubik_r1_20261106011120/logs/base_20261103_231337.log
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
rubik_r1_20261106011120/logs/base_20261103_233353.log
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
rubik_r1_20261106011120/logs/data_gpu0b.log
ADDED
|
@@ -0,0 +1,89 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 2 |
+
```
|
| 3 |
+
pip install accelerate
|
| 4 |
+
```
|
| 5 |
+
.
|
| 6 |
+
2026-11-03 19:57:24 [shard 13/128] 476 episodes in 33.1 min (0.24/s)
|
| 7 |
+
2026-11-03 20:31:07 [shard 24/128] 476 episodes in 33.7 min (0.24/s)
|
| 8 |
+
2026-11-03 21:04:19 [shard 46/128] 476 episodes in 33.2 min (0.24/s)
|
| 9 |
+
2026-11-03 21:36:55 [shard 64/128] 469 episodes in 32.6 min (0.24/s)
|
| 10 |
+
2026-11-03 22:09:10 [shard 83/128] 469 episodes in 32.3 min (0.24/s)
|
| 11 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 12 |
+
```
|
| 13 |
+
pip install accelerate
|
| 14 |
+
```
|
| 15 |
+
.
|
| 16 |
+
2026-11-03 22:44:15 [shard 105/128] 165 episodes in 10.7 min (0.26/s)
|
| 17 |
+
2026-11-03 23:04:44 [shard 120/128] 468 episodes in 20.5 min (0.38/s)
|
| 18 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 19 |
+
```
|
| 20 |
+
pip install accelerate
|
| 21 |
+
```
|
| 22 |
+
.
|
| 23 |
+
2026-11-03 23:33:38 [shard 75/128] 0 episodes in 0.0 min (0.00/s)
|
| 24 |
+
2026-11-03 23:33:38 [shard 78/128] 0 episodes in 0.0 min (0.00/s)
|
| 25 |
+
2026-11-03 23:33:38 [shard 80/128] 0 episodes in 0.0 min (0.00/s)
|
| 26 |
+
2026-11-03 23:33:38 [shard 83/128] 0 episodes in 0.0 min (0.00/s)
|
| 27 |
+
2026-11-03 23:33:38 [shard 85/128] 0 episodes in 0.0 min (0.00/s)
|
| 28 |
+
2026-11-03 23:33:38 [shard 87/128] 0 episodes in 0.0 min (0.00/s)
|
| 29 |
+
2026-11-03 23:33:38 [shard 89/128] 0 episodes in 0.0 min (0.00/s)
|
| 30 |
+
2026-11-03 23:33:38 [shard 90/128] 0 episodes in 0.0 min (0.00/s)
|
| 31 |
+
2026-11-03 23:33:38 [shard 92/128] 0 episodes in 0.0 min (0.00/s)
|
| 32 |
+
2026-11-03 23:33:38 [shard 94/128] 0 episodes in 0.0 min (0.00/s)
|
| 33 |
+
2026-11-03 23:33:38 [shard 96/128] 0 episodes in 0.0 min (0.00/s)
|
| 34 |
+
2026-11-03 23:33:38 [shard 98/128] 0 episodes in 0.0 min (0.00/s)
|
| 35 |
+
2026-11-03 23:33:38 [shard 100/128] 0 episodes in 0.0 min (0.00/s)
|
| 36 |
+
2026-11-03 23:33:38 [shard 102/128] 0 episodes in 0.0 min (0.00/s)
|
| 37 |
+
2026-11-03 23:33:38 [shard 104/128] 0 episodes in 0.0 min (0.00/s)
|
| 38 |
+
2026-11-03 23:33:38 [shard 106/128] 0 episodes in 0.0 min (0.00/s)
|
| 39 |
+
2026-11-03 23:33:38 [shard 108/128] 0 episodes in 0.0 min (0.00/s)
|
| 40 |
+
2026-11-03 23:33:38 [shard 110/128] 0 episodes in 0.0 min (0.00/s)
|
| 41 |
+
2026-11-03 23:33:38 [shard 112/128] 0 episodes in 0.0 min (0.00/s)
|
| 42 |
+
2026-11-03 23:33:38 [shard 114/128] 0 episodes in 0.0 min (0.00/s)
|
| 43 |
+
2026-11-03 23:33:38 [shard 115/128] 0 episodes in 0.0 min (0.00/s)
|
| 44 |
+
2026-11-03 23:33:38 [shard 117/128] 0 episodes in 0.0 min (0.00/s)
|
| 45 |
+
2026-11-03 23:33:38 [shard 119/128] 0 episodes in 0.0 min (0.00/s)
|
| 46 |
+
2026-11-03 23:33:38 [shard 121/128] 0 episodes in 0.0 min (0.00/s)
|
| 47 |
+
2026-11-03 23:33:38 [shard 123/128] 0 episodes in 0.0 min (0.00/s)
|
| 48 |
+
2026-11-03 23:33:38 [shard 125/128] 0 episodes in 0.0 min (0.00/s)
|
| 49 |
+
2026-11-03 23:33:38 [shard 127/128] 0 episodes in 0.0 min (0.00/s)
|
| 50 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 51 |
+
```
|
| 52 |
+
pip install accelerate
|
| 53 |
+
```
|
| 54 |
+
.
|
| 55 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 56 |
+
```
|
| 57 |
+
pip install accelerate
|
| 58 |
+
```
|
| 59 |
+
.
|
| 60 |
+
2026-11-05 12:11:33 [shard 19/128] 0 episodes in 0.0 min (0.00/s)
|
| 61 |
+
2026-11-05 12:11:33 [shard 24/128] 0 episodes in 0.0 min (0.00/s)
|
| 62 |
+
2026-11-05 12:11:33 [shard 30/128] 0 episodes in 0.0 min (0.00/s)
|
| 63 |
+
2026-11-05 12:11:33 [shard 36/128] 0 episodes in 0.0 min (0.00/s)
|
| 64 |
+
2026-11-05 12:11:33 [shard 45/128] 0 episodes in 0.0 min (0.00/s)
|
| 65 |
+
2026-11-05 12:11:33 [shard 56/128] 0 episodes in 0.0 min (0.00/s)
|
| 66 |
+
2026-11-05 12:11:33 [shard 68/128] 0 episodes in 0.0 min (0.00/s)
|
| 67 |
+
2026-11-05 12:11:33 [shard 77/128] 0 episodes in 0.0 min (0.00/s)
|
| 68 |
+
2026-11-05 12:11:33 [shard 86/128] 0 episodes in 0.0 min (0.00/s)
|
| 69 |
+
2026-11-05 12:11:33 [shard 95/128] 0 episodes in 0.0 min (0.00/s)
|
| 70 |
+
2026-11-05 12:11:33 [shard 104/128] 0 episodes in 0.0 min (0.00/s)
|
| 71 |
+
2026-11-05 12:11:33 [shard 115/128] 0 episodes in 0.0 min (0.00/s)
|
| 72 |
+
2026-11-05 12:11:34 [shard 123/128] 0 episodes in 0.0 min (0.00/s)
|
| 73 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 74 |
+
```
|
| 75 |
+
pip install accelerate
|
| 76 |
+
```
|
| 77 |
+
.
|
| 78 |
+
2026-11-05 17:49:23 [shard 35/128] 0 episodes in 0.0 min (0.00/s)
|
| 79 |
+
2026-11-05 17:49:23 [shard 42/128] 0 episodes in 0.0 min (0.00/s)
|
| 80 |
+
2026-11-05 17:49:23 [shard 49/128] 0 episodes in 0.0 min (0.00/s)
|
| 81 |
+
2026-11-05 17:49:23 [shard 57/128] 0 episodes in 0.0 min (0.00/s)
|
| 82 |
+
2026-11-05 17:49:23 [shard 65/128] 0 episodes in 0.0 min (0.00/s)
|
| 83 |
+
2026-11-05 17:49:23 [shard 76/128] 0 episodes in 0.0 min (0.00/s)
|
| 84 |
+
2026-11-05 17:49:23 [shard 83/128] 0 episodes in 0.0 min (0.00/s)
|
| 85 |
+
2026-11-05 17:49:23 [shard 90/128] 0 episodes in 0.0 min (0.00/s)
|
| 86 |
+
2026-11-05 17:49:23 [shard 98/128] 0 episodes in 0.0 min (0.00/s)
|
| 87 |
+
2026-11-05 17:49:23 [shard 105/128] 0 episodes in 0.0 min (0.00/s)
|
| 88 |
+
2026-11-05 17:49:23 [shard 113/128] 0 episodes in 0.0 min (0.00/s)
|
| 89 |
+
2026-11-05 17:49:23 [shard 120/128] 0 episodes in 0.0 min (0.00/s)
|
rubik_r1_20261106011120/logs/data_gpu1a.log
ADDED
|
@@ -0,0 +1,61 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 2 |
+
```
|
| 3 |
+
pip install accelerate
|
| 4 |
+
```
|
| 5 |
+
.
|
| 6 |
+
2026-11-03 19:58:19 [shard 11/128] 476 episodes in 34.0 min (0.23/s)
|
| 7 |
+
2026-11-03 20:32:48 [shard 29/128] 476 episodes in 34.5 min (0.23/s)
|
| 8 |
+
2026-11-03 21:06:52 [shard 52/128] 476 episodes in 34.1 min (0.23/s)
|
| 9 |
+
2026-11-03 21:39:15 [shard 68/128] 469 episodes in 32.4 min (0.24/s)
|
| 10 |
+
2026-11-03 22:12:03 [shard 92/128] 469 episodes in 32.8 min (0.24/s)
|
| 11 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 12 |
+
```
|
| 13 |
+
pip install accelerate
|
| 14 |
+
```
|
| 15 |
+
.
|
| 16 |
+
2026-11-03 22:51:45 [shard 107/128] 257 episodes in 18.2 min (0.24/s)
|
| 17 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 18 |
+
```
|
| 19 |
+
pip install accelerate
|
| 20 |
+
```
|
| 21 |
+
.
|
| 22 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 23 |
+
```
|
| 24 |
+
pip install accelerate
|
| 25 |
+
```
|
| 26 |
+
.
|
| 27 |
+
2026-11-04 06:08:11 [shard 0/128] 0 episodes in 0.0 min (0.00/s)
|
| 28 |
+
2026-11-04 06:08:11 [shard 1/128] 0 episodes in 0.0 min (0.00/s)
|
| 29 |
+
2026-11-04 06:08:11 [shard 2/128] 0 episodes in 0.0 min (0.00/s)
|
| 30 |
+
2026-11-04 06:08:11 [shard 3/128] 0 episodes in 0.0 min (0.00/s)
|
| 31 |
+
2026-11-04 06:08:11 [shard 4/128] 0 episodes in 0.0 min (0.00/s)
|
| 32 |
+
2026-11-04 06:08:11 [shard 6/128] 0 episodes in 0.0 min (0.00/s)
|
| 33 |
+
2026-11-04 06:08:11 [shard 9/128] 0 episodes in 0.0 min (0.00/s)
|
| 34 |
+
2026-11-04 06:08:11 [shard 13/128] 0 episodes in 0.0 min (0.00/s)
|
| 35 |
+
2026-11-04 06:08:11 [shard 17/128] 0 episodes in 0.0 min (0.00/s)
|
| 36 |
+
2026-11-04 06:08:11 [shard 22/128] 0 episodes in 0.0 min (0.00/s)
|
| 37 |
+
2026-11-04 06:08:11 [shard 28/128] 0 episodes in 0.0 min (0.00/s)
|
| 38 |
+
2026-11-04 06:08:12 [shard 33/128] 0 episodes in 0.0 min (0.00/s)
|
| 39 |
+
2026-11-04 06:08:12 [shard 39/128] 0 episodes in 0.0 min (0.00/s)
|
| 40 |
+
2026-11-04 06:08:12 [shard 44/128] 0 episodes in 0.0 min (0.00/s)
|
| 41 |
+
2026-11-04 06:08:12 [shard 49/128] 0 episodes in 0.0 min (0.00/s)
|
| 42 |
+
2026-11-04 06:08:12 [shard 56/128] 0 episodes in 0.0 min (0.00/s)
|
| 43 |
+
2026-11-04 06:08:12 [shard 61/128] 0 episodes in 0.0 min (0.00/s)
|
| 44 |
+
2026-11-04 06:08:12 [shard 68/128] 0 episodes in 0.0 min (0.00/s)
|
| 45 |
+
2026-11-04 06:08:12 [shard 74/128] 0 episodes in 0.0 min (0.00/s)
|
| 46 |
+
2026-11-04 06:08:12 [shard 82/128] 0 episodes in 0.0 min (0.00/s)
|
| 47 |
+
2026-11-04 06:08:12 [shard 89/128] 0 episodes in 0.0 min (0.00/s)
|
| 48 |
+
2026-11-04 06:08:12 [shard 95/128] 0 episodes in 0.0 min (0.00/s)
|
| 49 |
+
2026-11-04 06:08:12 [shard 103/128] 0 episodes in 0.0 min (0.00/s)
|
| 50 |
+
2026-11-04 06:08:12 [shard 111/128] 0 episodes in 0.0 min (0.00/s)
|
| 51 |
+
2026-11-04 06:08:12 [shard 119/128] 0 episodes in 0.0 min (0.00/s)
|
| 52 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 53 |
+
```
|
| 54 |
+
pip install accelerate
|
| 55 |
+
```
|
| 56 |
+
.
|
| 57 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 58 |
+
```
|
| 59 |
+
pip install accelerate
|
| 60 |
+
```
|
| 61 |
+
.
|
rubik_r1_20261106011120/logs/data_gpu1b.log
ADDED
|
@@ -0,0 +1,70 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 2 |
+
```
|
| 3 |
+
pip install accelerate
|
| 4 |
+
```
|
| 5 |
+
.
|
| 6 |
+
2026-11-03 19:58:02 [shard 6/128] 476 episodes in 33.8 min (0.23/s)
|
| 7 |
+
2026-11-03 20:31:51 [shard 27/128] 476 episodes in 33.8 min (0.23/s)
|
| 8 |
+
2026-11-03 21:06:03 [shard 50/128] 476 episodes in 34.2 min (0.23/s)
|
| 9 |
+
2026-11-03 21:38:22 [shard 67/128] 469 episodes in 32.3 min (0.24/s)
|
| 10 |
+
2026-11-03 22:11:33 [shard 88/128] 469 episodes in 33.2 min (0.24/s)
|
| 11 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 12 |
+
```
|
| 13 |
+
pip install accelerate
|
| 14 |
+
```
|
| 15 |
+
.
|
| 16 |
+
2026-11-03 22:53:53 [shard 115/128] 395 episodes in 20.3 min (0.32/s)
|
| 17 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 18 |
+
```
|
| 19 |
+
pip install accelerate
|
| 20 |
+
```
|
| 21 |
+
.
|
| 22 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 23 |
+
```
|
| 24 |
+
pip install accelerate
|
| 25 |
+
```
|
| 26 |
+
.
|
| 27 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 28 |
+
```
|
| 29 |
+
pip install accelerate
|
| 30 |
+
```
|
| 31 |
+
.
|
| 32 |
+
2026-11-05 12:11:33 [shard 0/128] 0 episodes in 0.0 min (0.00/s)
|
| 33 |
+
2026-11-05 12:11:33 [shard 1/128] 0 episodes in 0.0 min (0.00/s)
|
| 34 |
+
2026-11-05 12:11:33 [shard 2/128] 0 episodes in 0.0 min (0.00/s)
|
| 35 |
+
2026-11-05 12:11:33 [shard 3/128] 0 episodes in 0.0 min (0.00/s)
|
| 36 |
+
2026-11-05 12:11:33 [shard 4/128] 0 episodes in 0.0 min (0.00/s)
|
| 37 |
+
2026-11-05 12:11:33 [shard 5/128] 0 episodes in 0.0 min (0.00/s)
|
| 38 |
+
2026-11-05 12:11:33 [shard 6/128] 0 episodes in 0.0 min (0.00/s)
|
| 39 |
+
2026-11-05 12:11:33 [shard 7/128] 0 episodes in 0.0 min (0.00/s)
|
| 40 |
+
2026-11-05 12:11:33 [shard 8/128] 0 episodes in 0.0 min (0.00/s)
|
| 41 |
+
2026-11-05 12:11:33 [shard 9/128] 0 episodes in 0.0 min (0.00/s)
|
| 42 |
+
2026-11-05 12:11:33 [shard 10/128] 0 episodes in 0.0 min (0.00/s)
|
| 43 |
+
2026-11-05 12:11:33 [shard 11/128] 0 episodes in 0.0 min (0.00/s)
|
| 44 |
+
2026-11-05 12:11:33 [shard 12/128] 0 episodes in 0.0 min (0.00/s)
|
| 45 |
+
2026-11-05 12:11:33 [shard 13/128] 0 episodes in 0.0 min (0.00/s)
|
| 46 |
+
2026-11-05 12:11:33 [shard 14/128] 0 episodes in 0.0 min (0.00/s)
|
| 47 |
+
2026-11-05 12:11:33 [shard 15/128] 0 episodes in 0.0 min (0.00/s)
|
| 48 |
+
2026-11-05 12:11:33 [shard 16/128] 0 episodes in 0.0 min (0.00/s)
|
| 49 |
+
2026-11-05 12:11:33 [shard 17/128] 0 episodes in 0.0 min (0.00/s)
|
| 50 |
+
2026-11-05 12:11:33 [shard 20/128] 0 episodes in 0.0 min (0.00/s)
|
| 51 |
+
2026-11-05 12:11:33 [shard 23/128] 0 episodes in 0.0 min (0.00/s)
|
| 52 |
+
2026-11-05 12:11:33 [shard 29/128] 0 episodes in 0.0 min (0.00/s)
|
| 53 |
+
2026-11-05 12:11:33 [shard 34/128] 0 episodes in 0.0 min (0.00/s)
|
| 54 |
+
2026-11-05 12:11:33 [shard 40/128] 0 episodes in 0.0 min (0.00/s)
|
| 55 |
+
2026-11-05 12:11:33 [shard 46/128] 0 episodes in 0.0 min (0.00/s)
|
| 56 |
+
2026-11-05 12:11:33 [shard 53/128] 0 episodes in 0.0 min (0.00/s)
|
| 57 |
+
2026-11-05 12:11:33 [shard 62/128] 0 episodes in 0.0 min (0.00/s)
|
| 58 |
+
2026-11-05 12:11:33 [shard 70/128] 0 episodes in 0.0 min (0.00/s)
|
| 59 |
+
2026-11-05 12:11:33 [shard 79/128] 0 episodes in 0.0 min (0.00/s)
|
| 60 |
+
2026-11-05 12:11:33 [shard 88/128] 0 episodes in 0.0 min (0.00/s)
|
| 61 |
+
2026-11-05 12:11:33 [shard 97/128] 0 episodes in 0.0 min (0.00/s)
|
| 62 |
+
2026-11-05 12:11:33 [shard 106/128] 0 episodes in 0.0 min (0.00/s)
|
| 63 |
+
2026-11-05 12:11:33 [shard 112/128] 0 episodes in 0.0 min (0.00/s)
|
| 64 |
+
2026-11-05 12:11:34 [shard 121/128] 0 episodes in 0.0 min (0.00/s)
|
| 65 |
+
2026-11-05 12:11:34 [shard 127/128] 0 episodes in 0.0 min (0.00/s)
|
| 66 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 67 |
+
```
|
| 68 |
+
pip install accelerate
|
| 69 |
+
```
|
| 70 |
+
.
|
rubik_r1_20261106011120/logs/data_gpu2a.log
ADDED
|
@@ -0,0 +1,37 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 2 |
+
```
|
| 3 |
+
pip install accelerate
|
| 4 |
+
```
|
| 5 |
+
.
|
| 6 |
+
2026-11-03 19:58:41 [shard 1/128] 476 episodes in 34.4 min (0.23/s)
|
| 7 |
+
2026-11-03 20:33:05 [shard 30/128] 476 episodes in 34.4 min (0.23/s)
|
| 8 |
+
2026-11-03 21:08:03 [shard 53/128] 476 episodes in 35.0 min (0.23/s)
|
| 9 |
+
2026-11-03 21:41:29 [shard 69/128] 469 episodes in 33.4 min (0.23/s)
|
| 10 |
+
2026-11-03 22:15:02 [shard 93/128] 469 episodes in 33.5 min (0.23/s)
|
| 11 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 12 |
+
```
|
| 13 |
+
pip install accelerate
|
| 14 |
+
```
|
| 15 |
+
.
|
| 16 |
+
2026-11-03 22:49:35 [shard 109/128] 180 episodes in 16.0 min (0.19/s)
|
| 17 |
+
2026-11-03 23:10:53 [shard 125/128] 468 episodes in 21.3 min (0.37/s)
|
| 18 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 19 |
+
```
|
| 20 |
+
pip install accelerate
|
| 21 |
+
```
|
| 22 |
+
.
|
| 23 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 24 |
+
```
|
| 25 |
+
pip install accelerate
|
| 26 |
+
```
|
| 27 |
+
.
|
| 28 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 29 |
+
```
|
| 30 |
+
pip install accelerate
|
| 31 |
+
```
|
| 32 |
+
.
|
| 33 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 34 |
+
```
|
| 35 |
+
pip install accelerate
|
| 36 |
+
```
|
| 37 |
+
.
|
rubik_r1_20261106011120/logs/data_gpu3a.log
ADDED
|
@@ -0,0 +1,73 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 2 |
+
```
|
| 3 |
+
pip install accelerate
|
| 4 |
+
```
|
| 5 |
+
.
|
| 6 |
+
2026-11-03 19:57:51 [shard 5/128] 476 episodes in 33.6 min (0.24/s)
|
| 7 |
+
2026-11-03 20:31:34 [shard 26/128] 476 episodes in 33.7 min (0.24/s)
|
| 8 |
+
2026-11-03 21:05:38 [shard 48/128] 476 episodes in 34.1 min (0.23/s)
|
| 9 |
+
2026-11-03 21:38:49 [shard 66/128] 469 episodes in 33.2 min (0.24/s)
|
| 10 |
+
2026-11-03 22:12:34 [shard 89/128] 469 episodes in 33.7 min (0.23/s)
|
| 11 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 12 |
+
```
|
| 13 |
+
pip install accelerate
|
| 14 |
+
```
|
| 15 |
+
.
|
| 16 |
+
2026-11-03 22:48:56 [shard 108/128] 174 episodes in 15.4 min (0.19/s)
|
| 17 |
+
2026-11-03 23:06:28 [shard 123/128] 468 episodes in 17.5 min (0.44/s)
|
| 18 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 19 |
+
```
|
| 20 |
+
pip install accelerate
|
| 21 |
+
```
|
| 22 |
+
.
|
| 23 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 24 |
+
```
|
| 25 |
+
pip install accelerate
|
| 26 |
+
```
|
| 27 |
+
.
|
| 28 |
+
2026-11-04 06:08:11 [shard 18/128] 0 episodes in 0.0 min (0.00/s)
|
| 29 |
+
2026-11-04 06:08:11 [shard 24/128] 0 episodes in 0.0 min (0.00/s)
|
| 30 |
+
2026-11-04 06:08:12 [shard 30/128] 0 episodes in 0.0 min (0.00/s)
|
| 31 |
+
2026-11-04 06:08:12 [shard 36/128] 0 episodes in 0.0 min (0.00/s)
|
| 32 |
+
2026-11-04 06:08:12 [shard 41/128] 0 episodes in 0.0 min (0.00/s)
|
| 33 |
+
2026-11-04 06:08:12 [shard 46/128] 0 episodes in 0.0 min (0.00/s)
|
| 34 |
+
2026-11-04 06:08:12 [shard 52/128] 0 episodes in 0.0 min (0.00/s)
|
| 35 |
+
2026-11-04 06:08:12 [shard 58/128] 0 episodes in 0.0 min (0.00/s)
|
| 36 |
+
2026-11-04 06:08:12 [shard 65/128] 0 episodes in 0.0 min (0.00/s)
|
| 37 |
+
2026-11-04 06:08:12 [shard 72/128] 0 episodes in 0.0 min (0.00/s)
|
| 38 |
+
2026-11-04 06:08:12 [shard 80/128] 0 episodes in 0.0 min (0.00/s)
|
| 39 |
+
2026-11-04 06:08:12 [shard 88/128] 0 episodes in 0.0 min (0.00/s)
|
| 40 |
+
2026-11-04 06:08:12 [shard 96/128] 0 episodes in 0.0 min (0.00/s)
|
| 41 |
+
2026-11-04 06:08:12 [shard 104/128] 0 episodes in 0.0 min (0.00/s)
|
| 42 |
+
2026-11-04 06:08:12 [shard 112/128] 0 episodes in 0.0 min (0.00/s)
|
| 43 |
+
2026-11-04 06:08:12 [shard 121/128] 0 episodes in 0.0 min (0.00/s)
|
| 44 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 45 |
+
```
|
| 46 |
+
pip install accelerate
|
| 47 |
+
```
|
| 48 |
+
.
|
| 49 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 50 |
+
```
|
| 51 |
+
pip install accelerate
|
| 52 |
+
```
|
| 53 |
+
.
|
| 54 |
+
2026-11-05 17:49:23 [shard 11/128] 0 episodes in 0.0 min (0.00/s)
|
| 55 |
+
2026-11-05 17:49:23 [shard 13/128] 0 episodes in 0.0 min (0.00/s)
|
| 56 |
+
2026-11-05 17:49:23 [shard 16/128] 0 episodes in 0.0 min (0.00/s)
|
| 57 |
+
2026-11-05 17:49:23 [shard 19/128] 0 episodes in 0.0 min (0.00/s)
|
| 58 |
+
2026-11-05 17:49:23 [shard 22/128] 0 episodes in 0.0 min (0.00/s)
|
| 59 |
+
2026-11-05 17:49:23 [shard 26/128] 0 episodes in 0.0 min (0.00/s)
|
| 60 |
+
2026-11-05 17:49:23 [shard 31/128] 0 episodes in 0.0 min (0.00/s)
|
| 61 |
+
2026-11-05 17:49:23 [shard 38/128] 0 episodes in 0.0 min (0.00/s)
|
| 62 |
+
2026-11-05 17:49:23 [shard 44/128] 0 episodes in 0.0 min (0.00/s)
|
| 63 |
+
2026-11-05 17:49:23 [shard 52/128] 0 episodes in 0.0 min (0.00/s)
|
| 64 |
+
2026-11-05 17:49:23 [shard 60/128] 0 episodes in 0.0 min (0.00/s)
|
| 65 |
+
2026-11-05 17:49:23 [shard 67/128] 0 episodes in 0.0 min (0.00/s)
|
| 66 |
+
2026-11-05 17:49:23 [shard 73/128] 0 episodes in 0.0 min (0.00/s)
|
| 67 |
+
2026-11-05 17:49:23 [shard 82/128] 0 episodes in 0.0 min (0.00/s)
|
| 68 |
+
2026-11-05 17:49:23 [shard 89/128] 0 episodes in 0.0 min (0.00/s)
|
| 69 |
+
2026-11-05 17:49:23 [shard 95/128] 0 episodes in 0.0 min (0.00/s)
|
| 70 |
+
2026-11-05 17:49:23 [shard 104/128] 0 episodes in 0.0 min (0.00/s)
|
| 71 |
+
2026-11-05 17:49:23 [shard 110/128] 0 episodes in 0.0 min (0.00/s)
|
| 72 |
+
2026-11-05 17:49:23 [shard 117/128] 0 episodes in 0.0 min (0.00/s)
|
| 73 |
+
2026-11-05 17:49:23 [shard 127/128] 0 episodes in 0.0 min (0.00/s)
|
rubik_r1_20261106011120/logs/data_gpu3b.log
ADDED
|
@@ -0,0 +1,36 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 2 |
+
```
|
| 3 |
+
pip install accelerate
|
| 4 |
+
```
|
| 5 |
+
.
|
| 6 |
+
2026-11-03 19:58:06 [shard 4/128] 476 episodes in 33.8 min (0.23/s)
|
| 7 |
+
2026-11-03 20:31:36 [shard 28/128] 476 episodes in 33.5 min (0.24/s)
|
| 8 |
+
2026-11-03 21:05:16 [shard 49/128] 476 episodes in 33.7 min (0.24/s)
|
| 9 |
+
2026-11-03 21:37:55 [shard 65/128] 469 episodes in 32.7 min (0.24/s)
|
| 10 |
+
2026-11-03 22:10:45 [shard 86/128] 469 episodes in 32.8 min (0.24/s)
|
| 11 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 12 |
+
```
|
| 13 |
+
pip install accelerate
|
| 14 |
+
```
|
| 15 |
+
.
|
| 16 |
+
2026-11-03 23:00:51 [shard 114/128] 341 episodes in 27.3 min (0.21/s)
|
| 17 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 18 |
+
```
|
| 19 |
+
pip install accelerate
|
| 20 |
+
```
|
| 21 |
+
.
|
| 22 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 23 |
+
```
|
| 24 |
+
pip install accelerate
|
| 25 |
+
```
|
| 26 |
+
.
|
| 27 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 28 |
+
```
|
| 29 |
+
pip install accelerate
|
| 30 |
+
```
|
| 31 |
+
.
|
| 32 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 33 |
+
```
|
| 34 |
+
pip install accelerate
|
| 35 |
+
```
|
| 36 |
+
.
|
rubik_r1_20261106011120/logs/data_gpu5b.log
ADDED
|
@@ -0,0 +1,94 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 2 |
+
```
|
| 3 |
+
pip install accelerate
|
| 4 |
+
```
|
| 5 |
+
.
|
| 6 |
+
2026-11-03 19:45:22 [shard 9/128] 476 episodes in 21.1 min (0.38/s)
|
| 7 |
+
2026-11-03 20:06:19 [shard 17/128] 476 episodes in 20.9 min (0.38/s)
|
| 8 |
+
2026-11-03 20:27:25 [shard 33/128] 476 episodes in 21.1 min (0.38/s)
|
| 9 |
+
2026-11-03 20:48:09 [shard 42/128] 476 episodes in 20.7 min (0.38/s)
|
| 10 |
+
2026-11-03 21:08:58 [shard 56/128] 476 episodes in 20.8 min (0.38/s)
|
| 11 |
+
2026-11-03 21:29:10 [shard 72/128] 469 episodes in 20.2 min (0.39/s)
|
| 12 |
+
2026-11-03 21:49:38 [shard 79/128] 469 episodes in 20.5 min (0.38/s)
|
| 13 |
+
2026-11-03 22:10:07 [shard 95/128] 469 episodes in 20.5 min (0.38/s)
|
| 14 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 15 |
+
```
|
| 16 |
+
pip install accelerate
|
| 17 |
+
```
|
| 18 |
+
.
|
| 19 |
+
2026-11-03 22:47:57 [shard 110/128] 280 episodes in 14.4 min (0.32/s)
|
| 20 |
+
2026-11-03 23:00:09 [shard 122/128] 468 episodes in 12.2 min (0.64/s)
|
| 21 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 22 |
+
```
|
| 23 |
+
pip install accelerate
|
| 24 |
+
```
|
| 25 |
+
.
|
| 26 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 27 |
+
```
|
| 28 |
+
pip install accelerate
|
| 29 |
+
```
|
| 30 |
+
.
|
| 31 |
+
2026-11-04 06:08:11 [shard 11/128] 0 episodes in 0.0 min (0.00/s)
|
| 32 |
+
2026-11-04 06:08:11 [shard 15/128] 0 episodes in 0.0 min (0.00/s)
|
| 33 |
+
2026-11-04 06:08:11 [shard 20/128] 0 episodes in 0.0 min (0.00/s)
|
| 34 |
+
2026-11-04 06:08:11 [shard 25/128] 0 episodes in 0.0 min (0.00/s)
|
| 35 |
+
2026-11-04 06:08:11 [shard 29/128] 0 episodes in 0.0 min (0.00/s)
|
| 36 |
+
2026-11-04 06:08:12 [shard 35/128] 0 episodes in 0.0 min (0.00/s)
|
| 37 |
+
2026-11-04 06:08:12 [shard 40/128] 0 episodes in 0.0 min (0.00/s)
|
| 38 |
+
2026-11-04 06:08:12 [shard 45/128] 0 episodes in 0.0 min (0.00/s)
|
| 39 |
+
2026-11-04 06:08:12 [shard 50/128] 0 episodes in 0.0 min (0.00/s)
|
| 40 |
+
2026-11-04 06:08:12 [shard 55/128] 0 episodes in 0.0 min (0.00/s)
|
| 41 |
+
2026-11-04 06:08:12 [shard 60/128] 0 episodes in 0.0 min (0.00/s)
|
| 42 |
+
2026-11-04 06:08:12 [shard 67/128] 0 episodes in 0.0 min (0.00/s)
|
| 43 |
+
2026-11-04 06:08:12 [shard 73/128] 0 episodes in 0.0 min (0.00/s)
|
| 44 |
+
2026-11-04 06:08:12 [shard 81/128] 0 episodes in 0.0 min (0.00/s)
|
| 45 |
+
2026-11-04 06:08:12 [shard 87/128] 0 episodes in 0.0 min (0.00/s)
|
| 46 |
+
2026-11-04 06:08:12 [shard 94/128] 0 episodes in 0.0 min (0.00/s)
|
| 47 |
+
2026-11-04 06:08:12 [shard 102/128] 0 episodes in 0.0 min (0.00/s)
|
| 48 |
+
2026-11-04 06:08:12 [shard 110/128] 0 episodes in 0.0 min (0.00/s)
|
| 49 |
+
2026-11-04 06:08:12 [shard 118/128] 0 episodes in 0.0 min (0.00/s)
|
| 50 |
+
2026-11-04 06:08:12 [shard 125/128] 0 episodes in 0.0 min (0.00/s)
|
| 51 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 52 |
+
```
|
| 53 |
+
pip install accelerate
|
| 54 |
+
```
|
| 55 |
+
.
|
| 56 |
+
2026-11-05 12:11:33 [shard 18/128] 0 episodes in 0.0 min (0.00/s)
|
| 57 |
+
2026-11-05 12:11:33 [shard 22/128] 0 episodes in 0.0 min (0.00/s)
|
| 58 |
+
2026-11-05 12:11:33 [shard 26/128] 0 episodes in 0.0 min (0.00/s)
|
| 59 |
+
2026-11-05 12:11:33 [shard 32/128] 0 episodes in 0.0 min (0.00/s)
|
| 60 |
+
2026-11-05 12:11:33 [shard 38/128] 0 episodes in 0.0 min (0.00/s)
|
| 61 |
+
2026-11-05 12:11:33 [shard 47/128] 0 episodes in 0.0 min (0.00/s)
|
| 62 |
+
2026-11-05 12:11:33 [shard 55/128] 0 episodes in 0.0 min (0.00/s)
|
| 63 |
+
2026-11-05 12:11:33 [shard 63/128] 0 episodes in 0.0 min (0.00/s)
|
| 64 |
+
2026-11-05 12:11:33 [shard 72/128] 0 episodes in 0.0 min (0.00/s)
|
| 65 |
+
2026-11-05 12:11:33 [shard 81/128] 0 episodes in 0.0 min (0.00/s)
|
| 66 |
+
2026-11-05 12:11:33 [shard 90/128] 0 episodes in 0.0 min (0.00/s)
|
| 67 |
+
2026-11-05 12:11:33 [shard 99/128] 0 episodes in 0.0 min (0.00/s)
|
| 68 |
+
2026-11-05 12:11:33 [shard 108/128] 0 episodes in 0.0 min (0.00/s)
|
| 69 |
+
2026-11-05 12:11:34 [shard 117/128] 0 episodes in 0.0 min (0.00/s)
|
| 70 |
+
2026-11-05 12:11:34 [shard 125/128] 0 episodes in 0.0 min (0.00/s)
|
| 71 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 72 |
+
```
|
| 73 |
+
pip install accelerate
|
| 74 |
+
```
|
| 75 |
+
.
|
| 76 |
+
2026-11-05 17:49:23 [shard 15/128] 0 episodes in 0.0 min (0.00/s)
|
| 77 |
+
2026-11-05 17:49:23 [shard 20/128] 0 episodes in 0.0 min (0.00/s)
|
| 78 |
+
2026-11-05 17:49:23 [shard 23/128] 0 episodes in 0.0 min (0.00/s)
|
| 79 |
+
2026-11-05 17:49:23 [shard 27/128] 0 episodes in 0.0 min (0.00/s)
|
| 80 |
+
2026-11-05 17:49:23 [shard 32/128] 0 episodes in 0.0 min (0.00/s)
|
| 81 |
+
2026-11-05 17:49:23 [shard 39/128] 0 episodes in 0.0 min (0.00/s)
|
| 82 |
+
2026-11-05 17:49:23 [shard 46/128] 0 episodes in 0.0 min (0.00/s)
|
| 83 |
+
2026-11-05 17:49:23 [shard 53/128] 0 episodes in 0.0 min (0.00/s)
|
| 84 |
+
2026-11-05 17:49:23 [shard 59/128] 0 episodes in 0.0 min (0.00/s)
|
| 85 |
+
2026-11-05 17:49:23 [shard 64/128] 0 episodes in 0.0 min (0.00/s)
|
| 86 |
+
2026-11-05 17:49:23 [shard 71/128] 0 episodes in 0.0 min (0.00/s)
|
| 87 |
+
2026-11-05 17:49:23 [shard 77/128] 0 episodes in 0.0 min (0.00/s)
|
| 88 |
+
2026-11-05 17:49:23 [shard 85/128] 0 episodes in 0.0 min (0.00/s)
|
| 89 |
+
2026-11-05 17:49:23 [shard 92/128] 0 episodes in 0.0 min (0.00/s)
|
| 90 |
+
2026-11-05 17:49:23 [shard 99/128] 0 episodes in 0.0 min (0.00/s)
|
| 91 |
+
2026-11-05 17:49:23 [shard 107/128] 0 episodes in 0.0 min (0.00/s)
|
| 92 |
+
2026-11-05 17:49:23 [shard 114/128] 0 episodes in 0.0 min (0.00/s)
|
| 93 |
+
2026-11-05 17:49:23 [shard 118/128] 0 episodes in 0.0 min (0.00/s)
|
| 94 |
+
2026-11-05 17:49:23 [shard 124/128] 0 episodes in 0.0 min (0.00/s)
|
rubik_r1_20261106011120/logs/data_gpu6b.log
ADDED
|
@@ -0,0 +1,51 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 2 |
+
```
|
| 3 |
+
pip install accelerate
|
| 4 |
+
```
|
| 5 |
+
.
|
| 6 |
+
2026-11-03 19:51:13 [shard 12/128] 476 episodes in 26.9 min (0.29/s)
|
| 7 |
+
2026-11-03 20:17:43 [shard 23/128] 476 episodes in 26.5 min (0.30/s)
|
| 8 |
+
2026-11-03 20:45:05 [shard 39/128] 476 episodes in 27.4 min (0.29/s)
|
| 9 |
+
2026-11-03 21:12:27 [shard 55/128] 476 episodes in 27.4 min (0.29/s)
|
| 10 |
+
2026-11-03 21:38:22 [shard 75/128] 469 episodes in 25.9 min (0.30/s)
|
| 11 |
+
2026-11-03 22:04:18 [shard 87/128] 469 episodes in 25.9 min (0.30/s)
|
| 12 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 13 |
+
```
|
| 14 |
+
pip install accelerate
|
| 15 |
+
```
|
| 16 |
+
.
|
| 17 |
+
2026-11-03 22:33:33 [shard 90/128] 0 episodes in 0.0 min (0.00/s)
|
| 18 |
+
2026-11-03 22:33:33 [shard 97/128] 0 episodes in 0.0 min (0.00/s)
|
| 19 |
+
2026-11-03 22:41:33 [shard 102/128] 123 episodes in 8.0 min (0.26/s)
|
| 20 |
+
2026-11-03 22:56:23 [shard 118/128] 468 episodes in 14.8 min (0.53/s)
|
| 21 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 22 |
+
```
|
| 23 |
+
pip install accelerate
|
| 24 |
+
```
|
| 25 |
+
.
|
| 26 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 27 |
+
```
|
| 28 |
+
pip install accelerate
|
| 29 |
+
```
|
| 30 |
+
.
|
| 31 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 32 |
+
```
|
| 33 |
+
pip install accelerate
|
| 34 |
+
```
|
| 35 |
+
.
|
| 36 |
+
2026-11-05 12:11:33 [shard 27/128] 0 episodes in 0.0 min (0.00/s)
|
| 37 |
+
2026-11-05 12:11:33 [shard 33/128] 0 episodes in 0.0 min (0.00/s)
|
| 38 |
+
2026-11-05 12:11:33 [shard 41/128] 0 episodes in 0.0 min (0.00/s)
|
| 39 |
+
2026-11-05 12:11:33 [shard 49/128] 0 episodes in 0.0 min (0.00/s)
|
| 40 |
+
2026-11-05 12:11:33 [shard 58/128] 0 episodes in 0.0 min (0.00/s)
|
| 41 |
+
2026-11-05 12:11:33 [shard 66/128] 0 episodes in 0.0 min (0.00/s)
|
| 42 |
+
2026-11-05 12:11:33 [shard 75/128] 0 episodes in 0.0 min (0.00/s)
|
| 43 |
+
2026-11-05 12:11:33 [shard 84/128] 0 episodes in 0.0 min (0.00/s)
|
| 44 |
+
2026-11-05 12:11:33 [shard 93/128] 0 episodes in 0.0 min (0.00/s)
|
| 45 |
+
2026-11-05 12:11:33 [shard 102/128] 0 episodes in 0.0 min (0.00/s)
|
| 46 |
+
2026-11-05 12:11:34 [shard 113/128] 0 episodes in 0.0 min (0.00/s)
|
| 47 |
+
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with:
|
| 48 |
+
```
|
| 49 |
+
pip install accelerate
|
| 50 |
+
```
|
| 51 |
+
.
|
rubik_r1_20261106011120/logs/infer_rk_ffn_ckpt_004000.log
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
infer /mnt/localssd/lehany/wv/E1_rubik_20261104031751/rk_ffn/ckpt_004000.pt (ema, step 4000) | max_ctx 0 | moves [0, 2, 4, 8, 12, 16, 24] | ['free', 'tf'] | 50 steps | seed 0 -> /mnt/localssd/lehany/wv/E1_rubik_20261104031751/inference/rk_ffn/ckpt_004000
|
| 2 |
+
test/n00: 64 episodes, K 7, free, tf saved (65 s)
|
| 3 |
+
test/n02: 64 episodes, K 9, free, tf saved (151 s)
|
| 4 |
+
test/n04: 64 episodes, K 11, free, tf saved (259 s)
|
| 5 |
+
test/n08: 64 episodes, K 15, free, tf saved (409 s)
|
| 6 |
+
test/n12: 64 episodes, K 19, free, tf saved (603 s)
|
| 7 |
+
test/n16: 64 episodes, K 23, free, tf saved (839 s)
|
| 8 |
+
test/n24: 64 episodes, K 31, free, tf saved (1161 s)
|
| 9 |
+
done -> /mnt/localssd/lehany/wv/E1_rubik_20261104031751/inference/rk_ffn/ckpt_004000
|
rubik_r1_20261106011120/logs/infer_rk_ffn_ckpt_008000.log
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
infer /mnt/localssd/lehany/wv/E1_rubik_20261104031751/rk_ffn/ckpt_008000.pt (ema, step 8000) | max_ctx 0 | moves [0, 2, 4, 8, 12, 16, 24] | ['free', 'tf'] | 50 steps | seed 0 -> /mnt/localssd/lehany/wv/E1_rubik_20261104031751/inference/rk_ffn/ckpt_008000
|
| 2 |
+
test/n00: 64 episodes, K 7, free, tf saved (66 s)
|
| 3 |
+
test/n02: 64 episodes, K 9, free, tf saved (152 s)
|
| 4 |
+
test/n04: 64 episodes, K 11, free, tf saved (260 s)
|
| 5 |
+
test/n08: 64 episodes, K 15, free, tf saved (412 s)
|
| 6 |
+
test/n12: 64 episodes, K 19, free, tf saved (606 s)
|
| 7 |
+
test/n16: 64 episodes, K 23, free, tf saved (843 s)
|
| 8 |
+
test/n24: 64 episodes, K 31, free, tf saved (1166 s)
|
| 9 |
+
done -> /mnt/localssd/lehany/wv/E1_rubik_20261104031751/inference/rk_ffn/ckpt_008000
|
rubik_r1_20261106011120/logs/infer_rk_ffn_ckpt_012000.log
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
infer /mnt/localssd/lehany/wv/E1_rubik_20261104031751/rk_ffn/ckpt_012000.pt (ema, step 12000) | max_ctx 0 | moves [0, 2, 4, 8, 12, 16, 24] | ['free', 'tf'] | 50 steps | seed 0 -> /mnt/localssd/lehany/wv/E1_rubik_20261104031751/inference/rk_ffn/ckpt_012000
|
| 2 |
+
test/n00: 64 episodes, K 7, free, tf saved (65 s)
|
| 3 |
+
test/n02: 64 episodes, K 9, free, tf saved (152 s)
|
| 4 |
+
test/n04: 64 episodes, K 11, free, tf saved (260 s)
|
| 5 |
+
test/n08: 64 episodes, K 15, free, tf saved (411 s)
|
| 6 |
+
test/n12: 64 episodes, K 19, free, tf saved (606 s)
|
| 7 |
+
test/n16: 64 episodes, K 23, free, tf saved (843 s)
|
| 8 |
+
test/n24: 64 episodes, K 31, free, tf saved (1166 s)
|
| 9 |
+
done -> /mnt/localssd/lehany/wv/E1_rubik_20261104031751/inference/rk_ffn/ckpt_012000
|
rubik_r1_20261106011120/logs/infer_rk_k_ckpt_004000.log
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
infer /mnt/localssd/lehany/wv/E1_rubik_20261104031751/rk_k/ckpt_004000.pt (ema, step 4000) | max_ctx 0 | moves [0, 2, 4, 8, 12, 16, 24] | ['free', 'tf'] | 50 steps | seed 0 -> /mnt/localssd/lehany/wv/E1_rubik_20261104031751/inference/rk_k/ckpt_004000
|
| 2 |
+
test/n00: 64 episodes, K 7, free, tf saved (66 s)
|
| 3 |
+
test/n02: 64 episodes, K 9, free, tf saved (151 s)
|
| 4 |
+
test/n04: 64 episodes, K 11, free, tf saved (257 s)
|
| 5 |
+
test/n08: 64 episodes, K 15, free, tf saved (405 s)
|
| 6 |
+
test/n12: 64 episodes, K 19, free, tf saved (596 s)
|
| 7 |
+
test/n16: 64 episodes, K 23, free, tf saved (828 s)
|
| 8 |
+
test/n24: 64 episodes, K 31, free, tf saved (1145 s)
|
| 9 |
+
done -> /mnt/localssd/lehany/wv/E1_rubik_20261104031751/inference/rk_k/ckpt_004000
|
rubik_r1_20261106011120/logs/infer_rk_k_ckpt_008000.log
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
infer /mnt/localssd/lehany/wv/E1_rubik_20261104031751/rk_k/ckpt_008000.pt (ema, step 8000) | max_ctx 0 | moves [0, 2, 4, 8, 12, 16, 24] | ['free', 'tf'] | 50 steps | seed 0 -> /mnt/localssd/lehany/wv/E1_rubik_20261104031751/inference/rk_k/ckpt_008000
|
| 2 |
+
test/n00: 64 episodes, K 7, free, tf saved (64 s)
|
| 3 |
+
test/n02: 64 episodes, K 9, free, tf saved (149 s)
|
| 4 |
+
test/n04: 64 episodes, K 11, free, tf saved (255 s)
|
| 5 |
+
test/n08: 64 episodes, K 15, free, tf saved (404 s)
|
| 6 |
+
test/n12: 64 episodes, K 19, free, tf saved (595 s)
|
| 7 |
+
test/n16: 64 episodes, K 23, free, tf saved (828 s)
|
| 8 |
+
test/n24: 64 episodes, K 31, free, tf saved (1145 s)
|
| 9 |
+
done -> /mnt/localssd/lehany/wv/E1_rubik_20261104031751/inference/rk_k/ckpt_008000
|
rubik_r1_20261106011120/logs/infer_rk_k_ckpt_012000.log
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
infer /mnt/localssd/lehany/wv/E1_rubik_20261104031751/rk_k/ckpt_012000.pt (ema, step 12000) | max_ctx 0 | moves [0, 2, 4, 8, 12, 16, 24] | ['free', 'tf'] | 50 steps | seed 0 -> /mnt/localssd/lehany/wv/E1_rubik_20261104031751/inference/rk_k/ckpt_012000
|
| 2 |
+
test/n00: 64 episodes, K 7, free, tf saved (64 s)
|
| 3 |
+
test/n02: 64 episodes, K 9, free, tf saved (149 s)
|
| 4 |
+
test/n04: 64 episodes, K 11, free, tf saved (255 s)
|
| 5 |
+
test/n08: 64 episodes, K 15, free, tf saved (404 s)
|
| 6 |
+
test/n12: 64 episodes, K 19, free, tf saved (594 s)
|
| 7 |
+
test/n16: 64 episodes, K 23, free, tf saved (827 s)
|
| 8 |
+
test/n24: 64 episodes, K 31, free, tf saved (1144 s)
|
| 9 |
+
done -> /mnt/localssd/lehany/wv/E1_rubik_20261104031751/inference/rk_k/ckpt_012000
|
rubik_r1_20261106011120/logs/infer_rk_q_ckpt_004000.log
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
infer /mnt/localssd/lehany/wv/E1_rubik_20261104031751/rk_q/ckpt_004000.pt (ema, step 4000) | max_ctx 0 | moves [0, 2, 4, 8, 12, 16, 24] | ['free', 'tf'] | 50 steps | seed 0 -> /mnt/localssd/lehany/wv/E1_rubik_20261104031751/inference/rk_q/ckpt_004000
|
| 2 |
+
test/n00: 64 episodes, K 7, free, tf saved (65 s)
|
| 3 |
+
test/n02: 64 episodes, K 9, free, tf saved (150 s)
|
| 4 |
+
test/n04: 64 episodes, K 11, free, tf saved (256 s)
|
| 5 |
+
test/n08: 64 episodes, K 15, free, tf saved (405 s)
|
| 6 |
+
test/n12: 64 episodes, K 19, free, tf saved (596 s)
|
| 7 |
+
test/n16: 64 episodes, K 23, free, tf saved (830 s)
|
| 8 |
+
test/n24: 64 episodes, K 31, free, tf saved (1148 s)
|
| 9 |
+
done -> /mnt/localssd/lehany/wv/E1_rubik_20261104031751/inference/rk_q/ckpt_004000
|
rubik_r1_20261106011120/logs/infer_rk_q_ckpt_012000.log
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
infer /mnt/localssd/lehany/wv/E1_rubik_20261104031751/rk_q/ckpt_012000.pt (ema, step 12000) | max_ctx 0 | moves [0, 2, 4, 8, 12, 16, 24] | ['free', 'tf'] | 50 steps | seed 0 -> /mnt/localssd/lehany/wv/E1_rubik_20261104031751/inference/rk_q/ckpt_012000
|
| 2 |
+
test/n00: 64 episodes, K 7, free, tf saved (65 s)
|
| 3 |
+
test/n02: 64 episodes, K 9, free, tf saved (150 s)
|
| 4 |
+
test/n04: 64 episodes, K 11, free, tf saved (256 s)
|
| 5 |
+
test/n08: 64 episodes, K 15, free, tf saved (404 s)
|
| 6 |
+
test/n12: 64 episodes, K 19, free, tf saved (595 s)
|
| 7 |
+
test/n16: 64 episodes, K 23, free, tf saved (828 s)
|
| 8 |
+
test/n24: 64 episodes, K 31, free, tf saved (1145 s)
|
| 9 |
+
done -> /mnt/localssd/lehany/wv/E1_rubik_20261104031751/inference/rk_q/ckpt_012000
|
rubik_r1_20261106011120/logs/infer_rk_qkv_ckpt_004000.log
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
infer /mnt/localssd/lehany/wv/E1_rubik_20261104031751/rk_qkv/ckpt_004000.pt (ema, step 4000) | max_ctx 0 | moves [0, 2, 4, 8, 12, 16, 24] | ['free', 'tf'] | 50 steps | seed 0 -> /mnt/localssd/lehany/wv/E1_rubik_20261104031751/inference/rk_qkv/ckpt_004000
|
| 2 |
+
test/n00: 64 episodes, K 7, free, tf saved (70 s)
|
| 3 |
+
test/n02: 64 episodes, K 9, free, tf saved (162 s)
|
| 4 |
+
test/n04: 64 episodes, K 11, free, tf saved (277 s)
|
| 5 |
+
test/n08: 64 episodes, K 15, free, tf saved (438 s)
|
| 6 |
+
test/n12: 64 episodes, K 19, free, tf saved (644 s)
|
| 7 |
+
test/n16: 64 episodes, K 23, free, tf saved (897 s)
|
| 8 |
+
test/n24: 64 episodes, K 31, free, tf saved (1241 s)
|
| 9 |
+
done -> /mnt/localssd/lehany/wv/E1_rubik_20261104031751/inference/rk_qkv/ckpt_004000
|
rubik_r1_20261106011120/logs/infer_rk_qkv_ckpt_008000.log
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
infer /mnt/localssd/lehany/wv/E1_rubik_20261104031751/rk_qkv/ckpt_008000.pt (ema, step 8000) | max_ctx 0 | moves [0, 2, 4, 8, 12, 16, 24] | ['free', 'tf'] | 50 steps | seed 0 -> /mnt/localssd/lehany/wv/E1_rubik_20261104031751/inference/rk_qkv/ckpt_008000
|
| 2 |
+
test/n00: 64 episodes, K 7, free, tf saved (69 s)
|
| 3 |
+
test/n02: 64 episodes, K 9, free, tf saved (161 s)
|
| 4 |
+
test/n04: 64 episodes, K 11, free, tf saved (276 s)
|
| 5 |
+
test/n08: 64 episodes, K 15, free, tf saved (436 s)
|
| 6 |
+
test/n12: 64 episodes, K 19, free, tf saved (642 s)
|
| 7 |
+
test/n16: 64 episodes, K 23, free, tf saved (895 s)
|
| 8 |
+
test/n24: 64 episodes, K 31, free, tf saved (1238 s)
|
| 9 |
+
done -> /mnt/localssd/lehany/wv/E1_rubik_20261104031751/inference/rk_qkv/ckpt_008000
|
rubik_r1_20261106011120/logs/infer_rk_qkv_ckpt_012000.log
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
infer /mnt/localssd/lehany/wv/E1_rubik_20261104031751/rk_qkv/ckpt_012000.pt (ema, step 12000) | max_ctx 0 | moves [0, 2, 4, 8, 12, 16, 24] | ['free', 'tf'] | 50 steps | seed 0 -> /mnt/localssd/lehany/wv/E1_rubik_20261104031751/inference/rk_qkv/ckpt_012000
|
| 2 |
+
test/n00: 64 episodes, K 7, free, tf saved (70 s)
|
| 3 |
+
test/n02: 64 episodes, K 9, free, tf saved (162 s)
|
| 4 |
+
test/n04: 64 episodes, K 11, free, tf saved (277 s)
|
| 5 |
+
test/n08: 64 episodes, K 15, free, tf saved (438 s)
|
| 6 |
+
test/n12: 64 episodes, K 19, free, tf saved (644 s)
|
| 7 |
+
test/n16: 64 episodes, K 23, free, tf saved (897 s)
|
| 8 |
+
test/n24: 64 episodes, K 31, free, tf saved (1241 s)
|
| 9 |
+
done -> /mnt/localssd/lehany/wv/E1_rubik_20261104031751/inference/rk_qkv/ckpt_012000
|
rubik_r1_20261106011120/logs/infer_rk_qkv_norm_ckpt_004000.log
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
infer /mnt/localssd/lehany/wv/E1_rubik_20261104031751/rk_qkv_norm/ckpt_004000.pt (ema, step 4000) | max_ctx 0 | moves [0, 2, 4, 8, 12, 16, 24] | ['free', 'tf'] | 50 steps | seed 0 -> /mnt/localssd/lehany/wv/E1_rubik_20261104031751/inference/rk_qkv_norm/ckpt_004000
|
| 2 |
+
test/n00: 64 episodes, K 7, free, tf saved (74 s)
|
| 3 |
+
test/n02: 64 episodes, K 9, free, tf saved (173 s)
|
| 4 |
+
test/n04: 64 episodes, K 11, free, tf saved (295 s)
|
| 5 |
+
test/n08: 64 episodes, K 15, free, tf saved (467 s)
|
| 6 |
+
test/n12: 64 episodes, K 19, free, tf saved (688 s)
|
| 7 |
+
test/n16: 64 episodes, K 23, free, tf saved (957 s)
|
| 8 |
+
test/n24: 64 episodes, K 31, free, tf saved (1325 s)
|
| 9 |
+
done -> /mnt/localssd/lehany/wv/E1_rubik_20261104031751/inference/rk_qkv_norm/ckpt_004000
|
rubik_r1_20261106011120/logs/infer_rk_qkv_norm_ckpt_008000.log
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
infer /mnt/localssd/lehany/wv/E1_rubik_20261104031751/rk_qkv_norm/ckpt_008000.pt (ema, step 8000) | max_ctx 0 | moves [0, 2, 4, 8, 12, 16, 24] | ['free', 'tf'] | 50 steps | seed 0 -> /mnt/localssd/lehany/wv/E1_rubik_20261104031751/inference/rk_qkv_norm/ckpt_008000
|
| 2 |
+
test/n00: 64 episodes, K 7, free, tf saved (74 s)
|
| 3 |
+
test/n02: 64 episodes, K 9, free, tf saved (173 s)
|
| 4 |
+
test/n04: 64 episodes, K 11, free, tf saved (295 s)
|
| 5 |
+
test/n08: 64 episodes, K 15, free, tf saved (467 s)
|
| 6 |
+
test/n12: 64 episodes, K 19, free, tf saved (688 s)
|
| 7 |
+
test/n16: 64 episodes, K 23, free, tf saved (957 s)
|
| 8 |
+
test/n24: 64 episodes, K 31, free, tf saved (1325 s)
|
| 9 |
+
done -> /mnt/localssd/lehany/wv/E1_rubik_20261104031751/inference/rk_qkv_norm/ckpt_008000
|
rubik_r1_20261106011120/logs/infer_rk_qkv_norm_ckpt_012000.log
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
infer /mnt/localssd/lehany/wv/E1_rubik_20261104031751/rk_qkv_norm/ckpt_012000.pt (ema, step 12000) | max_ctx 0 | moves [0, 2, 4, 8, 12, 16, 24] | ['free', 'tf'] | 50 steps | seed 0 -> /mnt/localssd/lehany/wv/E1_rubik_20261104031751/inference/rk_qkv_norm/ckpt_012000
|
| 2 |
+
test/n00: 64 episodes, K 7, free, tf saved (74 s)
|
| 3 |
+
test/n02: 64 episodes, K 9, free, tf saved (173 s)
|
| 4 |
+
test/n04: 64 episodes, K 11, free, tf saved (295 s)
|
| 5 |
+
test/n08: 64 episodes, K 15, free, tf saved (467 s)
|
| 6 |
+
test/n12: 64 episodes, K 19, free, tf saved (688 s)
|
| 7 |
+
test/n16: 64 episodes, K 23, free, tf saved (957 s)
|
| 8 |
+
test/n24: 64 episodes, K 31, free, tf saved (1325 s)
|
| 9 |
+
done -> /mnt/localssd/lehany/wv/E1_rubik_20261104031751/inference/rk_qkv_norm/ckpt_012000
|
rubik_r1_20261106011120/logs/infer_rk_v_ckpt_004000.log
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
infer /mnt/localssd/lehany/wv/E1_rubik_20261104031751/rk_v/ckpt_004000.pt (ema, step 4000) | max_ctx 0 | moves [0, 2, 4, 8, 12, 16, 24] | ['free', 'tf'] | 50 steps | seed 0 -> /mnt/localssd/lehany/wv/E1_rubik_20261104031751/inference/rk_v/ckpt_004000
|
| 2 |
+
test/n00: 64 episodes, K 7, free, tf saved (64 s)
|
| 3 |
+
test/n02: 64 episodes, K 9, free, tf saved (149 s)
|
| 4 |
+
test/n04: 64 episodes, K 11, free, tf saved (255 s)
|
| 5 |
+
test/n08: 64 episodes, K 15, free, tf saved (403 s)
|
| 6 |
+
test/n12: 64 episodes, K 19, free, tf saved (593 s)
|
| 7 |
+
test/n16: 64 episodes, K 23, free, tf saved (825 s)
|
| 8 |
+
test/n24: 64 episodes, K 31, free, tf saved (1141 s)
|
| 9 |
+
done -> /mnt/localssd/lehany/wv/E1_rubik_20261104031751/inference/rk_v/ckpt_004000
|
rubik_r1_20261106011120/logs/infer_rk_v_norm_ckpt_012000.log
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
infer /mnt/localssd/lehany/wv/E1_rubik_20261104031751/rk_v_norm/ckpt_012000.pt (ema, step 12000) | max_ctx 0 | moves [0, 2, 4, 8, 12, 16, 24] | ['free', 'tf'] | 50 steps | seed 0 -> /mnt/localssd/lehany/wv/E1_rubik_20261104031751/inference/rk_v_norm/ckpt_012000
|
| 2 |
+
test/n00: 64 episodes, K 7, free, tf saved (66 s)
|
| 3 |
+
test/n02: 64 episodes, K 9, free, tf saved (153 s)
|
| 4 |
+
test/n04: 64 episodes, K 11, free, tf saved (262 s)
|
| 5 |
+
test/n08: 64 episodes, K 15, free, tf saved (414 s)
|
| 6 |
+
test/n12: 64 episodes, K 19, free, tf saved (610 s)
|
| 7 |
+
test/n16: 64 episodes, K 23, free, tf saved (849 s)
|
| 8 |
+
test/n24: 64 episodes, K 31, free, tf saved (1175 s)
|
| 9 |
+
done -> /mnt/localssd/lehany/wv/E1_rubik_20261104031751/inference/rk_v_norm/ckpt_012000
|
rubik_r1_20261106011120/logs/infer_rubik_base_ema_040000_ctx0.log
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
infer /mnt/localssd/lehany/wv/E1_rubik_20261104031751/rubik_base/ema_040000.pt (ema, step 40000) | max_ctx 0 | moves [0, 2, 4, 8, 12, 16, 24] | ['free', 'tf'] | 50 steps | seed 0 -> /mnt/localssd/lehany/wv/E1_rubik_20261104031751/inference/rubik_base/ema_040000_ctx0
|
| 2 |
+
test/n00: 64 episodes, K 7, free, tf saved (62 s)
|
| 3 |
+
test/n02: 64 episodes, K 9, free, tf saved (144 s)
|
| 4 |
+
test/n04: 64 episodes, K 11, free, tf saved (246 s)
|
| 5 |
+
test/n08: 64 episodes, K 15, free, tf saved (388 s)
|
| 6 |
+
test/n12: 64 episodes, K 19, free, tf saved (571 s)
|
| 7 |
+
test/n16: 64 episodes, K 23, free, tf saved (795 s)
|
| 8 |
+
test/n24: 64 episodes, K 31, free, tf saved (1100 s)
|
| 9 |
+
done -> /mnt/localssd/lehany/wv/E1_rubik_20261104031751/inference/rubik_base/ema_040000_ctx0
|
rubik_r1_20261106011120/logs/rk_q.log
ADDED
|
@@ -0,0 +1,271 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
run rk_q | MemConfig(proj='q', ffn=False, layers=(0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11), fast_heads=4, blocks=4, ffn_frac=0.25, ffn_blocks=8, loss='delta', key='raw', lr=0.001, renorm=True) | max_ctx 0 | new params 4.00M | fast weights / sample: writable 1.00M, state 3.98M (15 MB fp32) | start step 0
|
| 2 |
+
wandb: Currently logged in as: xsourse to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
|
| 3 |
+
wandb: creating run
|
| 4 |
+
wandb: Tracking run with wandb version 0.22.0
|
| 5 |
+
wandb: Run data is saved locally in /mnt/localssd/lehany/wv/E1_rubik_20261104031751/rk_q/wandb/run-20261104_071615-ywo1hfj6
|
| 6 |
+
wandb: Run `wandb offline` to turn off syncing.
|
| 7 |
+
wandb: Syncing run rk_q
|
| 8 |
+
wandb: ⭐️ View project at https://wandb.ai/xsourse/world-auto-visual
|
| 9 |
+
wandb: 🚀 View run at https://wandb.ai/xsourse/world-auto-visual/runs/ywo1hfj6
|
| 10 |
+
step 50 | loss 0.0619 | obs 0.0596 ret 0.0601 | gn 0.12 | dW_q 1.38e-04 | 2.72 s/step | 17 GB
|
| 11 |
+
step 100 | loss 0.0565 | obs 0.0553 ret 0.0591 | gn 0.53 | dW_q 1.39e-03 | 2.40 s/step | 17 GB
|
| 12 |
+
step 150 | loss 0.0543 | obs 0.0548 ret 0.0537 | gn 0.04 | dW_q 3.15e-03 | 2.48 s/step | 17 GB
|
| 13 |
+
step 200 | loss 0.0552 | obs 0.0556 ret 0.0537 | gn 0.04 | dW_q 6.17e-03 | 2.52 s/step | 17 GB
|
| 14 |
+
step 250 | loss 0.0535 | obs 0.0521 ret 0.0531 | gn 0.04 | dW_q 8.70e-03 | 2.39 s/step | 17 GB
|
| 15 |
+
step 300 | loss 0.0527 | obs 0.0531 ret 0.0534 | gn 0.04 | dW_q 1.14e-02 | 2.57 s/step | 17 GB
|
| 16 |
+
step 350 | loss 0.0540 | obs 0.0522 ret 0.0562 | gn 0.04 | dW_q 1.58e-02 | 2.59 s/step | 17 GB
|
| 17 |
+
step 400 | loss 0.0551 | obs 0.0533 ret 0.0534 | gn 0.04 | dW_q 1.50e-02 | 2.54 s/step | 17 GB
|
| 18 |
+
step 450 | loss 0.0544 | obs 0.0548 ret 0.0515 | gn 0.05 | dW_q 1.70e-02 | 2.49 s/step | 17 GB
|
| 19 |
+
step 500 | loss 0.0536 | obs 0.0521 ret 0.0571 | gn 0.04 | dW_q 2.42e-02 | 2.45 s/step | 17 GB
|
| 20 |
+
step 550 | loss 0.0532 | obs 0.0521 ret 0.0512 | gn 0.04 | dW_q 1.92e-02 | 2.49 s/step | 17 GB
|
| 21 |
+
step 600 | loss 0.0549 | obs 0.0530 ret 0.0535 | gn 0.05 | dW_q 2.78e-02 | 2.53 s/step | 17 GB
|
| 22 |
+
step 650 | loss 0.0532 | obs 0.0505 ret 0.0509 | gn 0.05 | dW_q 2.01e-02 | 2.38 s/step | 17 GB
|
| 23 |
+
step 700 | loss 0.0552 | obs 0.0543 ret 0.0550 | gn 0.05 | dW_q 2.31e-02 | 2.51 s/step | 17 GB
|
| 24 |
+
step 750 | loss 0.0538 | obs 0.0547 ret 0.0553 | gn 0.04 | dW_q 3.11e-02 | 2.64 s/step | 17 GB
|
| 25 |
+
step 800 | loss 0.0534 | obs 0.0515 ret 0.0544 | gn 0.05 | dW_q 3.76e-02 | 2.62 s/step | 17 GB
|
| 26 |
+
step 850 | loss 0.0537 | obs 0.0532 ret 0.0548 | gn 0.04 | dW_q 3.12e-02 | 2.57 s/step | 17 GB
|
| 27 |
+
step 900 | loss 0.0540 | obs 0.0516 ret 0.0583 | gn 0.05 | dW_q 4.00e-02 | 2.49 s/step | 17 GB
|
| 28 |
+
step 950 | loss 0.0546 | obs 0.0510 ret 0.0551 | gn 0.04 | dW_q 3.56e-02 | 2.35 s/step | 17 GB
|
| 29 |
+
step 1000 | loss 0.0539 | obs 0.0537 ret 0.0535 | gn 0.05 | dW_q 3.54e-02 | 2.44 s/step | 17 GB
|
| 30 |
+
step 1050 | loss 0.0534 | obs 0.0521 ret 0.0519 | gn 0.04 | dW_q 3.41e-02 | 2.53 s/step | 17 GB
|
| 31 |
+
step 1100 | loss 0.0535 | obs 0.0526 ret 0.0556 | gn 0.04 | dW_q 5.00e-02 | 2.59 s/step | 17 GB
|
| 32 |
+
step 1150 | loss 0.0524 | obs 0.0504 ret 0.0517 | gn 0.04 | dW_q 4.05e-02 | 2.44 s/step | 17 GB
|
| 33 |
+
step 1200 | loss 0.0535 | obs 0.0527 ret 0.0492 | gn 0.04 | dW_q 5.32e-02 | 2.55 s/step | 17 GB
|
| 34 |
+
step 1250 | loss 0.0531 | obs 0.0525 ret 0.0510 | gn 0.05 | dW_q 5.27e-02 | 2.40 s/step | 17 GB
|
| 35 |
+
step 1300 | loss 0.0529 | obs 0.0526 ret 0.0492 | gn 0.05 | dW_q 4.65e-02 | 2.51 s/step | 17 GB
|
| 36 |
+
step 1350 | loss 0.0544 | obs 0.0526 ret 0.0538 | gn 0.05 | dW_q 5.70e-02 | 2.42 s/step | 17 GB
|
| 37 |
+
step 1400 | loss 0.0517 | obs 0.0509 ret 0.0509 | gn 0.05 | dW_q 6.45e-02 | 2.58 s/step | 17 GB
|
| 38 |
+
step 1450 | loss 0.0502 | obs 0.0500 ret 0.0523 | gn 0.05 | dW_q 5.77e-02 | 2.47 s/step | 17 GB
|
| 39 |
+
step 1500 | loss 0.0498 | obs 0.0478 ret 0.0514 | gn 0.06 | dW_q 7.50e-02 | 2.54 s/step | 17 GB
|
| 40 |
+
step 1550 | loss 0.0504 | obs 0.0497 ret 0.0528 | gn 0.07 | dW_q 6.90e-02 | 2.45 s/step | 17 GB
|
| 41 |
+
step 1600 | loss 0.0472 | obs 0.0467 ret 0.0493 | gn 0.06 | dW_q 8.35e-02 | 2.41 s/step | 17 GB
|
| 42 |
+
step 1650 | loss 0.0441 | obs 0.0424 ret 0.0423 | gn 0.06 | dW_q 8.06e-02 | 2.46 s/step | 17 GB
|
| 43 |
+
step 1700 | loss 0.0444 | obs 0.0409 ret 0.0506 | gn 0.06 | dW_q 7.17e-02 | 2.50 s/step | 17 GB
|
| 44 |
+
step 1750 | loss 0.0415 | obs 0.0379 ret 0.0459 | gn 0.07 | dW_q 8.85e-02 | 2.57 s/step | 17 GB
|
| 45 |
+
step 1800 | loss 0.0377 | obs 0.0379 ret 0.0427 | gn 0.08 | dW_q 6.64e-02 | 2.46 s/step | 17 GB
|
| 46 |
+
step 1850 | loss 0.0347 | obs 0.0312 ret 0.0414 | gn 0.07 | dW_q 7.67e-02 | 2.49 s/step | 17 GB
|
| 47 |
+
step 1900 | loss 0.0345 | obs 0.0323 ret 0.0377 | gn 0.07 | dW_q 7.73e-02 | 2.59 s/step | 17 GB
|
| 48 |
+
step 1950 | loss 0.0310 | obs 0.0305 ret 0.0329 | gn 0.07 | dW_q 8.80e-02 | 2.52 s/step | 17 GB
|
| 49 |
+
step 2000 | loss 0.0292 | obs 0.0272 ret 0.0348 | gn 0.06 | dW_q 7.62e-02 | 2.33 s/step | 17 GB
|
| 50 |
+
step 2050 | loss 0.0271 | obs 0.0245 ret 0.0289 | gn 0.06 | dW_q 9.33e-02 | 2.66 s/step | 17 GB
|
| 51 |
+
run rk_q | MemConfig(proj='q', ffn=False, layers=(0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11), fast_heads=4, blocks=4, ffn_frac=0.25, ffn_blocks=8, loss='delta', key='raw', lr=0.001, renorm=True) | max_ctx 0 | new params 4.00M | fast weights / sample: writable 1.00M, state 3.98M (15 MB fp32) | start step 2000
|
| 52 |
+
wandb: Currently logged in as: xsourse to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
|
| 53 |
+
wandb: creating run
|
| 54 |
+
wandb: Tracking run with wandb version 0.22.0
|
| 55 |
+
wandb: Run data is saved locally in /mnt/localssd/lehany/wv/E1_rubik_20261104031751/rk_q/wandb/run-20261105_121204-ywo1hfj6
|
| 56 |
+
wandb: Run `wandb offline` to turn off syncing.
|
| 57 |
+
wandb: Resuming run rk_q
|
| 58 |
+
wandb: ⭐️ View project at https://wandb.ai/xsourse/world-auto-visual
|
| 59 |
+
wandb: 🚀 View run at https://wandb.ai/xsourse/world-auto-visual/runs/ywo1hfj6
|
| 60 |
+
step 2050 | loss 0.0261 | obs 0.0248 ret 0.0262 | gn 0.07 | dW_q 9.27e-02 | 2.57 s/step | 19 GB
|
| 61 |
+
step 2100 | loss 0.0249 | obs 0.0254 ret 0.0253 | gn 0.06 | dW_q 8.05e-02 | 2.45 s/step | 19 GB
|
| 62 |
+
step 2150 | loss 0.0231 | obs 0.0230 ret 0.0282 | gn 0.06 | dW_q 7.15e-02 | 2.44 s/step | 19 GB
|
| 63 |
+
step 2200 | loss 0.0222 | obs 0.0218 ret 0.0213 | gn 0.05 | dW_q 9.49e-02 | 2.63 s/step | 19 GB
|
| 64 |
+
step 2250 | loss 0.0199 | obs 0.0199 ret 0.0184 | gn 0.05 | dW_q 8.06e-02 | 2.55 s/step | 19 GB
|
| 65 |
+
step 2300 | loss 0.0191 | obs 0.0215 ret 0.0203 | gn 0.05 | dW_q 8.72e-02 | 2.48 s/step | 19 GB
|
| 66 |
+
step 2350 | loss 0.0193 | obs 0.0190 ret 0.0210 | gn 0.04 | dW_q 9.92e-02 | 2.51 s/step | 19 GB
|
| 67 |
+
step 2400 | loss 0.0207 | obs 0.0191 ret 0.0287 | gn 0.05 | dW_q 8.50e-02 | 2.56 s/step | 19 GB
|
| 68 |
+
step 2450 | loss 0.0186 | obs 0.0195 ret 0.0198 | gn 0.04 | dW_q 7.54e-02 | 2.41 s/step | 19 GB
|
| 69 |
+
step 2500 | loss 0.0179 | obs 0.0178 ret 0.0225 | gn 0.04 | dW_q 8.65e-02 | 2.43 s/step | 19 GB
|
| 70 |
+
step 2550 | loss 0.0169 | obs 0.0169 ret 0.0172 | gn 0.04 | dW_q 9.67e-02 | 2.58 s/step | 19 GB
|
| 71 |
+
step 2600 | loss 0.0197 | obs 0.0193 ret 0.0229 | gn 0.05 | dW_q 9.53e-02 | 2.47 s/step | 19 GB
|
| 72 |
+
step 2650 | loss 0.0180 | obs 0.0189 ret 0.0181 | gn 0.04 | dW_q 8.59e-02 | 2.50 s/step | 19 GB
|
| 73 |
+
step 2700 | loss 0.0183 | obs 0.0188 ret 0.0201 | gn 0.04 | dW_q 8.21e-02 | 2.33 s/step | 19 GB
|
| 74 |
+
step 2750 | loss 0.0189 | obs 0.0198 ret 0.0192 | gn 0.04 | dW_q 1.11e-01 | 2.55 s/step | 19 GB
|
| 75 |
+
step 2800 | loss 0.0173 | obs 0.0186 ret 0.0177 | gn 0.04 | dW_q 1.09e-01 | 2.62 s/step | 19 GB
|
| 76 |
+
step 2850 | loss 0.0173 | obs 0.0176 ret 0.0198 | gn 0.04 | dW_q 9.26e-02 | 2.73 s/step | 19 GB
|
| 77 |
+
step 2900 | loss 0.0177 | obs 0.0169 ret 0.0195 | gn 0.04 | dW_q 9.16e-02 | 2.35 s/step | 19 GB
|
| 78 |
+
step 2950 | loss 0.0175 | obs 0.0185 ret 0.0193 | gn 0.04 | dW_q 8.12e-02 | 2.32 s/step | 19 GB
|
| 79 |
+
step 3000 | loss 0.0172 | obs 0.0176 ret 0.0162 | gn 0.04 | dW_q 9.37e-02 | 2.39 s/step | 19 GB
|
| 80 |
+
step 3050 | loss 0.0174 | obs 0.0164 ret 0.0203 | gn 0.04 | dW_q 8.84e-02 | 2.53 s/step | 19 GB
|
| 81 |
+
step 3100 | loss 0.0170 | obs 0.0198 ret 0.0197 | gn 0.05 | dW_q 8.67e-02 | 2.43 s/step | 19 GB
|
| 82 |
+
step 3150 | loss 0.0164 | obs 0.0154 ret 0.0184 | gn 0.04 | dW_q 1.11e-01 | 2.34 s/step | 19 GB
|
| 83 |
+
step 3200 | loss 0.0162 | obs 0.0170 ret 0.0160 | gn 0.04 | dW_q 8.56e-02 | 2.26 s/step | 19 GB
|
| 84 |
+
step 3250 | loss 0.0163 | obs 0.0196 ret 0.0172 | gn 0.04 | dW_q 1.05e-01 | 2.56 s/step | 19 GB
|
| 85 |
+
step 3300 | loss 0.0178 | obs 0.0201 ret 0.0197 | gn 0.04 | dW_q 8.99e-02 | 2.44 s/step | 19 GB
|
| 86 |
+
step 3350 | loss 0.0164 | obs 0.0188 ret 0.0173 | gn 0.04 | dW_q 8.89e-02 | 2.46 s/step | 19 GB
|
| 87 |
+
step 3400 | loss 0.0170 | obs 0.0179 ret 0.0176 | gn 0.04 | dW_q 8.70e-02 | 2.39 s/step | 19 GB
|
| 88 |
+
step 3450 | loss 0.0166 | obs 0.0171 ret 0.0228 | gn 0.04 | dW_q 9.30e-02 | 2.45 s/step | 19 GB
|
| 89 |
+
step 3500 | loss 0.0152 | obs 0.0155 ret 0.0150 | gn 0.04 | dW_q 1.11e-01 | 2.48 s/step | 19 GB
|
| 90 |
+
step 3550 | loss 0.0175 | obs 0.0184 ret 0.0179 | gn 0.04 | dW_q 1.06e-01 | 2.76 s/step | 19 GB
|
| 91 |
+
step 3600 | loss 0.0153 | obs 0.0150 ret 0.0181 | gn 0.04 | dW_q 1.10e-01 | 2.51 s/step | 19 GB
|
| 92 |
+
step 3650 | loss 0.0168 | obs 0.0179 ret 0.0194 | gn 0.04 | dW_q 9.27e-02 | 2.53 s/step | 19 GB
|
| 93 |
+
step 3700 | loss 0.0154 | obs 0.0154 ret 0.0183 | gn 0.04 | dW_q 1.08e-01 | 2.63 s/step | 19 GB
|
| 94 |
+
step 3750 | loss 0.0159 | obs 0.0163 ret 0.0140 | gn 0.04 | dW_q 1.12e-01 | 2.38 s/step | 19 GB
|
| 95 |
+
step 3800 | loss 0.0180 | obs 0.0177 ret 0.0195 | gn 0.05 | dW_q 9.98e-02 | 2.18 s/step | 19 GB
|
| 96 |
+
step 3850 | loss 0.0157 | obs 0.0156 ret 0.0170 | gn 0.04 | dW_q 1.15e-01 | 2.51 s/step | 19 GB
|
| 97 |
+
step 3900 | loss 0.0158 | obs 0.0175 ret 0.0148 | gn 0.04 | dW_q 1.06e-01 | 2.38 s/step | 19 GB
|
| 98 |
+
step 3950 | loss 0.0180 | obs 0.0189 ret 0.0250 | gn 0.04 | dW_q 1.03e-01 | 2.45 s/step | 19 GB
|
| 99 |
+
step 4000 | loss 0.0165 | obs 0.0162 ret 0.0153 | gn 0.04 | dW_q 1.07e-01 | 2.49 s/step | 19 GB
|
| 100 |
+
step 4050 | loss 0.0168 | obs 0.0176 ret 0.0154 | gn 0.04 | dW_q 1.15e-01 | 2.67 s/step | 19 GB
|
| 101 |
+
step 4100 | loss 0.0158 | obs 0.0164 ret 0.0198 | gn 0.05 | dW_q 9.99e-02 | 2.59 s/step | 19 GB
|
| 102 |
+
step 4150 | loss 0.0168 | obs 0.0185 ret 0.0162 | gn 0.04 | dW_q 1.01e-01 | 2.52 s/step | 19 GB
|
| 103 |
+
step 4200 | loss 0.0168 | obs 0.0182 ret 0.0174 | gn 0.05 | dW_q 9.58e-02 | 2.62 s/step | 19 GB
|
| 104 |
+
step 4250 | loss 0.0158 | obs 0.0168 ret 0.0203 | gn 0.05 | dW_q 1.18e-01 | 2.52 s/step | 19 GB
|
| 105 |
+
step 4300 | loss 0.0165 | obs 0.0180 ret 0.0182 | gn 0.04 | dW_q 1.19e-01 | 2.39 s/step | 19 GB
|
| 106 |
+
step 4350 | loss 0.0165 | obs 0.0171 ret 0.0159 | gn 0.04 | dW_q 1.05e-01 | 2.75 s/step | 19 GB
|
| 107 |
+
step 4400 | loss 0.0152 | obs 0.0153 ret 0.0160 | gn 0.04 | dW_q 1.19e-01 | 2.45 s/step | 19 GB
|
| 108 |
+
step 4450 | loss 0.0156 | obs 0.0173 ret 0.0149 | gn 0.04 | dW_q 1.09e-01 | 2.45 s/step | 19 GB
|
| 109 |
+
step 4500 | loss 0.0157 | obs 0.0164 ret 0.0189 | gn 0.04 | dW_q 1.01e-01 | 2.50 s/step | 19 GB
|
| 110 |
+
step 4550 | loss 0.0149 | obs 0.0158 ret 0.0165 | gn 0.05 | dW_q 1.11e-01 | 2.37 s/step | 19 GB
|
| 111 |
+
step 4600 | loss 0.0163 | obs 0.0186 ret 0.0187 | gn 0.04 | dW_q 1.05e-01 | 2.36 s/step | 19 GB
|
| 112 |
+
step 4650 | loss 0.0160 | obs 0.0168 ret 0.0169 | gn 0.04 | dW_q 1.02e-01 | 2.70 s/step | 19 GB
|
| 113 |
+
step 4700 | loss 0.0164 | obs 0.0169 ret 0.0191 | gn 0.04 | dW_q 1.19e-01 | 2.63 s/step | 19 GB
|
| 114 |
+
step 4750 | loss 0.0177 | obs 0.0162 ret 0.0222 | gn 0.04 | dW_q 1.08e-01 | 2.62 s/step | 19 GB
|
| 115 |
+
step 4800 | loss 0.0160 | obs 0.0161 ret 0.0158 | gn 0.04 | dW_q 1.36e-01 | 2.43 s/step | 19 GB
|
| 116 |
+
step 4850 | loss 0.0165 | obs 0.0183 ret 0.0190 | gn 0.04 | dW_q 1.18e-01 | 2.48 s/step | 19 GB
|
| 117 |
+
step 4900 | loss 0.0156 | obs 0.0169 ret 0.0184 | gn 0.04 | dW_q 1.20e-01 | 2.47 s/step | 19 GB
|
| 118 |
+
step 4950 | loss 0.0152 | obs 0.0160 ret 0.0150 | gn 0.04 | dW_q 1.31e-01 | 2.73 s/step | 19 GB
|
| 119 |
+
step 5000 | loss 0.0152 | obs 0.0163 ret 0.0153 | gn 0.04 | dW_q 1.36e-01 | 2.43 s/step | 19 GB
|
| 120 |
+
step 5050 | loss 0.0150 | obs 0.0161 ret 0.0184 | gn 0.04 | dW_q 1.17e-01 | 2.50 s/step | 19 GB
|
| 121 |
+
step 5100 | loss 0.0153 | obs 0.0172 ret 0.0151 | gn 0.04 | dW_q 1.11e-01 | 2.47 s/step | 19 GB
|
| 122 |
+
step 5150 | loss 0.0150 | obs 0.0142 ret 0.0144 | gn 0.04 | dW_q 1.22e-01 | 2.31 s/step | 19 GB
|
| 123 |
+
step 5200 | loss 0.0156 | obs 0.0160 ret 0.0160 | gn 0.04 | dW_q 1.42e-01 | 2.50 s/step | 19 GB
|
| 124 |
+
step 5250 | loss 0.0151 | obs 0.0158 ret 0.0185 | gn 0.04 | dW_q 1.37e-01 | 2.46 s/step | 19 GB
|
| 125 |
+
step 5300 | loss 0.0165 | obs 0.0157 ret 0.0228 | gn 0.04 | dW_q 1.13e-01 | 2.44 s/step | 19 GB
|
| 126 |
+
step 5350 | loss 0.0155 | obs 0.0164 ret 0.0185 | gn 0.04 | dW_q 1.27e-01 | 2.51 s/step | 19 GB
|
| 127 |
+
step 5400 | loss 0.0156 | obs 0.0170 ret 0.0151 | gn 0.04 | dW_q 1.34e-01 | 2.50 s/step | 19 GB
|
| 128 |
+
step 5450 | loss 0.0154 | obs 0.0184 ret 0.0146 | gn 0.04 | dW_q 1.21e-01 | 2.41 s/step | 19 GB
|
| 129 |
+
step 5500 | loss 0.0162 | obs 0.0164 ret 0.0162 | gn 0.04 | dW_q 1.19e-01 | 2.42 s/step | 19 GB
|
| 130 |
+
step 5550 | loss 0.0156 | obs 0.0164 ret 0.0139 | gn 0.04 | dW_q 1.34e-01 | 2.46 s/step | 19 GB
|
| 131 |
+
step 5600 | loss 0.0158 | obs 0.0168 ret 0.0191 | gn 0.04 | dW_q 1.14e-01 | 2.50 s/step | 19 GB
|
| 132 |
+
step 5650 | loss 0.0163 | obs 0.0160 ret 0.0169 | gn 0.05 | dW_q 1.38e-01 | 2.54 s/step | 19 GB
|
| 133 |
+
step 5700 | loss 0.0161 | obs 0.0148 ret 0.0195 | gn 0.04 | dW_q 1.24e-01 | 2.43 s/step | 19 GB
|
| 134 |
+
step 5750 | loss 0.0151 | obs 0.0160 ret 0.0169 | gn 0.04 | dW_q 1.17e-01 | 2.55 s/step | 19 GB
|
| 135 |
+
step 5800 | loss 0.0163 | obs 0.0176 ret 0.0159 | gn 0.05 | dW_q 1.42e-01 | 2.70 s/step | 19 GB
|
| 136 |
+
step 5850 | loss 0.0155 | obs 0.0154 ret 0.0166 | gn 0.04 | dW_q 1.45e-01 | 2.63 s/step | 19 GB
|
| 137 |
+
step 5900 | loss 0.0161 | obs 0.0179 ret 0.0147 | gn 0.04 | dW_q 1.34e-01 | 2.66 s/step | 19 GB
|
| 138 |
+
step 5950 | loss 0.0165 | obs 0.0174 ret 0.0184 | gn 0.04 | dW_q 1.20e-01 | 2.71 s/step | 19 GB
|
| 139 |
+
step 6000 | loss 0.0156 | obs 0.0163 ret 0.0180 | gn 0.04 | dW_q 1.38e-01 | 2.86 s/step | 19 GB
|
| 140 |
+
step 6050 | loss 0.0161 | obs 0.0154 ret 0.0208 | gn 0.04 | dW_q 1.33e-01 | 2.79 s/step | 19 GB
|
| 141 |
+
step 6100 | loss 0.0156 | obs 0.0155 ret 0.0158 | gn 0.04 | dW_q 1.49e-01 | 2.69 s/step | 19 GB
|
| 142 |
+
step 6150 | loss 0.0160 | obs 0.0152 ret 0.0239 | gn 0.04 | dW_q 1.40e-01 | 2.71 s/step | 19 GB
|
| 143 |
+
step 6200 | loss 0.0157 | obs 0.0156 ret 0.0163 | gn 0.04 | dW_q 1.51e-01 | 2.55 s/step | 19 GB
|
| 144 |
+
step 6250 | loss 0.0164 | obs 0.0180 ret 0.0182 | gn 0.04 | dW_q 1.33e-01 | 2.71 s/step | 19 GB
|
| 145 |
+
step 6300 | loss 0.0161 | obs 0.0172 ret 0.0174 | gn 0.04 | dW_q 1.46e-01 | 2.77 s/step | 19 GB
|
| 146 |
+
step 6350 | loss 0.0147 | obs 0.0175 ret 0.0194 | gn 0.03 | dW_q 1.50e-01 | 2.53 s/step | 19 GB
|
| 147 |
+
step 6400 | loss 0.0148 | obs 0.0148 ret 0.0172 | gn 0.04 | dW_q 1.46e-01 | 2.52 s/step | 19 GB
|
| 148 |
+
step 6450 | loss 0.0149 | obs 0.0143 ret 0.0175 | gn 0.04 | dW_q 1.44e-01 | 2.57 s/step | 19 GB
|
| 149 |
+
step 6500 | loss 0.0155 | obs 0.0152 ret 0.0141 | gn 0.04 | dW_q 1.50e-01 | 2.67 s/step | 19 GB
|
| 150 |
+
step 6550 | loss 0.0149 | obs 0.0152 ret 0.0221 | gn 0.04 | dW_q 1.40e-01 | 2.53 s/step | 19 GB
|
| 151 |
+
step 6600 | loss 0.0165 | obs 0.0177 ret 0.0182 | gn 0.05 | dW_q 1.46e-01 | 2.87 s/step | 19 GB
|
| 152 |
+
step 6650 | loss 0.0163 | obs 0.0153 ret 0.0209 | gn 0.05 | dW_q 1.33e-01 | 2.85 s/step | 19 GB
|
| 153 |
+
step 6700 | loss 0.0159 | obs 0.0169 ret 0.0195 | gn 0.04 | dW_q 1.46e-01 | 2.78 s/step | 19 GB
|
| 154 |
+
step 6750 | loss 0.0153 | obs 0.0158 ret 0.0188 | gn 0.04 | dW_q 1.53e-01 | 2.49 s/step | 19 GB
|
| 155 |
+
step 6800 | loss 0.0144 | obs 0.0165 ret 0.0137 | gn 0.04 | dW_q 1.42e-01 | 2.43 s/step | 19 GB
|
| 156 |
+
step 6850 | loss 0.0163 | obs 0.0166 ret 0.0188 | gn 0.04 | dW_q 1.35e-01 | 2.39 s/step | 19 GB
|
| 157 |
+
step 6900 | loss 0.0148 | obs 0.0158 ret 0.0159 | gn 0.04 | dW_q 1.54e-01 | 2.68 s/step | 19 GB
|
| 158 |
+
step 6950 | loss 0.0159 | obs 0.0162 ret 0.0166 | gn 0.04 | dW_q 1.42e-01 | 2.51 s/step | 19 GB
|
| 159 |
+
step 7000 | loss 0.0156 | obs 0.0180 ret 0.0138 | gn 0.04 | dW_q 1.52e-01 | 2.35 s/step | 19 GB
|
| 160 |
+
step 7050 | loss 0.0151 | obs 0.0166 ret 0.0155 | gn 0.04 | dW_q 1.48e-01 | 2.58 s/step | 19 GB
|
| 161 |
+
step 7100 | loss 0.0143 | obs 0.0151 ret 0.0142 | gn 0.04 | dW_q 1.39e-01 | 2.49 s/step | 19 GB
|
| 162 |
+
step 7150 | loss 0.0158 | obs 0.0160 ret 0.0179 | gn 0.04 | dW_q 1.42e-01 | 2.65 s/step | 19 GB
|
| 163 |
+
step 7200 | loss 0.0157 | obs 0.0163 ret 0.0233 | gn 0.04 | dW_q 1.65e-01 | 2.52 s/step | 19 GB
|
| 164 |
+
step 7250 | loss 0.0167 | obs 0.0156 ret 0.0214 | gn 0.04 | dW_q 1.46e-01 | 2.51 s/step | 19 GB
|
| 165 |
+
step 7300 | loss 0.0157 | obs 0.0168 ret 0.0150 | gn 0.04 | dW_q 1.45e-01 | 2.50 s/step | 19 GB
|
| 166 |
+
step 7350 | loss 0.0148 | obs 0.0151 ret 0.0163 | gn 0.04 | dW_q 1.40e-01 | 2.51 s/step | 19 GB
|
| 167 |
+
step 7400 | loss 0.0145 | obs 0.0166 ret 0.0151 | gn 0.04 | dW_q 1.59e-01 | 2.48 s/step | 19 GB
|
| 168 |
+
step 7450 | loss 0.0145 | obs 0.0151 ret 0.0187 | gn 0.04 | dW_q 1.45e-01 | 2.35 s/step | 19 GB
|
| 169 |
+
step 7500 | loss 0.0155 | obs 0.0154 ret 0.0197 | gn 0.04 | dW_q 1.48e-01 | 2.55 s/step | 19 GB
|
| 170 |
+
step 7550 | loss 0.0145 | obs 0.0163 ret 0.0197 | gn 0.04 | dW_q 1.70e-01 | 2.64 s/step | 19 GB
|
| 171 |
+
step 7600 | loss 0.0152 | obs 0.0175 ret 0.0177 | gn 0.04 | dW_q 1.46e-01 | 2.37 s/step | 19 GB
|
| 172 |
+
step 7650 | loss 0.0148 | obs 0.0138 ret 0.0152 | gn 0.04 | dW_q 1.65e-01 | 2.50 s/step | 19 GB
|
| 173 |
+
step 7700 | loss 0.0168 | obs 0.0157 ret 0.0239 | gn 0.05 | dW_q 1.76e-01 | 2.32 s/step | 19 GB
|
| 174 |
+
step 7750 | loss 0.0159 | obs 0.0174 ret 0.0141 | gn 0.04 | dW_q 1.56e-01 | 2.29 s/step | 19 GB
|
| 175 |
+
step 7800 | loss 0.0152 | obs 0.0150 ret 0.0142 | gn 0.05 | dW_q 1.77e-01 | 2.45 s/step | 19 GB
|
| 176 |
+
step 7850 | loss 0.0145 | obs 0.0153 ret 0.0170 | gn 0.04 | dW_q 1.62e-01 | 2.62 s/step | 19 GB
|
| 177 |
+
step 7900 | loss 0.0138 | obs 0.0166 ret 0.0117 | gn 0.04 | dW_q 1.63e-01 | 2.46 s/step | 19 GB
|
| 178 |
+
step 7950 | loss 0.0142 | obs 0.0160 ret 0.0134 | gn 0.04 | dW_q 1.80e-01 | 2.41 s/step | 19 GB
|
| 179 |
+
step 8000 | loss 0.0148 | obs 0.0153 ret 0.0181 | gn 0.04 | dW_q 1.67e-01 | 2.32 s/step | 19 GB
|
| 180 |
+
step 8050 | loss 0.0145 | obs 0.0140 ret 0.0166 | gn 0.04 | dW_q 1.85e-01 | 2.58 s/step | 19 GB
|
| 181 |
+
step 8100 | loss 0.0163 | obs 0.0173 ret 0.0183 | gn 0.05 | dW_q 1.80e-01 | 2.39 s/step | 19 GB
|
| 182 |
+
step 8150 | loss 0.0154 | obs 0.0162 ret 0.0161 | gn 0.04 | dW_q 1.58e-01 | 2.60 s/step | 19 GB
|
| 183 |
+
step 8200 | loss 0.0141 | obs 0.0145 ret 0.0155 | gn 0.04 | dW_q 1.53e-01 | 2.42 s/step | 19 GB
|
| 184 |
+
step 8250 | loss 0.0138 | obs 0.0148 ret 0.0173 | gn 0.04 | dW_q 1.78e-01 | 2.47 s/step | 19 GB
|
| 185 |
+
step 8300 | loss 0.0143 | obs 0.0177 ret 0.0144 | gn 0.04 | dW_q 1.83e-01 | 2.49 s/step | 19 GB
|
| 186 |
+
step 8350 | loss 0.0148 | obs 0.0150 ret 0.0146 | gn 0.04 | dW_q 1.85e-01 | 2.40 s/step | 19 GB
|
| 187 |
+
step 8400 | loss 0.0153 | obs 0.0159 ret 0.0198 | gn 0.04 | dW_q 1.78e-01 | 2.55 s/step | 19 GB
|
| 188 |
+
step 8450 | loss 0.0158 | obs 0.0156 ret 0.0189 | gn 0.04 | dW_q 1.62e-01 | 2.36 s/step | 19 GB
|
| 189 |
+
step 8500 | loss 0.0157 | obs 0.0171 ret 0.0176 | gn 0.04 | dW_q 1.78e-01 | 2.41 s/step | 19 GB
|
| 190 |
+
step 8550 | loss 0.0147 | obs 0.0162 ret 0.0143 | gn 0.04 | dW_q 1.63e-01 | 2.49 s/step | 19 GB
|
| 191 |
+
step 8600 | loss 0.0144 | obs 0.0161 ret 0.0181 | gn 0.04 | dW_q 1.82e-01 | 2.29 s/step | 19 GB
|
| 192 |
+
step 8650 | loss 0.0145 | obs 0.0158 ret 0.0147 | gn 0.04 | dW_q 1.51e-01 | 2.28 s/step | 19 GB
|
| 193 |
+
step 8700 | loss 0.0138 | obs 0.0156 ret 0.0174 | gn 0.04 | dW_q 1.81e-01 | 2.51 s/step | 19 GB
|
| 194 |
+
step 8750 | loss 0.0150 | obs 0.0157 ret 0.0150 | gn 0.04 | dW_q 1.91e-01 | 2.51 s/step | 19 GB
|
| 195 |
+
step 8800 | loss 0.0138 | obs 0.0158 ret 0.0128 | gn 0.04 | dW_q 1.75e-01 | 2.41 s/step | 19 GB
|
| 196 |
+
step 8850 | loss 0.0140 | obs 0.0139 ret 0.0211 | gn 0.04 | dW_q 1.84e-01 | 2.44 s/step | 19 GB
|
| 197 |
+
step 8900 | loss 0.0144 | obs 0.0155 ret 0.0158 | gn 0.04 | dW_q 1.80e-01 | 2.40 s/step | 19 GB
|
| 198 |
+
step 8950 | loss 0.0136 | obs 0.0158 ret 0.0167 | gn 0.04 | dW_q 1.73e-01 | 2.49 s/step | 19 GB
|
| 199 |
+
step 9000 | loss 0.0148 | obs 0.0180 ret 0.0179 | gn 0.04 | dW_q 1.56e-01 | 2.47 s/step | 19 GB
|
| 200 |
+
step 9050 | loss 0.0152 | obs 0.0159 ret 0.0199 | gn 0.05 | dW_q 1.86e-01 | 2.45 s/step | 19 GB
|
| 201 |
+
step 9100 | loss 0.0160 | obs 0.0180 ret 0.0165 | gn 0.04 | dW_q 1.77e-01 | 2.36 s/step | 19 GB
|
| 202 |
+
step 9150 | loss 0.0140 | obs 0.0144 ret 0.0178 | gn 0.04 | dW_q 1.92e-01 | 2.42 s/step | 19 GB
|
| 203 |
+
step 9200 | loss 0.0138 | obs 0.0132 ret 0.0135 | gn 0.04 | dW_q 1.87e-01 | 2.57 s/step | 19 GB
|
| 204 |
+
step 9250 | loss 0.0151 | obs 0.0171 ret 0.0162 | gn 0.04 | dW_q 1.59e-01 | 2.39 s/step | 19 GB
|
| 205 |
+
step 9300 | loss 0.0151 | obs 0.0167 ret 0.0170 | gn 0.04 | dW_q 1.72e-01 | 2.34 s/step | 19 GB
|
| 206 |
+
step 9350 | loss 0.0162 | obs 0.0178 ret 0.0182 | gn 0.05 | dW_q 1.59e-01 | 2.60 s/step | 19 GB
|
| 207 |
+
step 9400 | loss 0.0158 | obs 0.0178 ret 0.0193 | gn 0.04 | dW_q 1.86e-01 | 2.34 s/step | 19 GB
|
| 208 |
+
step 9450 | loss 0.0145 | obs 0.0168 ret 0.0181 | gn 0.04 | dW_q 1.84e-01 | 2.40 s/step | 19 GB
|
| 209 |
+
step 9500 | loss 0.0141 | obs 0.0159 ret 0.0159 | gn 0.04 | dW_q 1.66e-01 | 2.53 s/step | 19 GB
|
| 210 |
+
step 9550 | loss 0.0145 | obs 0.0173 ret 0.0155 | gn 0.04 | dW_q 1.73e-01 | 2.61 s/step | 19 GB
|
| 211 |
+
step 9600 | loss 0.0146 | obs 0.0165 ret 0.0190 | gn 0.04 | dW_q 1.66e-01 | 2.44 s/step | 19 GB
|
| 212 |
+
step 9650 | loss 0.0145 | obs 0.0183 ret 0.0135 | gn 0.05 | dW_q 1.75e-01 | 2.30 s/step | 19 GB
|
| 213 |
+
step 9700 | loss 0.0144 | obs 0.0160 ret 0.0147 | gn 0.05 | dW_q 1.67e-01 | 2.28 s/step | 19 GB
|
| 214 |
+
step 9750 | loss 0.0134 | obs 0.0154 ret 0.0178 | gn 0.04 | dW_q 1.69e-01 | 2.50 s/step | 19 GB
|
| 215 |
+
step 9800 | loss 0.0156 | obs 0.0180 ret 0.0189 | gn 0.05 | dW_q 1.66e-01 | 2.37 s/step | 19 GB
|
| 216 |
+
step 9850 | loss 0.0134 | obs 0.0143 ret 0.0141 | gn 0.04 | dW_q 1.72e-01 | 2.31 s/step | 19 GB
|
| 217 |
+
step 9900 | loss 0.0149 | obs 0.0162 ret 0.0166 | gn 0.04 | dW_q 1.92e-01 | 2.38 s/step | 19 GB
|
| 218 |
+
step 9950 | loss 0.0131 | obs 0.0144 ret 0.0156 | gn 0.04 | dW_q 1.57e-01 | 2.50 s/step | 19 GB
|
| 219 |
+
step 10000 | loss 0.0149 | obs 0.0176 ret 0.0158 | gn 0.04 | dW_q 1.90e-01 | 2.29 s/step | 19 GB
|
| 220 |
+
run rk_q | MemConfig(proj='q', ffn=False, layers=(0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11), fast_heads=4, blocks=4, ffn_frac=0.25, ffn_blocks=8, loss='delta', key='raw', lr=0.001, renorm=True) | max_ctx 0 | new params 4.00M | fast weights / sample: writable 1.00M, state 3.98M (15 MB fp32) | start step 10000
|
| 221 |
+
wandb: Currently logged in as: xsourse to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
|
| 222 |
+
wandb: creating run
|
| 223 |
+
wandb: Tracking run with wandb version 0.22.0
|
| 224 |
+
wandb: Run data is saved locally in /mnt/localssd/lehany/wv/E1_rubik_20261104031751/rk_q/wandb/run-20261105_174950-ywo1hfj6
|
| 225 |
+
wandb: Run `wandb offline` to turn off syncing.
|
| 226 |
+
wandb: Resuming run rk_q
|
| 227 |
+
wandb: ⭐️ View project at https://wandb.ai/xsourse/world-auto-visual
|
| 228 |
+
wandb: 🚀 View run at https://wandb.ai/xsourse/world-auto-visual/runs/ywo1hfj6
|
| 229 |
+
step 10050 | loss 0.0136 | obs 0.0151 ret 0.0161 | gn 0.04 | dW_q 1.85e-01 | 2.76 s/step | 19 GB
|
| 230 |
+
step 10100 | loss 0.0144 | obs 0.0162 ret 0.0155 | gn 0.04 | dW_q 1.76e-01 | 2.64 s/step | 19 GB
|
| 231 |
+
step 10150 | loss 0.0138 | obs 0.0160 ret 0.0138 | gn 0.04 | dW_q 1.74e-01 | 2.42 s/step | 19 GB
|
| 232 |
+
step 10200 | loss 0.0145 | obs 0.0165 ret 0.0178 | gn 0.05 | dW_q 1.59e-01 | 2.44 s/step | 19 GB
|
| 233 |
+
step 10250 | loss 0.0131 | obs 0.0147 ret 0.0185 | gn 0.04 | dW_q 1.94e-01 | 2.47 s/step | 19 GB
|
| 234 |
+
step 10300 | loss 0.0125 | obs 0.0149 ret 0.0145 | gn 0.04 | dW_q 1.89e-01 | 2.38 s/step | 19 GB
|
| 235 |
+
step 10350 | loss 0.0135 | obs 0.0141 ret 0.0148 | gn 0.04 | dW_q 1.92e-01 | 2.51 s/step | 19 GB
|
| 236 |
+
step 10400 | loss 0.0153 | obs 0.0196 ret 0.0193 | gn 0.04 | dW_q 1.92e-01 | 2.61 s/step | 19 GB
|
| 237 |
+
step 10450 | loss 0.0137 | obs 0.0164 ret 0.0130 | gn 0.04 | dW_q 1.85e-01 | 2.49 s/step | 19 GB
|
| 238 |
+
step 10500 | loss 0.0140 | obs 0.0159 ret 0.0194 | gn 0.04 | dW_q 1.70e-01 | 2.55 s/step | 19 GB
|
| 239 |
+
step 10550 | loss 0.0129 | obs 0.0164 ret 0.0139 | gn 0.04 | dW_q 1.73e-01 | 2.58 s/step | 19 GB
|
| 240 |
+
step 10600 | loss 0.0153 | obs 0.0185 ret 0.0158 | gn 0.05 | dW_q 1.92e-01 | 2.38 s/step | 19 GB
|
| 241 |
+
step 10650 | loss 0.0135 | obs 0.0145 ret 0.0185 | gn 0.04 | dW_q 1.66e-01 | 2.39 s/step | 19 GB
|
| 242 |
+
step 10700 | loss 0.0145 | obs 0.0161 ret 0.0144 | gn 0.05 | dW_q 1.96e-01 | 2.37 s/step | 19 GB
|
| 243 |
+
step 10750 | loss 0.0149 | obs 0.0153 ret 0.0209 | gn 0.04 | dW_q 1.70e-01 | 2.35 s/step | 19 GB
|
| 244 |
+
step 10800 | loss 0.0138 | obs 0.0151 ret 0.0142 | gn 0.04 | dW_q 1.78e-01 | 2.41 s/step | 19 GB
|
| 245 |
+
step 10850 | loss 0.0134 | obs 0.0149 ret 0.0139 | gn 0.04 | dW_q 1.84e-01 | 2.44 s/step | 19 GB
|
| 246 |
+
step 10900 | loss 0.0144 | obs 0.0162 ret 0.0146 | gn 0.05 | dW_q 1.86e-01 | 2.41 s/step | 19 GB
|
| 247 |
+
step 10950 | loss 0.0145 | obs 0.0160 ret 0.0166 | gn 0.04 | dW_q 1.93e-01 | 2.46 s/step | 19 GB
|
| 248 |
+
step 11000 | loss 0.0143 | obs 0.0148 ret 0.0208 | gn 0.04 | dW_q 1.95e-01 | 2.49 s/step | 19 GB
|
| 249 |
+
step 11050 | loss 0.0138 | obs 0.0161 ret 0.0149 | gn 0.04 | dW_q 1.73e-01 | 2.52 s/step | 19 GB
|
| 250 |
+
step 11100 | loss 0.0134 | obs 0.0152 ret 0.0147 | gn 0.04 | dW_q 1.99e-01 | 2.45 s/step | 19 GB
|
| 251 |
+
step 11150 | loss 0.0136 | obs 0.0147 ret 0.0176 | gn 0.04 | dW_q 1.79e-01 | 2.41 s/step | 19 GB
|
| 252 |
+
step 11200 | loss 0.0127 | obs 0.0173 ret 0.0122 | gn 0.04 | dW_q 1.73e-01 | 2.45 s/step | 19 GB
|
| 253 |
+
step 11250 | loss 0.0136 | obs 0.0138 ret 0.0167 | gn 0.04 | dW_q 1.83e-01 | 2.34 s/step | 19 GB
|
| 254 |
+
step 11300 | loss 0.0142 | obs 0.0164 ret 0.0161 | gn 0.04 | dW_q 1.65e-01 | 2.44 s/step | 19 GB
|
| 255 |
+
step 11350 | loss 0.0138 | obs 0.0161 ret 0.0198 | gn 0.04 | dW_q 1.83e-01 | 2.42 s/step | 19 GB
|
| 256 |
+
step 11400 | loss 0.0141 | obs 0.0165 ret 0.0165 | gn 0.04 | dW_q 1.66e-01 | 2.48 s/step | 19 GB
|
| 257 |
+
step 11450 | loss 0.0134 | obs 0.0148 ret 0.0150 | gn 0.04 | dW_q 1.78e-01 | 2.49 s/step | 19 GB
|
| 258 |
+
step 11500 | loss 0.0124 | obs 0.0167 ret 0.0137 | gn 0.04 | dW_q 1.64e-01 | 2.44 s/step | 19 GB
|
| 259 |
+
step 11550 | loss 0.0147 | obs 0.0171 ret 0.0179 | gn 0.04 | dW_q 1.99e-01 | 2.57 s/step | 19 GB
|
| 260 |
+
step 11600 | loss 0.0134 | obs 0.0178 ret 0.0143 | gn 0.04 | dW_q 2.01e-01 | 2.61 s/step | 19 GB
|
| 261 |
+
step 11650 | loss 0.0135 | obs 0.0169 ret 0.0147 | gn 0.04 | dW_q 1.87e-01 | 2.42 s/step | 19 GB
|
| 262 |
+
step 11700 | loss 0.0135 | obs 0.0150 ret 0.0218 | gn 0.04 | dW_q 1.73e-01 | 2.51 s/step | 19 GB
|
| 263 |
+
step 11750 | loss 0.0142 | obs 0.0186 ret 0.0144 | gn 0.04 | dW_q 2.00e-01 | 2.58 s/step | 19 GB
|
| 264 |
+
step 11800 | loss 0.0140 | obs 0.0167 ret 0.0118 | gn 0.04 | dW_q 1.83e-01 | 2.37 s/step | 19 GB
|
| 265 |
+
step 11850 | loss 0.0123 | obs 0.0154 ret 0.0109 | gn 0.04 | dW_q 1.79e-01 | 2.51 s/step | 19 GB
|
| 266 |
+
step 11900 | loss 0.0146 | obs 0.0173 ret 0.0153 | gn 0.04 | dW_q 1.89e-01 | 2.47 s/step | 19 GB
|
| 267 |
+
step 11950 | loss 0.0142 | obs 0.0172 ret 0.0142 | gn 0.04 | dW_q 1.73e-01 | 2.32 s/step | 19 GB
|
| 268 |
+
step 12000 | loss 0.0141 | obs 0.0177 ret 0.0150 | gn 0.04 | dW_q 1.92e-01 | 2.47 s/step | 19 GB
|
| 269 |
+
[1;34mwandb[0m:
|
| 270 |
+
[1;34mwandb[0m: 🚀 View run [33mrk_q[0m at: [34m[0m
|
| 271 |
+
[1;34mwandb[0m: Find logs at: [1;35m../../../../../../wv/E1_rubik_20261104031751/rk_q/wandb/run-20261105_174950-ywo1hfj6/logs[0m
|
rubik_r1_20261106011120/logs/rk_qkv.log
ADDED
|
@@ -0,0 +1,286 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
run rk_qkv | MemConfig(proj='qkv', ffn=False, layers=(0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11), fast_heads=4, blocks=4, ffn_frac=0.25, ffn_blocks=8, loss='delta', key='raw', lr=0.001, renorm=True) | max_ctx 0 | new params 11.99M | fast weights / sample: writable 2.99M, state 11.94M (46 MB fp32) | start step 0
|
| 2 |
+
wandb: Currently logged in as: xsourse to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
|
| 3 |
+
wandb: creating run
|
| 4 |
+
wandb: Tracking run with wandb version 0.22.0
|
| 5 |
+
wandb: Run data is saved locally in /mnt/localssd/lehany/wv/E1_rubik_20261104031751/rk_qkv/wandb/run-20261104_071615-4zc2ujbq
|
| 6 |
+
wandb: Run `wandb offline` to turn off syncing.
|
| 7 |
+
wandb: Syncing run rk_qkv
|
| 8 |
+
wandb: ⭐️ View project at https://wandb.ai/xsourse/world-auto-visual
|
| 9 |
+
wandb: 🚀 View run at https://wandb.ai/xsourse/world-auto-visual/runs/4zc2ujbq
|
| 10 |
+
step 50 | loss 0.0619 | obs 0.0596 ret 0.0600 | gn 0.12 | dW_q 1.39e-04 dW_k 1.36e-04 dW_v 2.17e-04 | 3.37 s/step | 21 GB
|
| 11 |
+
step 100 | loss 0.0565 | obs 0.0553 ret 0.0588 | gn 0.07 | dW_q 1.30e-03 dW_k 1.40e-03 dW_v 1.82e-03 | 2.99 s/step | 21 GB
|
| 12 |
+
step 150 | loss 0.0543 | obs 0.0548 ret 0.0540 | gn 0.06 | dW_q 2.88e-03 dW_k 2.98e-03 dW_v 4.73e-03 | 3.06 s/step | 21 GB
|
| 13 |
+
step 200 | loss 0.0552 | obs 0.0555 ret 0.0537 | gn 0.05 | dW_q 5.12e-03 dW_k 5.63e-03 dW_v 7.63e-03 | 3.12 s/step | 21 GB
|
| 14 |
+
step 250 | loss 0.0534 | obs 0.0518 ret 0.0531 | gn 0.04 | dW_q 7.07e-03 dW_k 7.42e-03 dW_v 1.79e-02 | 2.98 s/step | 21 GB
|
| 15 |
+
step 300 | loss 0.0525 | obs 0.0525 ret 0.0533 | gn 0.04 | dW_q 1.08e-02 dW_k 8.72e-03 dW_v 2.10e-02 | 3.20 s/step | 21 GB
|
| 16 |
+
step 350 | loss 0.0537 | obs 0.0515 ret 0.0562 | gn 0.04 | dW_q 1.14e-02 dW_k 1.06e-02 dW_v 2.32e-02 | 3.23 s/step | 21 GB
|
| 17 |
+
step 400 | loss 0.0548 | obs 0.0528 ret 0.0533 | gn 0.04 | dW_q 1.22e-02 dW_k 1.24e-02 dW_v 2.69e-02 | 3.15 s/step | 21 GB
|
| 18 |
+
step 450 | loss 0.0541 | obs 0.0543 ret 0.0513 | gn 0.05 | dW_q 1.34e-02 dW_k 1.38e-02 dW_v 2.69e-02 | 3.11 s/step | 21 GB
|
| 19 |
+
step 500 | loss 0.0532 | obs 0.0513 ret 0.0568 | gn 0.05 | dW_q 2.25e-02 dW_k 1.93e-02 dW_v 3.76e-02 | 3.10 s/step | 21 GB
|
| 20 |
+
step 550 | loss 0.0526 | obs 0.0508 ret 0.0513 | gn 0.05 | dW_q 1.78e-02 dW_k 2.31e-02 dW_v 4.65e-02 | 3.15 s/step | 21 GB
|
| 21 |
+
step 600 | loss 0.0540 | obs 0.0508 ret 0.0532 | gn 0.05 | dW_q 2.65e-02 dW_k 2.64e-02 dW_v 5.89e-02 | 3.17 s/step | 21 GB
|
| 22 |
+
step 650 | loss 0.0516 | obs 0.0463 ret 0.0505 | gn 0.05 | dW_q 2.72e-02 dW_k 3.29e-02 dW_v 6.71e-02 | 2.99 s/step | 21 GB
|
| 23 |
+
step 700 | loss 0.0526 | obs 0.0478 ret 0.0545 | gn 0.06 | dW_q 3.10e-02 dW_k 4.10e-02 dW_v 8.53e-02 | 3.12 s/step | 21 GB
|
| 24 |
+
step 750 | loss 0.0493 | obs 0.0452 ret 0.0547 | gn 0.06 | dW_q 4.89e-02 dW_k 6.15e-02 dW_v 1.11e-01 | 3.37 s/step | 21 GB
|
| 25 |
+
step 800 | loss 0.0439 | obs 0.0401 ret 0.0532 | gn 0.11 | dW_q 4.83e-02 dW_k 8.93e-02 dW_v 1.32e-01 | 3.25 s/step | 21 GB
|
| 26 |
+
step 850 | loss 0.0382 | obs 0.0356 ret 0.0492 | gn 0.10 | dW_q 5.11e-02 dW_k 8.39e-02 dW_v 1.19e-01 | 3.17 s/step | 21 GB
|
| 27 |
+
step 900 | loss 0.0315 | obs 0.0280 ret 0.0437 | gn 0.08 | dW_q 5.72e-02 dW_k 8.14e-02 dW_v 1.54e-01 | 3.02 s/step | 21 GB
|
| 28 |
+
step 950 | loss 0.0279 | obs 0.0234 ret 0.0341 | gn 0.06 | dW_q 4.95e-02 dW_k 7.84e-02 dW_v 1.38e-01 | 2.88 s/step | 21 GB
|
| 29 |
+
step 1000 | loss 0.0247 | obs 0.0246 ret 0.0274 | gn 0.06 | dW_q 5.00e-02 dW_k 8.22e-02 dW_v 1.30e-01 | 3.02 s/step | 21 GB
|
| 30 |
+
step 1050 | loss 0.0223 | obs 0.0211 ret 0.0244 | gn 0.06 | dW_q 4.61e-02 dW_k 8.09e-02 dW_v 1.32e-01 | 3.13 s/step | 21 GB
|
| 31 |
+
step 1100 | loss 0.0210 | obs 0.0214 ret 0.0267 | gn 0.05 | dW_q 5.06e-02 dW_k 9.01e-02 dW_v 1.48e-01 | 3.20 s/step | 21 GB
|
| 32 |
+
step 1150 | loss 0.0187 | obs 0.0178 ret 0.0207 | gn 0.04 | dW_q 5.13e-02 dW_k 8.82e-02 dW_v 1.35e-01 | 3.04 s/step | 21 GB
|
| 33 |
+
step 1200 | loss 0.0174 | obs 0.0179 ret 0.0184 | gn 0.04 | dW_q 5.11e-02 dW_k 7.78e-02 dW_v 1.44e-01 | 3.16 s/step | 21 GB
|
| 34 |
+
step 1250 | loss 0.0183 | obs 0.0192 ret 0.0181 | gn 0.04 | dW_q 5.20e-02 dW_k 8.11e-02 dW_v 1.42e-01 | 2.97 s/step | 21 GB
|
| 35 |
+
step 1300 | loss 0.0179 | obs 0.0173 ret 0.0187 | gn 0.04 | dW_q 4.81e-02 dW_k 8.17e-02 dW_v 1.35e-01 | 3.12 s/step | 21 GB
|
| 36 |
+
step 1350 | loss 0.0187 | obs 0.0176 ret 0.0221 | gn 0.04 | dW_q 5.33e-02 dW_k 7.77e-02 dW_v 1.40e-01 | 3.01 s/step | 21 GB
|
| 37 |
+
step 1400 | loss 0.0165 | obs 0.0169 ret 0.0189 | gn 0.04 | dW_q 5.73e-02 dW_k 9.36e-02 dW_v 1.43e-01 | 3.21 s/step | 21 GB
|
| 38 |
+
step 1450 | loss 0.0160 | obs 0.0179 ret 0.0172 | gn 0.04 | dW_q 4.84e-02 dW_k 7.68e-02 dW_v 1.34e-01 | 3.04 s/step | 21 GB
|
| 39 |
+
step 1500 | loss 0.0167 | obs 0.0167 ret 0.0199 | gn 0.04 | dW_q 5.69e-02 dW_k 7.52e-02 dW_v 1.40e-01 | 3.17 s/step | 21 GB
|
| 40 |
+
step 1550 | loss 0.0177 | obs 0.0189 ret 0.0182 | gn 0.04 | dW_q 4.90e-02 dW_k 6.90e-02 dW_v 1.34e-01 | 3.08 s/step | 21 GB
|
| 41 |
+
step 1600 | loss 0.0158 | obs 0.0169 ret 0.0178 | gn 0.04 | dW_q 5.95e-02 dW_k 7.77e-02 dW_v 1.44e-01 | 2.99 s/step | 21 GB
|
| 42 |
+
step 1650 | loss 0.0150 | obs 0.0154 ret 0.0134 | gn 0.04 | dW_q 6.04e-02 dW_k 8.81e-02 dW_v 1.44e-01 | 3.10 s/step | 21 GB
|
| 43 |
+
run rk_qkv | MemConfig(proj='qkv', ffn=False, layers=(0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11), fast_heads=4, blocks=4, ffn_frac=0.25, ffn_blocks=8, loss='delta', key='raw', lr=0.001, renorm=True) | max_ctx 0 | new params 11.99M | fast weights / sample: writable 2.99M, state 11.94M (46 MB fp32) | start step 1500
|
| 44 |
+
wandb: Currently logged in as: xsourse to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
|
| 45 |
+
wandb: creating run
|
| 46 |
+
wandb: Tracking run with wandb version 0.22.0
|
| 47 |
+
wandb: Run data is saved locally in /mnt/localssd/lehany/wv/E1_rubik_20261104031751/rk_qkv/wandb/run-20261105_121205-4zc2ujbq
|
| 48 |
+
wandb: Run `wandb offline` to turn off syncing.
|
| 49 |
+
wandb: Resuming run rk_qkv
|
| 50 |
+
wandb: ⭐️ View project at https://wandb.ai/xsourse/world-auto-visual
|
| 51 |
+
wandb: 🚀 View run at https://wandb.ai/xsourse/world-auto-visual/runs/4zc2ujbq
|
| 52 |
+
step 1550 | loss 0.0162 | obs 0.0180 ret 0.0168 | gn 0.04 | dW_q 5.32e-02 dW_k 6.66e-02 dW_v 1.34e-01 | 3.54 s/step | 22 GB
|
| 53 |
+
wandb: WARNING Tried to log to step 1550 that is less than the current step 1601. Steps must be monotonically increasing, so this data will be ignored. See https://wandb.me/define-metric to log data out of order.
|
| 54 |
+
step 1600 | loss 0.0160 | obs 0.0170 ret 0.0154 | gn 0.04 | dW_q 6.08e-02 dW_k 7.31e-02 dW_v 1.40e-01 | 3.38 s/step | 22 GB
|
| 55 |
+
wandb: WARNING Tried to log to step 1600 that is less than the current step 1601. Steps must be monotonically increasing, so this data will be ignored. See https://wandb.me/define-metric to log data out of order.
|
| 56 |
+
step 1650 | loss 0.0163 | obs 0.0154 ret 0.0163 | gn 0.04 | dW_q 5.77e-02 dW_k 6.88e-02 dW_v 1.40e-01 | 3.42 s/step | 22 GB
|
| 57 |
+
step 1700 | loss 0.0158 | obs 0.0166 ret 0.0176 | gn 0.04 | dW_q 5.15e-02 dW_k 6.35e-02 dW_v 1.33e-01 | 3.18 s/step | 22 GB
|
| 58 |
+
step 1750 | loss 0.0156 | obs 0.0157 ret 0.0163 | gn 0.04 | dW_q 5.67e-02 dW_k 7.42e-02 dW_v 1.40e-01 | 3.34 s/step | 22 GB
|
| 59 |
+
step 1800 | loss 0.0156 | obs 0.0175 ret 0.0153 | gn 0.04 | dW_q 4.75e-02 dW_k 6.12e-02 dW_v 1.32e-01 | 3.42 s/step | 22 GB
|
| 60 |
+
step 1850 | loss 0.0158 | obs 0.0170 ret 0.0190 | gn 0.04 | dW_q 5.31e-02 dW_k 7.75e-02 dW_v 1.41e-01 | 3.36 s/step | 22 GB
|
| 61 |
+
step 1900 | loss 0.0168 | obs 0.0187 ret 0.0194 | gn 0.04 | dW_q 5.84e-02 dW_k 7.51e-02 dW_v 1.39e-01 | 3.23 s/step | 22 GB
|
| 62 |
+
step 1950 | loss 0.0164 | obs 0.0172 ret 0.0143 | gn 0.04 | dW_q 5.33e-02 dW_k 7.36e-02 dW_v 1.36e-01 | 2.93 s/step | 22 GB
|
| 63 |
+
step 2000 | loss 0.0153 | obs 0.0175 ret 0.0155 | gn 0.04 | dW_q 5.82e-02 dW_k 7.57e-02 dW_v 1.43e-01 | 3.13 s/step | 22 GB
|
| 64 |
+
step 2050 | loss 0.0150 | obs 0.0155 ret 0.0190 | gn 0.04 | dW_q 5.56e-02 dW_k 7.78e-02 dW_v 1.36e-01 | 3.32 s/step | 22 GB
|
| 65 |
+
step 2100 | loss 0.0171 | obs 0.0189 ret 0.0202 | gn 0.05 | dW_q 6.43e-02 dW_k 8.25e-02 dW_v 1.42e-01 | 3.28 s/step | 22 GB
|
| 66 |
+
step 2150 | loss 0.0152 | obs 0.0153 ret 0.0199 | gn 0.04 | dW_q 6.65e-02 dW_k 7.45e-02 dW_v 1.43e-01 | 3.25 s/step | 22 GB
|
| 67 |
+
step 2200 | loss 0.0169 | obs 0.0166 ret 0.0221 | gn 0.04 | dW_q 6.10e-02 dW_k 7.35e-02 dW_v 1.41e-01 | 3.34 s/step | 22 GB
|
| 68 |
+
step 2250 | loss 0.0156 | obs 0.0173 ret 0.0150 | gn 0.04 | dW_q 6.97e-02 dW_k 9.26e-02 dW_v 1.50e-01 | 3.02 s/step | 22 GB
|
| 69 |
+
step 2300 | loss 0.0157 | obs 0.0171 ret 0.0156 | gn 0.04 | dW_q 6.64e-02 dW_k 7.03e-02 dW_v 1.44e-01 | 3.16 s/step | 22 GB
|
| 70 |
+
step 2350 | loss 0.0153 | obs 0.0158 ret 0.0179 | gn 0.04 | dW_q 6.14e-02 dW_k 7.59e-02 dW_v 1.38e-01 | 2.94 s/step | 22 GB
|
| 71 |
+
step 2400 | loss 0.0150 | obs 0.0174 ret 0.0172 | gn 0.04 | dW_q 6.38e-02 dW_k 7.69e-02 dW_v 1.44e-01 | 3.21 s/step | 22 GB
|
| 72 |
+
step 2450 | loss 0.0164 | obs 0.0200 ret 0.0213 | gn 0.04 | dW_q 6.59e-02 dW_k 7.49e-02 dW_v 1.43e-01 | 3.04 s/step | 22 GB
|
| 73 |
+
step 2500 | loss 0.0150 | obs 0.0143 ret 0.0173 | gn 0.04 | dW_q 5.89e-02 dW_k 6.80e-02 dW_v 1.38e-01 | 3.12 s/step | 22 GB
|
| 74 |
+
step 2550 | loss 0.0155 | obs 0.0182 ret 0.0197 | gn 0.04 | dW_q 6.44e-02 dW_k 7.72e-02 dW_v 1.41e-01 | 3.16 s/step | 22 GB
|
| 75 |
+
step 2600 | loss 0.0143 | obs 0.0149 ret 0.0195 | gn 0.04 | dW_q 6.33e-02 dW_k 6.13e-02 dW_v 1.38e-01 | 3.42 s/step | 22 GB
|
| 76 |
+
step 2650 | loss 0.0141 | obs 0.0148 ret 0.0161 | gn 0.04 | dW_q 6.99e-02 dW_k 7.90e-02 dW_v 1.45e-01 | 3.32 s/step | 22 GB
|
| 77 |
+
step 2700 | loss 0.0130 | obs 0.0139 ret 0.0150 | gn 0.04 | dW_q 7.11e-02 dW_k 7.89e-02 dW_v 1.42e-01 | 3.56 s/step | 22 GB
|
| 78 |
+
step 2750 | loss 0.0149 | obs 0.0160 ret 0.0183 | gn 0.04 | dW_q 7.16e-02 dW_k 7.85e-02 dW_v 1.43e-01 | 3.22 s/step | 22 GB
|
| 79 |
+
step 2800 | loss 0.0154 | obs 0.0174 ret 0.0187 | gn 0.04 | dW_q 7.37e-02 dW_k 8.01e-02 dW_v 1.47e-01 | 3.46 s/step | 22 GB
|
| 80 |
+
step 2850 | loss 0.0154 | obs 0.0195 ret 0.0167 | gn 0.04 | dW_q 6.86e-02 dW_k 7.80e-02 dW_v 1.44e-01 | 3.04 s/step | 22 GB
|
| 81 |
+
step 2900 | loss 0.0147 | obs 0.0168 ret 0.0171 | gn 0.04 | dW_q 7.30e-02 dW_k 8.03e-02 dW_v 1.46e-01 | 3.03 s/step | 22 GB
|
| 82 |
+
step 2950 | loss 0.0133 | obs 0.0166 ret 0.0137 | gn 0.04 | dW_q 7.61e-02 dW_k 8.20e-02 dW_v 1.47e-01 | 3.16 s/step | 22 GB
|
| 83 |
+
step 3000 | loss 0.0142 | obs 0.0154 ret 0.0167 | gn 0.04 | dW_q 7.81e-02 dW_k 8.45e-02 dW_v 1.50e-01 | 3.31 s/step | 22 GB
|
| 84 |
+
step 3050 | loss 0.0153 | obs 0.0165 ret 0.0174 | gn 0.05 | dW_q 6.17e-02 dW_k 7.78e-02 dW_v 1.36e-01 | 3.25 s/step | 22 GB
|
| 85 |
+
step 3100 | loss 0.0145 | obs 0.0160 ret 0.0220 | gn 0.04 | dW_q 7.61e-02 dW_k 7.85e-02 dW_v 1.46e-01 | 3.65 s/step | 22 GB
|
| 86 |
+
step 3150 | loss 0.0126 | obs 0.0144 ret 0.0150 | gn 0.04 | dW_q 7.79e-02 dW_k 8.05e-02 dW_v 1.48e-01 | 3.10 s/step | 22 GB
|
| 87 |
+
step 3200 | loss 0.0151 | obs 0.0176 ret 0.0179 | gn 0.04 | dW_q 6.55e-02 dW_k 6.92e-02 dW_v 1.44e-01 | 3.16 s/step | 22 GB
|
| 88 |
+
step 3250 | loss 0.0153 | obs 0.0160 ret 0.0193 | gn 0.04 | dW_q 6.51e-02 dW_k 6.94e-02 dW_v 1.39e-01 | 2.95 s/step | 22 GB
|
| 89 |
+
step 3300 | loss 0.0143 | obs 0.0145 ret 0.0161 | gn 0.04 | dW_q 6.66e-02 dW_k 7.61e-02 dW_v 1.43e-01 | 3.00 s/step | 22 GB
|
| 90 |
+
step 3350 | loss 0.0135 | obs 0.0143 ret 0.0165 | gn 0.04 | dW_q 6.51e-02 dW_k 6.50e-02 dW_v 1.38e-01 | 2.97 s/step | 22 GB
|
| 91 |
+
step 3400 | loss 0.0155 | obs 0.0174 ret 0.0144 | gn 0.04 | dW_q 5.78e-02 dW_k 5.88e-02 dW_v 1.36e-01 | 3.01 s/step | 22 GB
|
| 92 |
+
step 3450 | loss 0.0153 | obs 0.0177 ret 0.0141 | gn 0.04 | dW_q 7.57e-02 dW_k 7.59e-02 dW_v 1.46e-01 | 3.05 s/step | 22 GB
|
| 93 |
+
step 3500 | loss 0.0148 | obs 0.0145 ret 0.0155 | gn 0.04 | dW_q 7.14e-02 dW_k 7.54e-02 dW_v 1.45e-01 | 3.02 s/step | 22 GB
|
| 94 |
+
step 3550 | loss 0.0141 | obs 0.0172 ret 0.0164 | gn 0.04 | dW_q 7.10e-02 dW_k 7.95e-02 dW_v 1.43e-01 | 3.14 s/step | 22 GB
|
| 95 |
+
step 3600 | loss 0.0151 | obs 0.0167 ret 0.0171 | gn 0.04 | dW_q 7.53e-02 dW_k 8.08e-02 dW_v 1.46e-01 | 3.21 s/step | 22 GB
|
| 96 |
+
step 3650 | loss 0.0143 | obs 0.0166 ret 0.0167 | gn 0.04 | dW_q 7.24e-02 dW_k 8.33e-02 dW_v 1.45e-01 | 3.54 s/step | 22 GB
|
| 97 |
+
step 3700 | loss 0.0144 | obs 0.0187 ret 0.0155 | gn 0.04 | dW_q 8.05e-02 dW_k 8.70e-02 dW_v 1.50e-01 | 3.34 s/step | 22 GB
|
| 98 |
+
step 3750 | loss 0.0145 | obs 0.0173 ret 0.0172 | gn 0.04 | dW_q 7.70e-02 dW_k 8.17e-02 dW_v 1.49e-01 | 3.45 s/step | 22 GB
|
| 99 |
+
step 3800 | loss 0.0129 | obs 0.0142 ret 0.0144 | gn 0.04 | dW_q 6.38e-02 dW_k 7.68e-02 dW_v 1.42e-01 | 3.27 s/step | 22 GB
|
| 100 |
+
step 3850 | loss 0.0147 | obs 0.0177 ret 0.0139 | gn 0.04 | dW_q 6.92e-02 dW_k 7.43e-02 dW_v 1.38e-01 | 3.12 s/step | 22 GB
|
| 101 |
+
step 3900 | loss 0.0124 | obs 0.0154 ret 0.0142 | gn 0.04 | dW_q 7.41e-02 dW_k 8.27e-02 dW_v 1.47e-01 | 3.25 s/step | 22 GB
|
| 102 |
+
step 3950 | loss 0.0142 | obs 0.0169 ret 0.0189 | gn 0.04 | dW_q 7.71e-02 dW_k 8.23e-02 dW_v 1.49e-01 | 3.27 s/step | 22 GB
|
| 103 |
+
step 4000 | loss 0.0129 | obs 0.0140 ret 0.0125 | gn 0.04 | dW_q 6.93e-02 dW_k 7.81e-02 dW_v 1.45e-01 | 2.99 s/step | 22 GB
|
| 104 |
+
step 4050 | loss 0.0132 | obs 0.0156 ret 0.0148 | gn 0.04 | dW_q 6.83e-02 dW_k 8.85e-02 dW_v 1.48e-01 | 3.91 s/step | 22 GB
|
| 105 |
+
step 4100 | loss 0.0139 | obs 0.0162 ret 0.0153 | gn 0.04 | dW_q 7.37e-02 dW_k 8.48e-02 dW_v 1.48e-01 | 3.14 s/step | 22 GB
|
| 106 |
+
step 4150 | loss 0.0143 | obs 0.0171 ret 0.0187 | gn 0.04 | dW_q 8.10e-02 dW_k 8.54e-02 dW_v 1.53e-01 | 3.07 s/step | 22 GB
|
| 107 |
+
step 4200 | loss 0.0141 | obs 0.0151 ret 0.0154 | gn 0.04 | dW_q 7.90e-02 dW_k 8.92e-02 dW_v 1.50e-01 | 3.09 s/step | 22 GB
|
| 108 |
+
step 4250 | loss 0.0156 | obs 0.0198 ret 0.0223 | gn 0.04 | dW_q 8.16e-02 dW_k 9.12e-02 dW_v 1.52e-01 | 2.97 s/step | 22 GB
|
| 109 |
+
step 4300 | loss 0.0127 | obs 0.0141 ret 0.0143 | gn 0.04 | dW_q 6.91e-02 dW_k 7.34e-02 dW_v 1.40e-01 | 2.89 s/step | 22 GB
|
| 110 |
+
step 4350 | loss 0.0144 | obs 0.0183 ret 0.0142 | gn 0.04 | dW_q 8.31e-02 dW_k 8.97e-02 dW_v 1.52e-01 | 3.89 s/step | 22 GB
|
| 111 |
+
step 4400 | loss 0.0138 | obs 0.0160 ret 0.0132 | gn 0.04 | dW_q 7.54e-02 dW_k 8.16e-02 dW_v 1.49e-01 | 3.54 s/step | 22 GB
|
| 112 |
+
step 4450 | loss 0.0126 | obs 0.0155 ret 0.0138 | gn 0.04 | dW_q 8.25e-02 dW_k 8.46e-02 dW_v 1.52e-01 | 3.65 s/step | 22 GB
|
| 113 |
+
step 4500 | loss 0.0138 | obs 0.0172 ret 0.0117 | gn 0.04 | dW_q 8.28e-02 dW_k 8.49e-02 dW_v 1.51e-01 | 3.56 s/step | 22 GB
|
| 114 |
+
step 4550 | loss 0.0130 | obs 0.0163 ret 0.0123 | gn 0.04 | dW_q 7.47e-02 dW_k 8.38e-02 dW_v 1.45e-01 | 3.57 s/step | 22 GB
|
| 115 |
+
step 4600 | loss 0.0122 | obs 0.0154 ret 0.0124 | gn 0.04 | dW_q 8.70e-02 dW_k 8.16e-02 dW_v 1.54e-01 | 3.58 s/step | 22 GB
|
| 116 |
+
step 4650 | loss 0.0123 | obs 0.0141 ret 0.0128 | gn 0.04 | dW_q 7.53e-02 dW_k 8.26e-02 dW_v 1.49e-01 | 3.57 s/step | 22 GB
|
| 117 |
+
step 4700 | loss 0.0134 | obs 0.0148 ret 0.0198 | gn 0.04 | dW_q 7.85e-02 dW_k 8.44e-02 dW_v 1.47e-01 | 3.48 s/step | 22 GB
|
| 118 |
+
step 4750 | loss 0.0138 | obs 0.0185 ret 0.0120 | gn 0.04 | dW_q 7.05e-02 dW_k 7.78e-02 dW_v 1.45e-01 | 3.03 s/step | 22 GB
|
| 119 |
+
step 4800 | loss 0.0136 | obs 0.0163 ret 0.0146 | gn 0.04 | dW_q 9.12e-02 dW_k 9.21e-02 dW_v 1.58e-01 | 3.18 s/step | 22 GB
|
| 120 |
+
step 4850 | loss 0.0133 | obs 0.0157 ret 0.0150 | gn 0.03 | dW_q 8.67e-02 dW_k 9.53e-02 dW_v 1.45e-01 | 3.15 s/step | 22 GB
|
| 121 |
+
step 4900 | loss 0.0137 | obs 0.0165 ret 0.0183 | gn 0.04 | dW_q 8.01e-02 dW_k 8.05e-02 dW_v 1.50e-01 | 3.11 s/step | 22 GB
|
| 122 |
+
step 4950 | loss 0.0139 | obs 0.0175 ret 0.0151 | gn 0.04 | dW_q 7.02e-02 dW_k 7.69e-02 dW_v 1.41e-01 | 3.02 s/step | 22 GB
|
| 123 |
+
step 5000 | loss 0.0143 | obs 0.0172 ret 0.0142 | gn 0.04 | dW_q 8.12e-02 dW_k 8.97e-02 dW_v 1.50e-01 | 2.88 s/step | 22 GB
|
| 124 |
+
step 5050 | loss 0.0133 | obs 0.0164 ret 0.0168 | gn 0.04 | dW_q 8.85e-02 dW_k 8.38e-02 dW_v 1.55e-01 | 4.01 s/step | 22 GB
|
| 125 |
+
step 5100 | loss 0.0140 | obs 0.0170 ret 0.0159 | gn 0.04 | dW_q 9.06e-02 dW_k 8.71e-02 dW_v 1.59e-01 | 3.59 s/step | 22 GB
|
| 126 |
+
step 5150 | loss 0.0139 | obs 0.0156 ret 0.0223 | gn 0.04 | dW_q 8.49e-02 dW_k 8.31e-02 dW_v 1.56e-01 | 3.26 s/step | 22 GB
|
| 127 |
+
step 5200 | loss 0.0136 | obs 0.0173 ret 0.0101 | gn 0.04 | dW_q 8.94e-02 dW_k 8.74e-02 dW_v 1.53e-01 | 2.99 s/step | 22 GB
|
| 128 |
+
step 5250 | loss 0.0132 | obs 0.0140 ret 0.0176 | gn 0.04 | dW_q 7.40e-02 dW_k 7.85e-02 dW_v 1.45e-01 | 3.08 s/step | 22 GB
|
| 129 |
+
step 5300 | loss 0.0142 | obs 0.0182 ret 0.0119 | gn 0.04 | dW_q 8.30e-02 dW_k 7.94e-02 dW_v 1.51e-01 | 2.90 s/step | 22 GB
|
| 130 |
+
step 5350 | loss 0.0133 | obs 0.0145 ret 0.0158 | gn 0.04 | dW_q 9.50e-02 dW_k 9.72e-02 dW_v 1.56e-01 | 3.07 s/step | 22 GB
|
| 131 |
+
step 5400 | loss 0.0139 | obs 0.0171 ret 0.0118 | gn 0.04 | dW_q 9.60e-02 dW_k 8.33e-02 dW_v 1.57e-01 | 3.07 s/step | 22 GB
|
| 132 |
+
step 5450 | loss 0.0141 | obs 0.0162 ret 0.0123 | gn 0.04 | dW_q 7.76e-02 dW_k 7.30e-02 dW_v 1.45e-01 | 3.39 s/step | 22 GB
|
| 133 |
+
step 5500 | loss 0.0134 | obs 0.0163 ret 0.0117 | gn 0.04 | dW_q 1.03e-01 dW_k 9.29e-02 dW_v 1.59e-01 | 3.44 s/step | 22 GB
|
| 134 |
+
step 5550 | loss 0.0134 | obs 0.0155 ret 0.0179 | gn 0.04 | dW_q 1.01e-01 dW_k 9.11e-02 dW_v 1.60e-01 | 3.12 s/step | 22 GB
|
| 135 |
+
step 5600 | loss 0.0134 | obs 0.0163 ret 0.0126 | gn 0.04 | dW_q 9.82e-02 dW_k 9.25e-02 dW_v 1.61e-01 | 3.14 s/step | 22 GB
|
| 136 |
+
step 5650 | loss 0.0132 | obs 0.0151 ret 0.0145 | gn 0.04 | dW_q 1.09e-01 dW_k 9.45e-02 dW_v 1.65e-01 | 2.93 s/step | 22 GB
|
| 137 |
+
step 5700 | loss 0.0132 | obs 0.0158 ret 0.0134 | gn 0.04 | dW_q 8.83e-02 dW_k 8.43e-02 dW_v 1.57e-01 | 3.18 s/step | 22 GB
|
| 138 |
+
step 5750 | loss 0.0142 | obs 0.0181 ret 0.0167 | gn 0.04 | dW_q 9.61e-02 dW_k 9.46e-02 dW_v 1.56e-01 | 3.06 s/step | 22 GB
|
| 139 |
+
step 5800 | loss 0.0142 | obs 0.0185 ret 0.0169 | gn 0.04 | dW_q 9.97e-02 dW_k 9.57e-02 dW_v 1.60e-01 | 3.17 s/step | 22 GB
|
| 140 |
+
step 5850 | loss 0.0128 | obs 0.0155 ret 0.0119 | gn 0.03 | dW_q 9.39e-02 dW_k 8.38e-02 dW_v 1.50e-01 | 3.05 s/step | 22 GB
|
| 141 |
+
step 5900 | loss 0.0126 | obs 0.0143 ret 0.0172 | gn 0.04 | dW_q 9.78e-02 dW_k 8.43e-02 dW_v 1.58e-01 | 2.97 s/step | 22 GB
|
| 142 |
+
step 5950 | loss 0.0127 | obs 0.0157 ret 0.0097 | gn 0.04 | dW_q 9.93e-02 dW_k 8.57e-02 dW_v 1.58e-01 | 2.95 s/step | 22 GB
|
| 143 |
+
step 6000 | loss 0.0132 | obs 0.0152 ret 0.0173 | gn 0.04 | dW_q 7.92e-02 dW_k 6.91e-02 dW_v 1.44e-01 | 2.95 s/step | 22 GB
|
| 144 |
+
step 6050 | loss 0.0127 | obs 0.0149 ret 0.0139 | gn 0.04 | dW_q 8.64e-02 dW_k 7.67e-02 dW_v 1.49e-01 | 3.02 s/step | 22 GB
|
| 145 |
+
step 6100 | loss 0.0141 | obs 0.0174 ret 0.0163 | gn 0.04 | dW_q 1.07e-01 dW_k 8.96e-02 dW_v 1.64e-01 | 3.33 s/step | 22 GB
|
| 146 |
+
step 6150 | loss 0.0135 | obs 0.0155 ret 0.0159 | gn 0.04 | dW_q 1.02e-01 dW_k 8.61e-02 dW_v 1.62e-01 | 3.20 s/step | 22 GB
|
| 147 |
+
step 6200 | loss 0.0133 | obs 0.0163 ret 0.0151 | gn 0.04 | dW_q 1.05e-01 dW_k 9.59e-02 dW_v 1.63e-01 | 3.02 s/step | 22 GB
|
| 148 |
+
step 6250 | loss 0.0129 | obs 0.0153 ret 0.0111 | gn 0.04 | dW_q 9.10e-02 dW_k 8.69e-02 dW_v 1.51e-01 | 2.95 s/step | 22 GB
|
| 149 |
+
step 6300 | loss 0.0120 | obs 0.0157 ret 0.0134 | gn 0.04 | dW_q 9.63e-02 dW_k 8.61e-02 dW_v 1.58e-01 | 2.99 s/step | 22 GB
|
| 150 |
+
step 6350 | loss 0.0125 | obs 0.0156 ret 0.0126 | gn 0.03 | dW_q 1.10e-01 dW_k 9.57e-02 dW_v 1.67e-01 | 3.26 s/step | 22 GB
|
| 151 |
+
step 6400 | loss 0.0132 | obs 0.0171 ret 0.0134 | gn 0.05 | dW_q 9.73e-02 dW_k 9.29e-02 dW_v 1.54e-01 | 3.08 s/step | 22 GB
|
| 152 |
+
step 6450 | loss 0.0133 | obs 0.0181 ret 0.0153 | gn 0.04 | dW_q 1.02e-01 dW_k 9.56e-02 dW_v 1.63e-01 | 3.15 s/step | 22 GB
|
| 153 |
+
step 6500 | loss 0.0126 | obs 0.0160 ret 0.0134 | gn 0.04 | dW_q 9.38e-02 dW_k 8.08e-02 dW_v 1.53e-01 | 2.91 s/step | 22 GB
|
| 154 |
+
step 6550 | loss 0.0127 | obs 0.0148 ret 0.0122 | gn 0.04 | dW_q 1.01e-01 dW_k 8.98e-02 dW_v 1.65e-01 | 3.24 s/step | 22 GB
|
| 155 |
+
step 6600 | loss 0.0123 | obs 0.0150 ret 0.0103 | gn 0.04 | dW_q 1.04e-01 dW_k 9.06e-02 dW_v 1.60e-01 | 2.95 s/step | 22 GB
|
| 156 |
+
step 6650 | loss 0.0120 | obs 0.0150 ret 0.0124 | gn 0.04 | dW_q 9.68e-02 dW_k 8.72e-02 dW_v 1.52e-01 | 2.82 s/step | 22 GB
|
| 157 |
+
step 6700 | loss 0.0133 | obs 0.0200 ret 0.0118 | gn 0.04 | dW_q 1.08e-01 dW_k 9.19e-02 dW_v 1.61e-01 | 3.07 s/step | 22 GB
|
| 158 |
+
step 6750 | loss 0.0143 | obs 0.0177 ret 0.0130 | gn 0.04 | dW_q 1.03e-01 dW_k 8.76e-02 dW_v 1.57e-01 | 3.02 s/step | 22 GB
|
| 159 |
+
step 6800 | loss 0.0128 | obs 0.0169 ret 0.0139 | gn 0.04 | dW_q 1.01e-01 dW_k 8.49e-02 dW_v 1.57e-01 | 3.00 s/step | 22 GB
|
| 160 |
+
step 6850 | loss 0.0124 | obs 0.0156 ret 0.0142 | gn 0.04 | dW_q 8.35e-02 dW_k 7.50e-02 dW_v 1.44e-01 | 3.31 s/step | 22 GB
|
| 161 |
+
step 6900 | loss 0.0123 | obs 0.0147 ret 0.0115 | gn 0.04 | dW_q 1.19e-01 dW_k 9.25e-02 dW_v 1.65e-01 | 3.16 s/step | 22 GB
|
| 162 |
+
step 6950 | loss 0.0121 | obs 0.0164 ret 0.0120 | gn 0.04 | dW_q 9.94e-02 dW_k 8.90e-02 dW_v 1.59e-01 | 2.87 s/step | 22 GB
|
| 163 |
+
step 7000 | loss 0.0133 | obs 0.0163 ret 0.0134 | gn 0.04 | dW_q 1.05e-01 dW_k 8.24e-02 dW_v 1.55e-01 | 3.03 s/step | 22 GB
|
| 164 |
+
step 7050 | loss 0.0121 | obs 0.0159 ret 0.0128 | gn 0.04 | dW_q 1.04e-01 dW_k 9.04e-02 dW_v 1.58e-01 | 3.08 s/step | 22 GB
|
| 165 |
+
step 7100 | loss 0.0132 | obs 0.0156 ret 0.0131 | gn 0.04 | dW_q 9.59e-02 dW_k 8.02e-02 dW_v 1.52e-01 | 2.98 s/step | 22 GB
|
| 166 |
+
step 7150 | loss 0.0117 | obs 0.0156 ret 0.0086 | gn 0.04 | dW_q 8.55e-02 dW_k 7.12e-02 dW_v 1.47e-01 | 2.84 s/step | 22 GB
|
| 167 |
+
step 7200 | loss 0.0143 | obs 0.0174 ret 0.0131 | gn 0.04 | dW_q 9.63e-02 dW_k 8.06e-02 dW_v 1.53e-01 | 3.04 s/step | 22 GB
|
| 168 |
+
step 7250 | loss 0.0123 | obs 0.0152 ret 0.0152 | gn 0.04 | dW_q 1.11e-01 dW_k 9.31e-02 dW_v 1.66e-01 | 3.16 s/step | 22 GB
|
| 169 |
+
step 7300 | loss 0.0130 | obs 0.0147 ret 0.0117 | gn 0.04 | dW_q 9.35e-02 dW_k 7.58e-02 dW_v 1.49e-01 | 3.16 s/step | 22 GB
|
| 170 |
+
step 7350 | loss 0.0125 | obs 0.0169 ret 0.0122 | gn 0.04 | dW_q 1.08e-01 dW_k 8.63e-02 dW_v 1.65e-01 | 3.11 s/step | 22 GB
|
| 171 |
+
step 7400 | loss 0.0112 | obs 0.0146 ret 0.0131 | gn 0.04 | dW_q 8.71e-02 dW_k 8.42e-02 dW_v 1.47e-01 | 2.83 s/step | 22 GB
|
| 172 |
+
step 7450 | loss 0.0115 | obs 0.0146 ret 0.0106 | gn 0.04 | dW_q 9.73e-02 dW_k 8.71e-02 dW_v 1.59e-01 | 2.95 s/step | 22 GB
|
| 173 |
+
step 7500 | loss 0.0120 | obs 0.0170 ret 0.0109 | gn 0.04 | dW_q 8.39e-02 dW_k 6.81e-02 dW_v 1.47e-01 | 2.93 s/step | 22 GB
|
| 174 |
+
step 7550 | loss 0.0111 | obs 0.0145 ret 0.0088 | gn 0.05 | dW_q 1.02e-01 dW_k 8.97e-02 dW_v 1.58e-01 | 2.92 s/step | 22 GB
|
| 175 |
+
step 7600 | loss 0.0143 | obs 0.0176 ret 0.0138 | gn 0.05 | dW_q 1.00e-01 dW_k 8.50e-02 dW_v 1.55e-01 | 2.96 s/step | 22 GB
|
| 176 |
+
step 7650 | loss 0.0129 | obs 0.0156 ret 0.0133 | gn 0.04 | dW_q 1.10e-01 dW_k 9.49e-02 dW_v 1.68e-01 | 3.07 s/step | 22 GB
|
| 177 |
+
step 7700 | loss 0.0122 | obs 0.0157 ret 0.0130 | gn 0.04 | dW_q 9.22e-02 dW_k 7.57e-02 dW_v 1.52e-01 | 3.00 s/step | 22 GB
|
| 178 |
+
step 7750 | loss 0.0115 | obs 0.0150 ret 0.0121 | gn 0.04 | dW_q 1.01e-01 dW_k 8.41e-02 dW_v 1.63e-01 | 3.00 s/step | 22 GB
|
| 179 |
+
step 7800 | loss 0.0113 | obs 0.0127 ret 0.0117 | gn 0.04 | dW_q 1.03e-01 dW_k 9.16e-02 dW_v 1.57e-01 | 3.11 s/step | 22 GB
|
| 180 |
+
run rk_qkv | MemConfig(proj='qkv', ffn=False, layers=(0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11), fast_heads=4, blocks=4, ffn_frac=0.25, ffn_blocks=8, loss='delta', key='raw', lr=0.001, renorm=True) | max_ctx 0 | new params 11.99M | fast weights / sample: writable 2.99M, state 11.94M (46 MB fp32) | start step 7500
|
| 181 |
+
wandb: Currently logged in as: xsourse to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
|
| 182 |
+
wandb: creating run
|
| 183 |
+
wandb: Tracking run with wandb version 0.22.0
|
| 184 |
+
wandb: Run data is saved locally in /mnt/localssd/lehany/wv/E1_rubik_20261104031751/rk_qkv/wandb/run-20261105_174950-4zc2ujbq
|
| 185 |
+
wandb: Run `wandb offline` to turn off syncing.
|
| 186 |
+
wandb: Resuming run rk_qkv
|
| 187 |
+
wandb: ⭐️ View project at https://wandb.ai/xsourse/world-auto-visual
|
| 188 |
+
wandb: 🚀 View run at https://wandb.ai/xsourse/world-auto-visual/runs/4zc2ujbq
|
| 189 |
+
step 7550 | loss 0.0119 | obs 0.0153 ret 0.0118 | gn 0.04 | dW_q 8.39e-02 dW_k 7.76e-02 dW_v 1.46e-01 | 3.41 s/step | 22 GB
|
| 190 |
+
wandb: WARNING Tried to log to step 7550 that is less than the current step 7751. Steps must be monotonically increasing, so this data will be ignored. See https://wandb.me/define-metric to log data out of order.
|
| 191 |
+
step 7600 | loss 0.0123 | obs 0.0152 ret 0.0158 | gn 0.04 | dW_q 1.12e-01 dW_k 9.03e-02 dW_v 1.64e-01 | 3.56 s/step | 22 GB
|
| 192 |
+
wandb: WARNING Tried to log to step 7600 that is less than the current step 7751. Steps must be monotonically increasing, so this data will be ignored. See https://wandb.me/define-metric to log data out of order.
|
| 193 |
+
step 7650 | loss 0.0123 | obs 0.0166 ret 0.0104 | gn 0.04 | dW_q 1.13e-01 dW_k 9.06e-02 dW_v 1.70e-01 | 3.35 s/step | 22 GB
|
| 194 |
+
wandb: WARNING Tried to log to step 7650 that is less than the current step 7751. Steps must be monotonically increasing, so this data will be ignored. See https://wandb.me/define-metric to log data out of order.
|
| 195 |
+
step 7700 | loss 0.0123 | obs 0.0174 ret 0.0124 | gn 0.04 | dW_q 1.12e-01 dW_k 8.64e-02 dW_v 1.65e-01 | 3.21 s/step | 22 GB
|
| 196 |
+
wandb: WARNING Tried to log to step 7700 that is less than the current step 7751. Steps must be monotonically increasing, so this data will be ignored. See https://wandb.me/define-metric to log data out of order.
|
| 197 |
+
step 7750 | loss 0.0117 | obs 0.0152 ret 0.0137 | gn 0.04 | dW_q 8.60e-02 dW_k 7.09e-02 dW_v 1.46e-01 | 3.27 s/step | 22 GB
|
| 198 |
+
wandb: WARNING Tried to log to step 7750 that is less than the current step 7751. Steps must be monotonically increasing, so this data will be ignored. See https://wandb.me/define-metric to log data out of order.
|
| 199 |
+
step 7800 | loss 0.0119 | obs 0.0154 ret 0.0118 | gn 0.04 | dW_q 9.22e-02 dW_k 7.65e-02 dW_v 1.53e-01 | 3.22 s/step | 22 GB
|
| 200 |
+
step 7850 | loss 0.0120 | obs 0.0152 ret 0.0079 | gn 0.03 | dW_q 1.09e-01 dW_k 9.10e-02 dW_v 1.66e-01 | 3.51 s/step | 22 GB
|
| 201 |
+
step 7900 | loss 0.0133 | obs 0.0169 ret 0.0164 | gn 0.04 | dW_q 1.13e-01 dW_k 9.33e-02 dW_v 1.68e-01 | 3.35 s/step | 22 GB
|
| 202 |
+
step 7950 | loss 0.0126 | obs 0.0172 ret 0.0130 | gn 0.04 | dW_q 1.12e-01 dW_k 9.49e-02 dW_v 1.70e-01 | 3.13 s/step | 22 GB
|
| 203 |
+
step 8000 | loss 0.0116 | obs 0.0142 ret 0.0130 | gn 0.04 | dW_q 9.85e-02 dW_k 8.23e-02 dW_v 1.57e-01 | 3.02 s/step | 22 GB
|
| 204 |
+
step 8050 | loss 0.0112 | obs 0.0148 ret 0.0080 | gn 0.04 | dW_q 1.20e-01 dW_k 9.95e-02 dW_v 1.75e-01 | 3.19 s/step | 22 GB
|
| 205 |
+
step 8100 | loss 0.0132 | obs 0.0159 ret 0.0146 | gn 0.04 | dW_q 1.01e-01 dW_k 8.93e-02 dW_v 1.55e-01 | 3.07 s/step | 22 GB
|
| 206 |
+
step 8150 | loss 0.0120 | obs 0.0141 ret 0.0112 | gn 0.04 | dW_q 1.06e-01 dW_k 9.20e-02 dW_v 1.64e-01 | 3.33 s/step | 22 GB
|
| 207 |
+
step 8200 | loss 0.0135 | obs 0.0177 ret 0.0178 | gn 0.04 | dW_q 1.16e-01 dW_k 8.78e-02 dW_v 1.66e-01 | 3.37 s/step | 22 GB
|
| 208 |
+
step 8250 | loss 0.0123 | obs 0.0162 ret 0.0118 | gn 0.04 | dW_q 1.05e-01 dW_k 8.96e-02 dW_v 1.64e-01 | 3.19 s/step | 22 GB
|
| 209 |
+
step 8300 | loss 0.0124 | obs 0.0155 ret 0.0149 | gn 0.04 | dW_q 1.20e-01 dW_k 9.41e-02 dW_v 1.70e-01 | 3.14 s/step | 22 GB
|
| 210 |
+
step 8350 | loss 0.0122 | obs 0.0157 ret 0.0122 | gn 0.04 | dW_q 1.15e-01 dW_k 9.68e-02 dW_v 1.73e-01 | 2.99 s/step | 22 GB
|
| 211 |
+
step 8400 | loss 0.0123 | obs 0.0165 ret 0.0143 | gn 0.04 | dW_q 1.09e-01 dW_k 8.75e-02 dW_v 1.66e-01 | 3.29 s/step | 22 GB
|
| 212 |
+
step 8450 | loss 0.0132 | obs 0.0151 ret 0.0189 | gn 0.04 | dW_q 1.15e-01 dW_k 9.50e-02 dW_v 1.71e-01 | 3.18 s/step | 22 GB
|
| 213 |
+
step 8500 | loss 0.0125 | obs 0.0165 ret 0.0122 | gn 0.04 | dW_q 1.11e-01 dW_k 8.85e-02 dW_v 1.64e-01 | 3.00 s/step | 22 GB
|
| 214 |
+
step 8550 | loss 0.0120 | obs 0.0155 ret 0.0107 | gn 0.04 | dW_q 1.17e-01 dW_k 9.49e-02 dW_v 1.66e-01 | 3.21 s/step | 22 GB
|
| 215 |
+
step 8600 | loss 0.0112 | obs 0.0148 ret 0.0109 | gn 0.04 | dW_q 1.20e-01 dW_k 9.75e-02 dW_v 1.76e-01 | 3.18 s/step | 22 GB
|
| 216 |
+
step 8650 | loss 0.0117 | obs 0.0152 ret 0.0107 | gn 0.04 | dW_q 1.01e-01 dW_k 8.65e-02 dW_v 1.60e-01 | 3.09 s/step | 22 GB
|
| 217 |
+
step 8700 | loss 0.0111 | obs 0.0147 ret 0.0109 | gn 0.03 | dW_q 1.00e-01 dW_k 8.41e-02 dW_v 1.57e-01 | 3.05 s/step | 22 GB
|
| 218 |
+
step 8750 | loss 0.0126 | obs 0.0156 ret 0.0158 | gn 0.04 | dW_q 1.02e-01 dW_k 8.49e-02 dW_v 1.59e-01 | 3.16 s/step | 22 GB
|
| 219 |
+
step 8800 | loss 0.0122 | obs 0.0139 ret 0.0140 | gn 0.04 | dW_q 1.18e-01 dW_k 9.55e-02 dW_v 1.73e-01 | 3.22 s/step | 22 GB
|
| 220 |
+
step 8850 | loss 0.0129 | obs 0.0169 ret 0.0157 | gn 0.04 | dW_q 1.05e-01 dW_k 8.29e-02 dW_v 1.63e-01 | 3.00 s/step | 22 GB
|
| 221 |
+
step 8900 | loss 0.0118 | obs 0.0152 ret 0.0107 | gn 0.04 | dW_q 1.12e-01 dW_k 8.88e-02 dW_v 1.65e-01 | 3.07 s/step | 22 GB
|
| 222 |
+
step 8950 | loss 0.0114 | obs 0.0153 ret 0.0080 | gn 0.03 | dW_q 1.03e-01 dW_k 8.47e-02 dW_v 1.58e-01 | 3.20 s/step | 22 GB
|
| 223 |
+
step 9000 | loss 0.0120 | obs 0.0138 ret 0.0154 | gn 0.04 | dW_q 1.09e-01 dW_k 9.09e-02 dW_v 1.69e-01 | 3.07 s/step | 22 GB
|
| 224 |
+
step 9050 | loss 0.0122 | obs 0.0170 ret 0.0110 | gn 0.04 | dW_q 9.39e-02 dW_k 8.08e-02 dW_v 1.55e-01 | 3.16 s/step | 22 GB
|
| 225 |
+
step 9100 | loss 0.0116 | obs 0.0152 ret 0.0101 | gn 0.03 | dW_q 1.06e-01 dW_k 9.11e-02 dW_v 1.66e-01 | 3.14 s/step | 22 GB
|
| 226 |
+
step 9150 | loss 0.0112 | obs 0.0153 ret 0.0114 | gn 0.04 | dW_q 9.15e-02 dW_k 8.16e-02 dW_v 1.53e-01 | 3.14 s/step | 22 GB
|
| 227 |
+
step 9200 | loss 0.0120 | obs 0.0171 ret 0.0133 | gn 0.04 | dW_q 9.87e-02 dW_k 8.67e-02 dW_v 1.63e-01 | 3.08 s/step | 22 GB
|
| 228 |
+
step 9250 | loss 0.0130 | obs 0.0155 ret 0.0086 | gn 0.04 | dW_q 1.15e-01 dW_k 9.05e-02 dW_v 1.69e-01 | 3.19 s/step | 22 GB
|
| 229 |
+
step 9300 | loss 0.0118 | obs 0.0156 ret 0.0088 | gn 0.04 | dW_q 1.29e-01 dW_k 1.03e-01 dW_v 1.88e-01 | 3.25 s/step | 22 GB
|
| 230 |
+
step 9350 | loss 0.0111 | obs 0.0141 ret 0.0101 | gn 0.03 | dW_q 8.67e-02 dW_k 6.98e-02 dW_v 1.48e-01 | 3.23 s/step | 22 GB
|
| 231 |
+
step 9400 | loss 0.0130 | obs 0.0170 ret 0.0115 | gn 0.04 | dW_q 8.80e-02 dW_k 7.39e-02 dW_v 1.49e-01 | 2.99 s/step | 22 GB
|
| 232 |
+
step 9450 | loss 0.0132 | obs 0.0159 ret 0.0149 | gn 0.04 | dW_q 1.16e-01 dW_k 9.67e-02 dW_v 1.77e-01 | 3.16 s/step | 22 GB
|
| 233 |
+
step 9500 | loss 0.0124 | obs 0.0142 ret 0.0121 | gn 0.03 | dW_q 1.27e-01 dW_k 1.03e-01 dW_v 1.77e-01 | 3.22 s/step | 22 GB
|
| 234 |
+
step 9550 | loss 0.0124 | obs 0.0171 ret 0.0144 | gn 0.04 | dW_q 1.05e-01 dW_k 8.83e-02 dW_v 1.67e-01 | 3.32 s/step | 22 GB
|
| 235 |
+
step 9600 | loss 0.0127 | obs 0.0168 ret 0.0089 | gn 0.04 | dW_q 1.24e-01 dW_k 9.17e-02 dW_v 1.71e-01 | 3.13 s/step | 22 GB
|
| 236 |
+
step 9650 | loss 0.0122 | obs 0.0155 ret 0.0126 | gn 0.04 | dW_q 1.34e-01 dW_k 1.05e-01 dW_v 1.86e-01 | 3.15 s/step | 22 GB
|
| 237 |
+
step 9700 | loss 0.0125 | obs 0.0144 ret 0.0119 | gn 0.04 | dW_q 9.36e-02 dW_k 8.44e-02 dW_v 1.59e-01 | 3.18 s/step | 22 GB
|
| 238 |
+
step 9750 | loss 0.0124 | obs 0.0139 ret 0.0112 | gn 0.04 | dW_q 8.84e-02 dW_k 7.63e-02 dW_v 1.51e-01 | 3.16 s/step | 22 GB
|
| 239 |
+
step 9800 | loss 0.0112 | obs 0.0136 ret 0.0103 | gn 0.04 | dW_q 9.73e-02 dW_k 8.23e-02 dW_v 1.62e-01 | 3.13 s/step | 22 GB
|
| 240 |
+
step 9850 | loss 0.0123 | obs 0.0166 ret 0.0110 | gn 0.04 | dW_q 9.82e-02 dW_k 8.27e-02 dW_v 1.55e-01 | 3.25 s/step | 22 GB
|
| 241 |
+
step 9900 | loss 0.0106 | obs 0.0138 ret 0.0116 | gn 0.04 | dW_q 1.02e-01 dW_k 8.79e-02 dW_v 1.64e-01 | 3.16 s/step | 22 GB
|
| 242 |
+
step 9950 | loss 0.0116 | obs 0.0150 ret 0.0131 | gn 0.04 | dW_q 1.17e-01 dW_k 1.01e-01 dW_v 1.79e-01 | 3.06 s/step | 22 GB
|
| 243 |
+
step 10000 | loss 0.0119 | obs 0.0149 ret 0.0126 | gn 0.04 | dW_q 1.25e-01 dW_k 1.02e-01 dW_v 1.79e-01 | 3.20 s/step | 22 GB
|
| 244 |
+
step 10050 | loss 0.0109 | obs 0.0136 ret 0.0088 | gn 0.03 | dW_q 1.13e-01 dW_k 8.81e-02 dW_v 1.67e-01 | 3.46 s/step | 22 GB
|
| 245 |
+
step 10100 | loss 0.0126 | obs 0.0160 ret 0.0143 | gn 0.04 | dW_q 1.15e-01 dW_k 1.02e-01 dW_v 1.79e-01 | 2.97 s/step | 22 GB
|
| 246 |
+
step 10150 | loss 0.0127 | obs 0.0151 ret 0.0103 | gn 0.04 | dW_q 1.11e-01 dW_k 9.52e-02 dW_v 1.72e-01 | 3.15 s/step | 22 GB
|
| 247 |
+
step 10200 | loss 0.0128 | obs 0.0138 ret 0.0159 | gn 0.04 | dW_q 1.07e-01 dW_k 8.92e-02 dW_v 1.63e-01 | 3.12 s/step | 22 GB
|
| 248 |
+
step 10250 | loss 0.0139 | obs 0.0179 ret 0.0145 | gn 0.04 | dW_q 1.09e-01 dW_k 9.47e-02 dW_v 1.70e-01 | 3.31 s/step | 22 GB
|
| 249 |
+
step 10300 | loss 0.0111 | obs 0.0139 ret 0.0115 | gn 0.04 | dW_q 1.04e-01 dW_k 8.38e-02 dW_v 1.62e-01 | 2.97 s/step | 22 GB
|
| 250 |
+
step 10350 | loss 0.0130 | obs 0.0182 ret 0.0102 | gn 0.04 | dW_q 1.15e-01 dW_k 9.61e-02 dW_v 1.76e-01 | 3.32 s/step | 22 GB
|
| 251 |
+
step 10400 | loss 0.0121 | obs 0.0169 ret 0.0089 | gn 0.04 | dW_q 1.00e-01 dW_k 8.84e-02 dW_v 1.63e-01 | 3.08 s/step | 22 GB
|
| 252 |
+
step 10450 | loss 0.0108 | obs 0.0142 ret 0.0104 | gn 0.04 | dW_q 1.02e-01 dW_k 8.69e-02 dW_v 1.59e-01 | 3.11 s/step | 22 GB
|
| 253 |
+
step 10500 | loss 0.0123 | obs 0.0140 ret 0.0123 | gn 0.04 | dW_q 1.41e-01 dW_k 1.05e-01 dW_v 1.87e-01 | 3.19 s/step | 22 GB
|
| 254 |
+
step 10550 | loss 0.0107 | obs 0.0137 ret 0.0109 | gn 0.03 | dW_q 1.18e-01 dW_k 9.99e-02 dW_v 1.77e-01 | 3.29 s/step | 22 GB
|
| 255 |
+
step 10600 | loss 0.0105 | obs 0.0148 ret 0.0107 | gn 0.03 | dW_q 1.05e-01 dW_k 9.21e-02 dW_v 1.70e-01 | 2.93 s/step | 22 GB
|
| 256 |
+
step 10650 | loss 0.0103 | obs 0.0139 ret 0.0106 | gn 0.04 | dW_q 9.13e-02 dW_k 8.18e-02 dW_v 1.51e-01 | 3.22 s/step | 22 GB
|
| 257 |
+
step 10700 | loss 0.0120 | obs 0.0157 ret 0.0098 | gn 0.03 | dW_q 1.10e-01 dW_k 1.00e-01 dW_v 1.85e-01 | 3.21 s/step | 22 GB
|
| 258 |
+
step 10750 | loss 0.0122 | obs 0.0153 ret 0.0127 | gn 0.04 | dW_q 1.12e-01 dW_k 1.04e-01 dW_v 1.75e-01 | 3.26 s/step | 22 GB
|
| 259 |
+
step 10800 | loss 0.0115 | obs 0.0164 ret 0.0110 | gn 0.03 | dW_q 1.31e-01 dW_k 1.07e-01 dW_v 1.89e-01 | 3.32 s/step | 22 GB
|
| 260 |
+
step 10850 | loss 0.0122 | obs 0.0150 ret 0.0115 | gn 0.03 | dW_q 9.16e-02 dW_k 7.21e-02 dW_v 1.51e-01 | 3.36 s/step | 22 GB
|
| 261 |
+
step 10900 | loss 0.0120 | obs 0.0165 ret 0.0120 | gn 0.04 | dW_q 1.11e-01 dW_k 9.94e-02 dW_v 1.76e-01 | 3.00 s/step | 22 GB
|
| 262 |
+
step 10950 | loss 0.0123 | obs 0.0169 ret 0.0130 | gn 0.03 | dW_q 1.11e-01 dW_k 9.51e-02 dW_v 1.77e-01 | 3.00 s/step | 22 GB
|
| 263 |
+
step 11000 | loss 0.0116 | obs 0.0144 ret 0.0100 | gn 0.04 | dW_q 1.03e-01 dW_k 8.49e-02 dW_v 1.64e-01 | 3.18 s/step | 22 GB
|
| 264 |
+
step 11050 | loss 0.0112 | obs 0.0150 ret 0.0107 | gn 0.04 | dW_q 1.18e-01 dW_k 1.00e-01 dW_v 1.84e-01 | 3.21 s/step | 22 GB
|
| 265 |
+
step 11100 | loss 0.0123 | obs 0.0175 ret 0.0095 | gn 0.04 | dW_q 1.16e-01 dW_k 1.05e-01 dW_v 1.81e-01 | 3.17 s/step | 22 GB
|
| 266 |
+
step 11150 | loss 0.0128 | obs 0.0167 ret 0.0188 | gn 0.04 | dW_q 1.06e-01 dW_k 9.62e-02 dW_v 1.68e-01 | 3.13 s/step | 22 GB
|
| 267 |
+
step 11200 | loss 0.0121 | obs 0.0157 ret 0.0128 | gn 0.04 | dW_q 1.12e-01 dW_k 9.44e-02 dW_v 1.71e-01 | 3.30 s/step | 22 GB
|
| 268 |
+
step 11250 | loss 0.0113 | obs 0.0158 ret 0.0081 | gn 0.04 | dW_q 1.11e-01 dW_k 8.69e-02 dW_v 1.68e-01 | 3.25 s/step | 22 GB
|
| 269 |
+
step 11300 | loss 0.0135 | obs 0.0172 ret 0.0135 | gn 0.04 | dW_q 9.48e-02 dW_k 7.77e-02 dW_v 1.56e-01 | 3.15 s/step | 22 GB
|
| 270 |
+
step 11350 | loss 0.0117 | obs 0.0171 ret 0.0104 | gn 0.04 | dW_q 1.02e-01 dW_k 8.75e-02 dW_v 1.71e-01 | 3.16 s/step | 22 GB
|
| 271 |
+
step 11400 | loss 0.0131 | obs 0.0164 ret 0.0142 | gn 0.04 | dW_q 1.21e-01 dW_k 1.04e-01 dW_v 1.88e-01 | 3.35 s/step | 22 GB
|
| 272 |
+
step 11450 | loss 0.0125 | obs 0.0165 ret 0.0141 | gn 0.04 | dW_q 9.36e-02 dW_k 8.01e-02 dW_v 1.59e-01 | 3.11 s/step | 22 GB
|
| 273 |
+
step 11500 | loss 0.0114 | obs 0.0144 ret 0.0104 | gn 0.04 | dW_q 1.05e-01 dW_k 8.75e-02 dW_v 1.69e-01 | 3.00 s/step | 22 GB
|
| 274 |
+
step 11550 | loss 0.0114 | obs 0.0166 ret 0.0111 | gn 0.03 | dW_q 1.18e-01 dW_k 1.05e-01 dW_v 1.94e-01 | 3.24 s/step | 22 GB
|
| 275 |
+
step 11600 | loss 0.0120 | obs 0.0169 ret 0.0147 | gn 0.03 | dW_q 8.84e-02 dW_k 7.40e-02 dW_v 1.50e-01 | 3.24 s/step | 22 GB
|
| 276 |
+
step 11650 | loss 0.0115 | obs 0.0141 ret 0.0123 | gn 0.04 | dW_q 1.24e-01 dW_k 1.01e-01 dW_v 1.82e-01 | 3.02 s/step | 22 GB
|
| 277 |
+
step 11700 | loss 0.0122 | obs 0.0160 ret 0.0130 | gn 0.03 | dW_q 1.13e-01 dW_k 9.79e-02 dW_v 1.77e-01 | 3.20 s/step | 22 GB
|
| 278 |
+
step 11750 | loss 0.0128 | obs 0.0166 ret 0.0118 | gn 0.04 | dW_q 1.29e-01 dW_k 1.12e-01 dW_v 2.02e-01 | 3.23 s/step | 22 GB
|
| 279 |
+
step 11800 | loss 0.0119 | obs 0.0148 ret 0.0148 | gn 0.03 | dW_q 9.40e-02 dW_k 8.45e-02 dW_v 1.60e-01 | 3.22 s/step | 22 GB
|
| 280 |
+
step 11850 | loss 0.0118 | obs 0.0161 ret 0.0093 | gn 0.03 | dW_q 1.10e-01 dW_k 9.27e-02 dW_v 1.78e-01 | 3.04 s/step | 22 GB
|
| 281 |
+
step 11900 | loss 0.0110 | obs 0.0153 ret 0.0109 | gn 0.04 | dW_q 1.25e-01 dW_k 1.07e-01 dW_v 1.81e-01 | 3.20 s/step | 22 GB
|
| 282 |
+
step 11950 | loss 0.0114 | obs 0.0157 ret 0.0118 | gn 0.04 | dW_q 1.08e-01 dW_k 9.00e-02 dW_v 1.71e-01 | 3.17 s/step | 22 GB
|
| 283 |
+
step 12000 | loss 0.0122 | obs 0.0182 ret 0.0116 | gn 0.04 | dW_q 1.38e-01 dW_k 1.14e-01 dW_v 1.85e-01 | 3.09 s/step | 22 GB
|
| 284 |
+
[1;34mwandb[0m:
|
| 285 |
+
[1;34mwandb[0m: 🚀 View run [33mrk_qkv[0m at: [34m[0m
|
| 286 |
+
[1;34mwandb[0m: Find logs at: [1;35m../../../../../../wv/E1_rubik_20261104031751/rk_qkv/wandb/run-20261105_174950-4zc2ujbq/logs[0m
|
rubik_r1_20261106011120/logs/rk_qkv_norm.log
ADDED
|
@@ -0,0 +1,290 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
run rk_qkv_norm | MemConfig(proj='qkv', ffn=False, layers=(0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11), fast_heads=4, blocks=4, ffn_frac=0.25, ffn_blocks=8, loss='delta', key='norm', lr=0.5, renorm=True) | max_ctx 0 | new params 11.99M | fast weights / sample: writable 2.99M, state 11.94M (46 MB fp32) | start step 0
|
| 2 |
+
wandb: Currently logged in as: xsourse to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
|
| 3 |
+
wandb: creating run
|
| 4 |
+
wandb: Tracking run with wandb version 0.22.0
|
| 5 |
+
wandb: Run data is saved locally in /mnt/localssd/lehany/wv/E1_rubik_20261104031751/rk_qkv_norm/wandb/run-20261104_071615-816i4yzf
|
| 6 |
+
wandb: Run `wandb offline` to turn off syncing.
|
| 7 |
+
wandb: Syncing run rk_qkv_norm
|
| 8 |
+
wandb: ⭐️ View project at https://wandb.ai/xsourse/world-auto-visual
|
| 9 |
+
wandb: 🚀 View run at https://wandb.ai/xsourse/world-auto-visual/runs/816i4yzf
|
| 10 |
+
step 50 | loss 0.0619 | obs 0.0596 ret 0.0600 | gn 0.12 | dW_q 3.10e-03 dW_k 3.26e-03 dW_v 3.75e-03 | 3.77 s/step | 21 GB
|
| 11 |
+
step 100 | loss 0.0564 | obs 0.0553 ret 0.0586 | gn 0.04 | dW_q 1.32e-02 dW_k 1.50e-02 dW_v 1.59e-02 | 3.34 s/step | 21 GB
|
| 12 |
+
step 150 | loss 0.0543 | obs 0.0548 ret 0.0536 | gn 0.04 | dW_q 2.31e-02 dW_k 2.56e-02 dW_v 2.78e-02 | 3.44 s/step | 21 GB
|
| 13 |
+
step 200 | loss 0.0552 | obs 0.0555 ret 0.0536 | gn 0.04 | dW_q 3.85e-02 dW_k 4.06e-02 dW_v 4.45e-02 | 3.48 s/step | 21 GB
|
| 14 |
+
step 250 | loss 0.0535 | obs 0.0520 ret 0.0531 | gn 0.04 | dW_q 4.62e-02 dW_k 4.85e-02 dW_v 5.48e-02 | 3.34 s/step | 21 GB
|
| 15 |
+
step 300 | loss 0.0526 | obs 0.0529 ret 0.0533 | gn 0.04 | dW_q 6.52e-02 dW_k 6.78e-02 dW_v 7.79e-02 | 3.61 s/step | 21 GB
|
| 16 |
+
step 350 | loss 0.0539 | obs 0.0519 ret 0.0562 | gn 0.04 | dW_q 8.05e-02 dW_k 8.56e-02 dW_v 9.72e-02 | 3.62 s/step | 21 GB
|
| 17 |
+
step 400 | loss 0.0549 | obs 0.0531 ret 0.0533 | gn 0.04 | dW_q 7.89e-02 dW_k 8.58e-02 dW_v 9.52e-02 | 3.52 s/step | 21 GB
|
| 18 |
+
step 450 | loss 0.0543 | obs 0.0546 ret 0.0514 | gn 0.05 | dW_q 8.75e-02 dW_k 9.57e-02 dW_v 1.03e-01 | 3.49 s/step | 21 GB
|
| 19 |
+
step 500 | loss 0.0534 | obs 0.0519 ret 0.0570 | gn 0.04 | dW_q 1.32e-01 dW_k 1.33e-01 dW_v 1.58e-01 | 3.45 s/step | 21 GB
|
| 20 |
+
step 550 | loss 0.0532 | obs 0.0520 ret 0.0528 | gn 0.07 | dW_q 8.60e-02 dW_k 9.19e-02 dW_v 1.02e-01 | 3.48 s/step | 21 GB
|
| 21 |
+
step 600 | loss 0.0548 | obs 0.0530 ret 0.0534 | gn 0.04 | dW_q 1.30e-01 dW_k 1.29e-01 dW_v 1.61e-01 | 3.56 s/step | 21 GB
|
| 22 |
+
step 650 | loss 0.0531 | obs 0.0504 ret 0.0508 | gn 0.05 | dW_q 1.14e-01 dW_k 1.20e-01 dW_v 1.41e-01 | 3.32 s/step | 21 GB
|
| 23 |
+
step 700 | loss 0.0550 | obs 0.0542 ret 0.0549 | gn 0.05 | dW_q 1.23e-01 dW_k 1.27e-01 dW_v 1.51e-01 | 3.46 s/step | 21 GB
|
| 24 |
+
step 750 | loss 0.0536 | obs 0.0545 ret 0.0551 | gn 0.04 | dW_q 1.73e-01 dW_k 1.65e-01 dW_v 2.16e-01 | 3.71 s/step | 21 GB
|
| 25 |
+
step 800 | loss 0.0531 | obs 0.0512 ret 0.0542 | gn 0.04 | dW_q 1.86e-01 dW_k 1.69e-01 dW_v 2.34e-01 | 3.67 s/step | 21 GB
|
| 26 |
+
step 850 | loss 0.0534 | obs 0.0529 ret 0.0546 | gn 0.04 | dW_q 1.64e-01 dW_k 1.54e-01 dW_v 2.04e-01 | 3.57 s/step | 21 GB
|
| 27 |
+
step 900 | loss 0.0536 | obs 0.0512 ret 0.0583 | gn 0.05 | dW_q 2.24e-01 dW_k 1.96e-01 dW_v 2.84e-01 | 3.40 s/step | 21 GB
|
| 28 |
+
step 950 | loss 0.0542 | obs 0.0507 ret 0.0548 | gn 0.05 | dW_q 1.79e-01 dW_k 1.70e-01 dW_v 2.23e-01 | 3.27 s/step | 21 GB
|
| 29 |
+
step 1000 | loss 0.0534 | obs 0.0533 ret 0.0531 | gn 0.05 | dW_q 1.56e-01 dW_k 1.56e-01 dW_v 1.92e-01 | 3.39 s/step | 21 GB
|
| 30 |
+
step 1050 | loss 0.0529 | obs 0.0516 ret 0.0515 | gn 0.04 | dW_q 1.71e-01 dW_k 1.67e-01 dW_v 2.10e-01 | 3.52 s/step | 21 GB
|
| 31 |
+
step 1100 | loss 0.0528 | obs 0.0520 ret 0.0551 | gn 0.04 | dW_q 2.58e-01 dW_k 2.38e-01 dW_v 3.24e-01 | 3.59 s/step | 21 GB
|
| 32 |
+
step 1150 | loss 0.0516 | obs 0.0496 ret 0.0510 | gn 0.04 | dW_q 1.84e-01 dW_k 1.85e-01 dW_v 2.27e-01 | 3.41 s/step | 21 GB
|
| 33 |
+
step 1200 | loss 0.0525 | obs 0.0516 ret 0.0482 | gn 0.04 | dW_q 2.35e-01 dW_k 2.29e-01 dW_v 2.92e-01 | 3.54 s/step | 21 GB
|
| 34 |
+
step 1250 | loss 0.0518 | obs 0.0510 ret 0.0498 | gn 0.05 | dW_q 2.65e-01 dW_k 2.55e-01 dW_v 3.28e-01 | 3.34 s/step | 21 GB
|
| 35 |
+
step 1300 | loss 0.0512 | obs 0.0506 ret 0.0478 | gn 0.05 | dW_q 2.10e-01 dW_k 2.14e-01 dW_v 2.50e-01 | 3.49 s/step | 21 GB
|
| 36 |
+
step 1350 | loss 0.0523 | obs 0.0502 ret 0.0521 | gn 0.05 | dW_q 2.63e-01 dW_k 2.66e-01 dW_v 3.18e-01 | 3.36 s/step | 21 GB
|
| 37 |
+
step 1400 | loss 0.0499 | obs 0.0485 ret 0.0491 | gn 0.05 | dW_q 3.10e-01 dW_k 3.01e-01 dW_v 3.80e-01 | 3.59 s/step | 21 GB
|
| 38 |
+
step 1450 | loss 0.0484 | obs 0.0480 ret 0.0502 | gn 0.05 | dW_q 2.29e-01 dW_k 2.35e-01 dW_v 2.72e-01 | 3.42 s/step | 21 GB
|
| 39 |
+
run rk_qkv_norm | MemConfig(proj='qkv', ffn=False, layers=(0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11), fast_heads=4, blocks=4, ffn_frac=0.25, ffn_blocks=8, loss='delta', key='norm', lr=0.5, renorm=True) | max_ctx 0 | new params 11.99M | fast weights / sample: writable 2.99M, state 11.94M (46 MB fp32) | start step 1000
|
| 40 |
+
wandb: Currently logged in as: xsourse to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
|
| 41 |
+
wandb: creating run
|
| 42 |
+
wandb: Tracking run with wandb version 0.22.0
|
| 43 |
+
wandb: Run data is saved locally in /mnt/localssd/lehany/wv/E1_rubik_20261104031751/rk_qkv_norm/wandb/run-20261105_121205-816i4yzf
|
| 44 |
+
wandb: Run `wandb offline` to turn off syncing.
|
| 45 |
+
wandb: Resuming run rk_qkv_norm
|
| 46 |
+
wandb: ⭐️ View project at https://wandb.ai/xsourse/world-auto-visual
|
| 47 |
+
wandb: 🚀 View run at https://wandb.ai/xsourse/world-auto-visual/runs/816i4yzf
|
| 48 |
+
step 1050 | loss 0.0524 | obs 0.0519 ret 0.0521 | gn 0.04 | dW_q 2.70e-01 dW_k 2.41e-01 dW_v 3.42e-01 | 3.91 s/step | 22 GB
|
| 49 |
+
wandb: WARNING Tried to log to step 1050 that is less than the current step 1401. Steps must be monotonically increasing, so this data will be ignored. See https://wandb.me/define-metric to log data out of order.
|
| 50 |
+
step 1100 | loss 0.0532 | obs 0.0514 ret 0.0518 | gn 0.05 | dW_q 2.27e-01 dW_k 2.15e-01 dW_v 2.87e-01 | 3.50 s/step | 22 GB
|
| 51 |
+
wandb: WARNING Tried to log to step 1100 that is less than the current step 1401. Steps must be monotonically increasing, so this data will be ignored. See https://wandb.me/define-metric to log data out of order.
|
| 52 |
+
step 1150 | loss 0.0522 | obs 0.0524 ret 0.0537 | gn 0.05 | dW_q 2.31e-01 dW_k 2.19e-01 dW_v 2.89e-01 | 3.55 s/step | 22 GB
|
| 53 |
+
wandb: WARNING Tried to log to step 1150 that is less than the current step 1401. Steps must be monotonically increasing, so this data will be ignored. See https://wandb.me/define-metric to log data out of order.
|
| 54 |
+
step 1200 | loss 0.0522 | obs 0.0492 ret 0.0528 | gn 0.05 | dW_q 2.74e-01 dW_k 2.59e-01 dW_v 3.47e-01 | 3.79 s/step | 22 GB
|
| 55 |
+
wandb: WARNING Tried to log to step 1200 that is less than the current step 1401. Steps must be monotonically increasing, so this data will be ignored. See https://wandb.me/define-metric to log data out of order.
|
| 56 |
+
step 1250 | loss 0.0509 | obs 0.0497 ret 0.0458 | gn 0.05 | dW_q 2.70e-01 dW_k 2.57e-01 dW_v 3.36e-01 | 3.75 s/step | 22 GB
|
| 57 |
+
wandb: WARNING Tried to log to step 1250 that is less than the current step 1401. Steps must be monotonically increasing, so this data will be ignored. See https://wandb.me/define-metric to log data out of order.
|
| 58 |
+
step 1300 | loss 0.0505 | obs 0.0481 ret 0.0548 | gn 0.05 | dW_q 2.26e-01 dW_k 2.28e-01 dW_v 2.72e-01 | 3.60 s/step | 22 GB
|
| 59 |
+
wandb: WARNING Tried to log to step 1300 that is less than the current step 1401. Steps must be monotonically increasing, so this data will be ignored. See https://wandb.me/define-metric to log data out of order.
|
| 60 |
+
step 1350 | loss 0.0504 | obs 0.0503 ret 0.0490 | gn 0.05 | dW_q 3.13e-01 dW_k 2.99e-01 dW_v 3.88e-01 | 3.59 s/step | 22 GB
|
| 61 |
+
wandb: WARNING Tried to log to step 1350 that is less than the current step 1401. Steps must be monotonically increasing, so this data will be ignored. See https://wandb.me/define-metric to log data out of order.
|
| 62 |
+
step 1400 | loss 0.0511 | obs 0.0489 ret 0.0514 | gn 0.05 | dW_q 2.83e-01 dW_k 2.80e-01 dW_v 3.44e-01 | 3.37 s/step | 22 GB
|
| 63 |
+
wandb: WARNING Tried to log to step 1400 that is less than the current step 1401. Steps must be monotonically increasing, so this data will be ignored. See https://wandb.me/define-metric to log data out of order.
|
| 64 |
+
step 1450 | loss 0.0497 | obs 0.0500 ret 0.0474 | gn 0.05 | dW_q 3.17e-01 dW_k 3.12e-01 dW_v 3.88e-01 | 3.33 s/step | 22 GB
|
| 65 |
+
step 1500 | loss 0.0488 | obs 0.0461 ret 0.0507 | gn 0.05 | dW_q 2.53e-01 dW_k 2.63e-01 dW_v 3.06e-01 | 3.27 s/step | 22 GB
|
| 66 |
+
step 1550 | loss 0.0477 | obs 0.0470 ret 0.0466 | gn 0.06 | dW_q 3.30e-01 dW_k 3.35e-01 dW_v 4.03e-01 | 3.81 s/step | 22 GB
|
| 67 |
+
step 1600 | loss 0.0485 | obs 0.0475 ret 0.0453 | gn 0.06 | dW_q 3.73e-01 dW_k 3.79e-01 dW_v 4.55e-01 | 3.78 s/step | 22 GB
|
| 68 |
+
step 1650 | loss 0.0472 | obs 0.0457 ret 0.0486 | gn 0.06 | dW_q 2.94e-01 dW_k 3.12e-01 dW_v 3.54e-01 | 3.47 s/step | 22 GB
|
| 69 |
+
step 1700 | loss 0.0470 | obs 0.0436 ret 0.0501 | gn 0.05 | dW_q 2.85e-01 dW_k 3.06e-01 dW_v 3.41e-01 | 3.51 s/step | 22 GB
|
| 70 |
+
step 1750 | loss 0.0467 | obs 0.0433 ret 0.0497 | gn 0.06 | dW_q 3.78e-01 dW_k 3.93e-01 dW_v 4.61e-01 | 3.54 s/step | 22 GB
|
| 71 |
+
step 1800 | loss 0.0457 | obs 0.0460 ret 0.0473 | gn 0.06 | dW_q 2.85e-01 dW_k 3.07e-01 dW_v 3.46e-01 | 3.51 s/step | 22 GB
|
| 72 |
+
step 1850 | loss 0.0436 | obs 0.0392 ret 0.0481 | gn 0.06 | dW_q 3.72e-01 dW_k 3.95e-01 dW_v 4.51e-01 | 3.32 s/step | 22 GB
|
| 73 |
+
step 1900 | loss 0.0438 | obs 0.0396 ret 0.0431 | gn 0.06 | dW_q 3.05e-01 dW_k 3.32e-01 dW_v 3.76e-01 | 3.55 s/step | 22 GB
|
| 74 |
+
step 1950 | loss 0.0437 | obs 0.0392 ret 0.0442 | gn 0.06 | dW_q 3.94e-01 dW_k 4.16e-01 dW_v 4.89e-01 | 3.58 s/step | 22 GB
|
| 75 |
+
step 2000 | loss 0.0411 | obs 0.0375 ret 0.0417 | gn 0.06 | dW_q 3.99e-01 dW_k 4.28e-01 dW_v 4.91e-01 | 3.54 s/step | 22 GB
|
| 76 |
+
step 2050 | loss 0.0414 | obs 0.0360 ret 0.0444 | gn 0.06 | dW_q 3.75e-01 dW_k 4.07e-01 dW_v 4.64e-01 | 3.65 s/step | 22 GB
|
| 77 |
+
step 2100 | loss 0.0398 | obs 0.0360 ret 0.0355 | gn 0.06 | dW_q 3.62e-01 dW_k 3.94e-01 dW_v 4.48e-01 | 3.60 s/step | 22 GB
|
| 78 |
+
step 2150 | loss 0.0387 | obs 0.0351 ret 0.0350 | gn 0.06 | dW_q 4.76e-01 dW_k 5.01e-01 dW_v 5.91e-01 | 3.71 s/step | 22 GB
|
| 79 |
+
step 2200 | loss 0.0384 | obs 0.0330 ret 0.0369 | gn 0.07 | dW_q 4.41e-01 dW_k 4.76e-01 dW_v 5.48e-01 | 3.50 s/step | 22 GB
|
| 80 |
+
step 2250 | loss 0.0379 | obs 0.0319 ret 0.0361 | gn 0.07 | dW_q 4.46e-01 dW_k 4.83e-01 dW_v 5.52e-01 | 3.29 s/step | 22 GB
|
| 81 |
+
step 2300 | loss 0.0369 | obs 0.0332 ret 0.0350 | gn 0.07 | dW_q 3.84e-01 dW_k 4.28e-01 dW_v 4.71e-01 | 3.53 s/step | 22 GB
|
| 82 |
+
step 2350 | loss 0.0368 | obs 0.0334 ret 0.0317 | gn 0.07 | dW_q 5.24e-01 dW_k 5.60e-01 dW_v 6.59e-01 | 3.49 s/step | 22 GB
|
| 83 |
+
step 2400 | loss 0.0352 | obs 0.0354 ret 0.0278 | gn 0.08 | dW_q 3.76e-01 dW_k 4.14e-01 dW_v 4.69e-01 | 3.41 s/step | 22 GB
|
| 84 |
+
step 2450 | loss 0.0335 | obs 0.0306 ret 0.0276 | gn 0.07 | dW_q 4.32e-01 dW_k 4.75e-01 dW_v 5.27e-01 | 3.71 s/step | 22 GB
|
| 85 |
+
step 2500 | loss 0.0323 | obs 0.0270 ret 0.0281 | gn 0.06 | dW_q 4.40e-01 dW_k 4.83e-01 dW_v 5.39e-01 | 3.60 s/step | 22 GB
|
| 86 |
+
step 2550 | loss 0.0341 | obs 0.0295 ret 0.0296 | gn 0.07 | dW_q 5.67e-01 dW_k 6.06e-01 dW_v 7.02e-01 | 3.40 s/step | 22 GB
|
| 87 |
+
step 2600 | loss 0.0324 | obs 0.0277 ret 0.0302 | gn 0.06 | dW_q 5.46e-01 dW_k 5.84e-01 dW_v 6.79e-01 | 3.52 s/step | 22 GB
|
| 88 |
+
step 2650 | loss 0.0311 | obs 0.0268 ret 0.0271 | gn 0.07 | dW_q 5.86e-01 dW_k 6.24e-01 dW_v 7.42e-01 | 3.46 s/step | 22 GB
|
| 89 |
+
step 2700 | loss 0.0312 | obs 0.0264 ret 0.0269 | gn 0.07 | dW_q 5.54e-01 dW_k 5.97e-01 dW_v 6.77e-01 | 3.37 s/step | 22 GB
|
| 90 |
+
step 2750 | loss 0.0308 | obs 0.0269 ret 0.0220 | gn 0.06 | dW_q 4.45e-01 dW_k 4.92e-01 dW_v 5.53e-01 | 3.46 s/step | 22 GB
|
| 91 |
+
step 2800 | loss 0.0307 | obs 0.0255 ret 0.0237 | gn 0.06 | dW_q 4.45e-01 dW_k 4.89e-01 dW_v 5.46e-01 | 3.55 s/step | 22 GB
|
| 92 |
+
step 2850 | loss 0.0282 | obs 0.0241 ret 0.0208 | gn 0.05 | dW_q 5.71e-01 dW_k 6.10e-01 dW_v 7.04e-01 | 3.80 s/step | 22 GB
|
| 93 |
+
step 2900 | loss 0.0292 | obs 0.0246 ret 0.0243 | gn 0.06 | dW_q 6.16e-01 dW_k 6.57e-01 dW_v 7.49e-01 | 3.93 s/step | 22 GB
|
| 94 |
+
step 2950 | loss 0.0295 | obs 0.0268 ret 0.0242 | gn 0.06 | dW_q 4.29e-01 dW_k 4.80e-01 dW_v 5.36e-01 | 4.02 s/step | 22 GB
|
| 95 |
+
step 3000 | loss 0.0283 | obs 0.0242 ret 0.0252 | gn 0.05 | dW_q 5.64e-01 dW_k 6.07e-01 dW_v 7.01e-01 | 3.58 s/step | 22 GB
|
| 96 |
+
step 3050 | loss 0.0268 | obs 0.0221 ret 0.0241 | gn 0.06 | dW_q 4.64e-01 dW_k 5.15e-01 dW_v 5.65e-01 | 3.83 s/step | 22 GB
|
| 97 |
+
step 3100 | loss 0.0273 | obs 0.0245 ret 0.0249 | gn 0.05 | dW_q 5.33e-01 dW_k 5.78e-01 dW_v 6.58e-01 | 3.54 s/step | 22 GB
|
| 98 |
+
step 3150 | loss 0.0269 | obs 0.0245 ret 0.0219 | gn 0.05 | dW_q 4.78e-01 dW_k 5.29e-01 dW_v 5.98e-01 | 3.66 s/step | 22 GB
|
| 99 |
+
step 3200 | loss 0.0269 | obs 0.0222 ret 0.0175 | gn 0.05 | dW_q 5.77e-01 dW_k 6.23e-01 dW_v 7.17e-01 | 4.04 s/step | 22 GB
|
| 100 |
+
step 3250 | loss 0.0263 | obs 0.0238 ret 0.0230 | gn 0.05 | dW_q 6.63e-01 dW_k 6.95e-01 dW_v 8.34e-01 | 3.55 s/step | 22 GB
|
| 101 |
+
step 3300 | loss 0.0244 | obs 0.0218 ret 0.0194 | gn 0.05 | dW_q 4.91e-01 dW_k 5.44e-01 dW_v 6.13e-01 | 3.97 s/step | 22 GB
|
| 102 |
+
step 3350 | loss 0.0258 | obs 0.0219 ret 0.0208 | gn 0.06 | dW_q 6.67e-01 dW_k 7.07e-01 dW_v 8.30e-01 | 3.55 s/step | 22 GB
|
| 103 |
+
step 3400 | loss 0.0243 | obs 0.0213 ret 0.0195 | gn 0.05 | dW_q 5.68e-01 dW_k 6.20e-01 dW_v 6.95e-01 | 3.67 s/step | 22 GB
|
| 104 |
+
step 3450 | loss 0.0238 | obs 0.0218 ret 0.0194 | gn 0.05 | dW_q 4.85e-01 dW_k 5.43e-01 dW_v 6.00e-01 | 3.38 s/step | 22 GB
|
| 105 |
+
step 3500 | loss 0.0232 | obs 0.0182 ret 0.0206 | gn 0.05 | dW_q 6.08e-01 dW_k 6.54e-01 dW_v 7.53e-01 | 3.42 s/step | 22 GB
|
| 106 |
+
step 3550 | loss 0.0226 | obs 0.0188 ret 0.0208 | gn 0.05 | dW_q 6.83e-01 dW_k 7.23e-01 dW_v 8.47e-01 | 4.14 s/step | 22 GB
|
| 107 |
+
step 3600 | loss 0.0233 | obs 0.0192 ret 0.0173 | gn 0.05 | dW_q 4.55e-01 dW_k 5.15e-01 dW_v 5.62e-01 | 3.86 s/step | 22 GB
|
| 108 |
+
step 3650 | loss 0.0236 | obs 0.0211 ret 0.0183 | gn 0.05 | dW_q 5.88e-01 dW_k 6.37e-01 dW_v 7.24e-01 | 4.36 s/step | 22 GB
|
| 109 |
+
step 3700 | loss 0.0229 | obs 0.0196 ret 0.0207 | gn 0.05 | dW_q 5.66e-01 dW_k 6.16e-01 dW_v 6.99e-01 | 4.12 s/step | 22 GB
|
| 110 |
+
step 3750 | loss 0.0236 | obs 0.0223 ret 0.0244 | gn 0.06 | dW_q 5.99e-01 dW_k 6.47e-01 dW_v 7.39e-01 | 3.86 s/step | 22 GB
|
| 111 |
+
step 3800 | loss 0.0212 | obs 0.0202 ret 0.0211 | gn 0.05 | dW_q 5.09e-01 dW_k 5.64e-01 dW_v 6.34e-01 | 3.96 s/step | 22 GB
|
| 112 |
+
step 3850 | loss 0.0217 | obs 0.0185 ret 0.0235 | gn 0.05 | dW_q 7.12e-01 dW_k 7.53e-01 dW_v 8.82e-01 | 4.12 s/step | 22 GB
|
| 113 |
+
step 3900 | loss 0.0202 | obs 0.0184 ret 0.0169 | gn 0.05 | dW_q 6.41e-01 dW_k 6.87e-01 dW_v 7.98e-01 | 3.80 s/step | 22 GB
|
| 114 |
+
step 3950 | loss 0.0203 | obs 0.0191 ret 0.0203 | gn 0.05 | dW_q 5.02e-01 dW_k 5.59e-01 dW_v 6.22e-01 | 3.38 s/step | 22 GB
|
| 115 |
+
step 4000 | loss 0.0196 | obs 0.0180 ret 0.0176 | gn 0.05 | dW_q 5.41e-01 dW_k 5.95e-01 dW_v 6.67e-01 | 3.60 s/step | 22 GB
|
| 116 |
+
step 4050 | loss 0.0194 | obs 0.0181 ret 0.0206 | gn 0.05 | dW_q 6.83e-01 dW_k 7.32e-01 dW_v 8.45e-01 | 3.80 s/step | 22 GB
|
| 117 |
+
step 4100 | loss 0.0188 | obs 0.0191 ret 0.0170 | gn 0.05 | dW_q 6.65e-01 dW_k 7.15e-01 dW_v 8.23e-01 | 3.52 s/step | 22 GB
|
| 118 |
+
step 4150 | loss 0.0175 | obs 0.0196 ret 0.0169 | gn 0.05 | dW_q 4.77e-01 dW_k 5.38e-01 dW_v 5.95e-01 | 3.58 s/step | 22 GB
|
| 119 |
+
step 4200 | loss 0.0188 | obs 0.0186 ret 0.0184 | gn 0.05 | dW_q 5.48e-01 dW_k 6.07e-01 dW_v 6.83e-01 | 4.11 s/step | 22 GB
|
| 120 |
+
step 4250 | loss 0.0178 | obs 0.0171 ret 0.0199 | gn 0.05 | dW_q 5.09e-01 dW_k 5.73e-01 dW_v 6.39e-01 | 3.76 s/step | 22 GB
|
| 121 |
+
step 4300 | loss 0.0195 | obs 0.0208 ret 0.0184 | gn 0.05 | dW_q 5.71e-01 dW_k 6.29e-01 dW_v 7.00e-01 | 3.56 s/step | 22 GB
|
| 122 |
+
step 4350 | loss 0.0184 | obs 0.0192 ret 0.0188 | gn 0.05 | dW_q 5.43e-01 dW_k 6.03e-01 dW_v 6.70e-01 | 3.16 s/step | 22 GB
|
| 123 |
+
step 4400 | loss 0.0174 | obs 0.0172 ret 0.0196 | gn 0.06 | dW_q 4.60e-01 dW_k 5.19e-01 dW_v 5.83e-01 | 3.40 s/step | 22 GB
|
| 124 |
+
step 4450 | loss 0.0182 | obs 0.0182 ret 0.0189 | gn 0.05 | dW_q 7.02e-01 dW_k 7.54e-01 dW_v 8.82e-01 | 3.40 s/step | 22 GB
|
| 125 |
+
step 4500 | loss 0.0183 | obs 0.0180 ret 0.0163 | gn 0.05 | dW_q 6.30e-01 dW_k 6.87e-01 dW_v 7.78e-01 | 3.23 s/step | 22 GB
|
| 126 |
+
step 4550 | loss 0.0173 | obs 0.0177 ret 0.0219 | gn 0.05 | dW_q 6.45e-01 dW_k 7.02e-01 dW_v 8.10e-01 | 3.95 s/step | 22 GB
|
| 127 |
+
step 4600 | loss 0.0180 | obs 0.0167 ret 0.0187 | gn 0.05 | dW_q 4.96e-01 dW_k 5.61e-01 dW_v 6.04e-01 | 3.49 s/step | 22 GB
|
| 128 |
+
step 4650 | loss 0.0181 | obs 0.0189 ret 0.0158 | gn 0.05 | dW_q 6.31e-01 dW_k 6.94e-01 dW_v 7.80e-01 | 3.29 s/step | 22 GB
|
| 129 |
+
step 4700 | loss 0.0175 | obs 0.0172 ret 0.0209 | gn 0.04 | dW_q 7.17e-01 dW_k 7.68e-01 dW_v 8.94e-01 | 3.62 s/step | 22 GB
|
| 130 |
+
step 4750 | loss 0.0172 | obs 0.0170 ret 0.0183 | gn 0.07 | dW_q 6.61e-01 dW_k 7.20e-01 dW_v 8.12e-01 | 3.51 s/step | 22 GB
|
| 131 |
+
step 4800 | loss 0.0180 | obs 0.0195 ret 0.0200 | gn 0.05 | dW_q 6.34e-01 dW_k 6.97e-01 dW_v 7.91e-01 | 3.38 s/step | 22 GB
|
| 132 |
+
step 4850 | loss 0.0180 | obs 0.0180 ret 0.0185 | gn 0.05 | dW_q 6.10e-01 dW_k 6.72e-01 dW_v 7.70e-01 | 3.41 s/step | 22 GB
|
| 133 |
+
step 4900 | loss 0.0171 | obs 0.0188 ret 0.0208 | gn 0.04 | dW_q 7.39e-01 dW_k 7.95e-01 dW_v 9.43e-01 | 3.49 s/step | 22 GB
|
| 134 |
+
step 4950 | loss 0.0184 | obs 0.0198 ret 0.0207 | gn 0.05 | dW_q 6.10e-01 dW_k 6.75e-01 dW_v 7.72e-01 | 3.39 s/step | 22 GB
|
| 135 |
+
step 5000 | loss 0.0169 | obs 0.0168 ret 0.0208 | gn 0.05 | dW_q 4.73e-01 dW_k 5.41e-01 dW_v 5.99e-01 | 3.67 s/step | 22 GB
|
| 136 |
+
step 5050 | loss 0.0168 | obs 0.0171 ret 0.0211 | gn 0.05 | dW_q 6.10e-01 dW_k 6.77e-01 dW_v 7.76e-01 | 3.58 s/step | 22 GB
|
| 137 |
+
step 5100 | loss 0.0167 | obs 0.0161 ret 0.0186 | gn 0.04 | dW_q 7.18e-01 dW_k 7.84e-01 dW_v 9.17e-01 | 3.49 s/step | 22 GB
|
| 138 |
+
step 5150 | loss 0.0172 | obs 0.0170 ret 0.0174 | gn 0.05 | dW_q 5.40e-01 dW_k 6.15e-01 dW_v 6.92e-01 | 3.41 s/step | 22 GB
|
| 139 |
+
step 5200 | loss 0.0165 | obs 0.0162 ret 0.0205 | gn 0.04 | dW_q 6.57e-01 dW_k 7.25e-01 dW_v 8.32e-01 | 3.40 s/step | 22 GB
|
| 140 |
+
step 5250 | loss 0.0181 | obs 0.0201 ret 0.0190 | gn 0.05 | dW_q 5.49e-01 dW_k 6.20e-01 dW_v 7.06e-01 | 3.40 s/step | 22 GB
|
| 141 |
+
step 5300 | loss 0.0169 | obs 0.0173 ret 0.0164 | gn 0.05 | dW_q 6.62e-01 dW_k 7.33e-01 dW_v 8.52e-01 | 3.36 s/step | 22 GB
|
| 142 |
+
step 5350 | loss 0.0174 | obs 0.0203 ret 0.0205 | gn 0.04 | dW_q 6.87e-01 dW_k 7.62e-01 dW_v 8.81e-01 | 3.72 s/step | 22 GB
|
| 143 |
+
step 5400 | loss 0.0164 | obs 0.0170 ret 0.0206 | gn 0.04 | dW_q 6.64e-01 dW_k 7.34e-01 dW_v 8.52e-01 | 3.58 s/step | 22 GB
|
| 144 |
+
step 5450 | loss 0.0160 | obs 0.0159 ret 0.0213 | gn 0.04 | dW_q 6.03e-01 dW_k 6.73e-01 dW_v 7.77e-01 | 3.45 s/step | 22 GB
|
| 145 |
+
step 5500 | loss 0.0165 | obs 0.0178 ret 0.0168 | gn 0.05 | dW_q 6.64e-01 dW_k 7.37e-01 dW_v 8.44e-01 | 3.32 s/step | 22 GB
|
| 146 |
+
step 5550 | loss 0.0164 | obs 0.0173 ret 0.0198 | gn 0.04 | dW_q 6.97e-01 dW_k 7.69e-01 dW_v 9.00e-01 | 3.33 s/step | 22 GB
|
| 147 |
+
step 5600 | loss 0.0172 | obs 0.0169 ret 0.0182 | gn 0.04 | dW_q 5.00e-01 dW_k 5.78e-01 dW_v 6.30e-01 | 3.52 s/step | 22 GB
|
| 148 |
+
step 5650 | loss 0.0170 | obs 0.0187 ret 0.0155 | gn 0.04 | dW_q 7.13e-01 dW_k 7.94e-01 dW_v 9.24e-01 | 3.38 s/step | 22 GB
|
| 149 |
+
step 5700 | loss 0.0165 | obs 0.0157 ret 0.0221 | gn 0.04 | dW_q 6.66e-01 dW_k 7.48e-01 dW_v 8.55e-01 | 3.56 s/step | 22 GB
|
| 150 |
+
step 5750 | loss 0.0164 | obs 0.0179 ret 0.0169 | gn 0.05 | dW_q 5.98e-01 dW_k 6.77e-01 dW_v 7.63e-01 | 3.40 s/step | 22 GB
|
| 151 |
+
step 5800 | loss 0.0150 | obs 0.0157 ret 0.0193 | gn 0.04 | dW_q 5.97e-01 dW_k 6.77e-01 dW_v 7.68e-01 | 3.42 s/step | 22 GB
|
| 152 |
+
step 5850 | loss 0.0158 | obs 0.0173 ret 0.0180 | gn 0.04 | dW_q 6.97e-01 dW_k 7.73e-01 dW_v 9.18e-01 | 3.43 s/step | 22 GB
|
| 153 |
+
step 5900 | loss 0.0167 | obs 0.0172 ret 0.0212 | gn 0.04 | dW_q 6.08e-01 dW_k 6.91e-01 dW_v 7.97e-01 | 3.45 s/step | 22 GB
|
| 154 |
+
step 5950 | loss 0.0167 | obs 0.0159 ret 0.0225 | gn 0.04 | dW_q 6.61e-01 dW_k 7.52e-01 dW_v 8.82e-01 | 3.37 s/step | 22 GB
|
| 155 |
+
step 6000 | loss 0.0161 | obs 0.0176 ret 0.0158 | gn 0.04 | dW_q 6.29e-01 dW_k 7.11e-01 dW_v 8.24e-01 | 3.61 s/step | 22 GB
|
| 156 |
+
step 6050 | loss 0.0155 | obs 0.0154 ret 0.0170 | gn 0.04 | dW_q 5.70e-01 dW_k 6.62e-01 dW_v 7.53e-01 | 3.43 s/step | 22 GB
|
| 157 |
+
step 6100 | loss 0.0154 | obs 0.0159 ret 0.0207 | gn 0.04 | dW_q 5.17e-01 dW_k 6.07e-01 dW_v 6.87e-01 | 3.41 s/step | 22 GB
|
| 158 |
+
step 6150 | loss 0.0164 | obs 0.0182 ret 0.0159 | gn 0.05 | dW_q 5.09e-01 dW_k 5.96e-01 dW_v 6.64e-01 | 3.38 s/step | 22 GB
|
| 159 |
+
step 6200 | loss 0.0163 | obs 0.0142 ret 0.0255 | gn 0.04 | dW_q 7.14e-01 dW_k 8.10e-01 dW_v 9.51e-01 | 3.33 s/step | 22 GB
|
| 160 |
+
step 6250 | loss 0.0170 | obs 0.0173 ret 0.0247 | gn 0.04 | dW_q 4.71e-01 dW_k 5.54e-01 dW_v 6.26e-01 | 3.27 s/step | 22 GB
|
| 161 |
+
step 6300 | loss 0.0157 | obs 0.0189 ret 0.0164 | gn 0.04 | dW_q 6.02e-01 dW_k 6.99e-01 dW_v 8.09e-01 | 3.34 s/step | 22 GB
|
| 162 |
+
step 6350 | loss 0.0162 | obs 0.0155 ret 0.0216 | gn 0.04 | dW_q 6.97e-01 dW_k 7.99e-01 dW_v 9.51e-01 | 3.41 s/step | 22 GB
|
| 163 |
+
step 6400 | loss 0.0146 | obs 0.0164 ret 0.0162 | gn 0.04 | dW_q 4.70e-01 dW_k 5.51e-01 dW_v 6.25e-01 | 3.49 s/step | 22 GB
|
| 164 |
+
step 6450 | loss 0.0154 | obs 0.0172 ret 0.0170 | gn 0.04 | dW_q 5.57e-01 dW_k 6.58e-01 dW_v 7.58e-01 | 3.40 s/step | 22 GB
|
| 165 |
+
step 6500 | loss 0.0162 | obs 0.0196 ret 0.0193 | gn 0.04 | dW_q 5.86e-01 dW_k 6.88e-01 dW_v 7.80e-01 | 3.65 s/step | 22 GB
|
| 166 |
+
step 6550 | loss 0.0147 | obs 0.0151 ret 0.0202 | gn 0.04 | dW_q 6.84e-01 dW_k 7.89e-01 dW_v 9.25e-01 | 3.69 s/step | 22 GB
|
| 167 |
+
step 6600 | loss 0.0163 | obs 0.0187 ret 0.0141 | gn 0.05 | dW_q 5.76e-01 dW_k 6.72e-01 dW_v 7.85e-01 | 3.21 s/step | 22 GB
|
| 168 |
+
run rk_qkv_norm | MemConfig(proj='qkv', ffn=False, layers=(0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11), fast_heads=4, blocks=4, ffn_frac=0.25, ffn_blocks=8, loss='delta', key='norm', lr=0.5, renorm=True) | max_ctx 0 | new params 11.99M | fast weights / sample: writable 2.99M, state 11.94M (46 MB fp32) | start step 6500
|
| 169 |
+
wandb: Currently logged in as: xsourse to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
|
| 170 |
+
wandb: creating run
|
| 171 |
+
wandb: Tracking run with wandb version 0.22.0
|
| 172 |
+
wandb: Run data is saved locally in /mnt/localssd/lehany/wv/E1_rubik_20261104031751/rk_qkv_norm/wandb/run-20261105_174950-816i4yzf
|
| 173 |
+
wandb: Run `wandb offline` to turn off syncing.
|
| 174 |
+
wandb: Resuming run rk_qkv_norm
|
| 175 |
+
wandb: ⭐️ View project at https://wandb.ai/xsourse/world-auto-visual
|
| 176 |
+
wandb: 🚀 View run at https://wandb.ai/xsourse/world-auto-visual/runs/816i4yzf
|
| 177 |
+
step 6550 | loss 0.0151 | obs 0.0149 ret 0.0183 | gn 0.04 | dW_q 6.32e-01 dW_k 7.31e-01 dW_v 8.39e-01 | 3.97 s/step | 22 GB
|
| 178 |
+
wandb: WARNING Tried to log to step 6550 that is less than the current step 6551. Steps must be monotonically increasing, so this data will be ignored. See https://wandb.me/define-metric to log data out of order.
|
| 179 |
+
step 6600 | loss 0.0155 | obs 0.0174 ret 0.0143 | gn 0.04 | dW_q 4.70e-01 dW_k 5.62e-01 dW_v 6.40e-01 | 3.50 s/step | 22 GB
|
| 180 |
+
step 6650 | loss 0.0149 | obs 0.0162 ret 0.0174 | gn 0.04 | dW_q 4.95e-01 dW_k 5.91e-01 dW_v 6.64e-01 | 3.61 s/step | 22 GB
|
| 181 |
+
step 6700 | loss 0.0157 | obs 0.0175 ret 0.0195 | gn 0.04 | dW_q 6.27e-01 dW_k 7.42e-01 dW_v 8.67e-01 | 3.42 s/step | 22 GB
|
| 182 |
+
step 6750 | loss 0.0148 | obs 0.0160 ret 0.0164 | gn 0.04 | dW_q 5.17e-01 dW_k 6.21e-01 dW_v 7.21e-01 | 3.61 s/step | 22 GB
|
| 183 |
+
step 6800 | loss 0.0146 | obs 0.0158 ret 0.0170 | gn 0.04 | dW_q 6.05e-01 dW_k 7.10e-01 dW_v 8.27e-01 | 3.83 s/step | 22 GB
|
| 184 |
+
step 6850 | loss 0.0153 | obs 0.0176 ret 0.0198 | gn 0.04 | dW_q 6.14e-01 dW_k 7.29e-01 dW_v 8.59e-01 | 3.48 s/step | 22 GB
|
| 185 |
+
step 6900 | loss 0.0164 | obs 0.0171 ret 0.0153 | gn 0.04 | dW_q 4.89e-01 dW_k 5.91e-01 dW_v 6.74e-01 | 3.45 s/step | 22 GB
|
| 186 |
+
step 6950 | loss 0.0153 | obs 0.0182 ret 0.0166 | gn 0.04 | dW_q 5.02e-01 dW_k 6.04e-01 dW_v 6.91e-01 | 3.49 s/step | 22 GB
|
| 187 |
+
step 7000 | loss 0.0153 | obs 0.0155 ret 0.0157 | gn 0.04 | dW_q 5.53e-01 dW_k 6.67e-01 dW_v 7.65e-01 | 3.36 s/step | 22 GB
|
| 188 |
+
step 7050 | loss 0.0140 | obs 0.0170 ret 0.0147 | gn 0.04 | dW_q 4.59e-01 dW_k 5.65e-01 dW_v 6.45e-01 | 3.47 s/step | 22 GB
|
| 189 |
+
step 7100 | loss 0.0168 | obs 0.0183 ret 0.0161 | gn 0.06 | dW_q 5.24e-01 dW_k 6.40e-01 dW_v 7.45e-01 | 3.60 s/step | 22 GB
|
| 190 |
+
step 7150 | loss 0.0154 | obs 0.0180 ret 0.0193 | gn 0.04 | dW_q 6.03e-01 dW_k 7.18e-01 dW_v 8.32e-01 | 3.54 s/step | 22 GB
|
| 191 |
+
step 7200 | loss 0.0161 | obs 0.0176 ret 0.0204 | gn 0.04 | dW_q 5.65e-01 dW_k 6.83e-01 dW_v 8.05e-01 | 3.42 s/step | 22 GB
|
| 192 |
+
step 7250 | loss 0.0166 | obs 0.0169 ret 0.0174 | gn 0.04 | dW_q 5.47e-01 dW_k 6.63e-01 dW_v 7.64e-01 | 3.40 s/step | 22 GB
|
| 193 |
+
step 7300 | loss 0.0154 | obs 0.0175 ret 0.0169 | gn 0.04 | dW_q 5.61e-01 dW_k 6.81e-01 dW_v 7.97e-01 | 3.45 s/step | 22 GB
|
| 194 |
+
step 7350 | loss 0.0157 | obs 0.0175 ret 0.0192 | gn 0.04 | dW_q 6.38e-01 dW_k 7.69e-01 dW_v 8.98e-01 | 3.66 s/step | 22 GB
|
| 195 |
+
step 7400 | loss 0.0151 | obs 0.0164 ret 0.0152 | gn 0.04 | dW_q 6.21e-01 dW_k 7.41e-01 dW_v 8.72e-01 | 3.46 s/step | 22 GB
|
| 196 |
+
step 7450 | loss 0.0164 | obs 0.0172 ret 0.0197 | gn 0.04 | dW_q 5.92e-01 dW_k 7.27e-01 dW_v 8.44e-01 | 3.52 s/step | 22 GB
|
| 197 |
+
step 7500 | loss 0.0149 | obs 0.0160 ret 0.0163 | gn 0.04 | dW_q 5.41e-01 dW_k 6.69e-01 dW_v 7.70e-01 | 3.33 s/step | 22 GB
|
| 198 |
+
step 7550 | loss 0.0157 | obs 0.0181 ret 0.0194 | gn 0.04 | dW_q 6.74e-01 dW_k 8.09e-01 dW_v 9.53e-01 | 3.79 s/step | 22 GB
|
| 199 |
+
step 7600 | loss 0.0147 | obs 0.0170 ret 0.0176 | gn 0.04 | dW_q 5.79e-01 dW_k 7.02e-01 dW_v 7.90e-01 | 3.46 s/step | 22 GB
|
| 200 |
+
step 7650 | loss 0.0149 | obs 0.0173 ret 0.0174 | gn 0.04 | dW_q 5.80e-01 dW_k 7.06e-01 dW_v 8.17e-01 | 3.32 s/step | 22 GB
|
| 201 |
+
step 7700 | loss 0.0141 | obs 0.0141 ret 0.0160 | gn 0.03 | dW_q 6.51e-01 dW_k 7.79e-01 dW_v 9.11e-01 | 3.59 s/step | 22 GB
|
| 202 |
+
step 7750 | loss 0.0150 | obs 0.0138 ret 0.0198 | gn 0.04 | dW_q 6.28e-01 dW_k 7.63e-01 dW_v 8.96e-01 | 3.55 s/step | 22 GB
|
| 203 |
+
step 7800 | loss 0.0150 | obs 0.0154 ret 0.0196 | gn 0.04 | dW_q 4.66e-01 dW_k 5.72e-01 dW_v 6.55e-01 | 3.39 s/step | 22 GB
|
| 204 |
+
step 7850 | loss 0.0158 | obs 0.0182 ret 0.0174 | gn 0.05 | dW_q 5.47e-01 dW_k 6.60e-01 dW_v 7.61e-01 | 3.45 s/step | 22 GB
|
| 205 |
+
step 7900 | loss 0.0147 | obs 0.0165 ret 0.0132 | gn 0.04 | dW_q 4.74e-01 dW_k 5.67e-01 dW_v 6.56e-01 | 3.10 s/step | 22 GB
|
| 206 |
+
step 7950 | loss 0.0147 | obs 0.0165 ret 0.0153 | gn 0.04 | dW_q 5.27e-01 dW_k 6.37e-01 dW_v 7.33e-01 | 3.52 s/step | 22 GB
|
| 207 |
+
step 8000 | loss 0.0137 | obs 0.0146 ret 0.0155 | gn 0.04 | dW_q 5.47e-01 dW_k 6.72e-01 dW_v 7.65e-01 | 3.46 s/step | 22 GB
|
| 208 |
+
step 8050 | loss 0.0167 | obs 0.0171 ret 0.0183 | gn 0.05 | dW_q 5.53e-01 dW_k 6.52e-01 dW_v 7.41e-01 | 3.55 s/step | 22 GB
|
| 209 |
+
step 8100 | loss 0.0144 | obs 0.0159 ret 0.0166 | gn 0.04 | dW_q 5.56e-01 dW_k 6.75e-01 dW_v 7.70e-01 | 3.43 s/step | 22 GB
|
| 210 |
+
step 8150 | loss 0.0142 | obs 0.0153 ret 0.0183 | gn 0.04 | dW_q 5.93e-01 dW_k 7.19e-01 dW_v 8.43e-01 | 3.42 s/step | 22 GB
|
| 211 |
+
step 8200 | loss 0.0145 | obs 0.0160 ret 0.0154 | gn 0.04 | dW_q 6.19e-01 dW_k 7.39e-01 dW_v 8.75e-01 | 3.37 s/step | 22 GB
|
| 212 |
+
step 8250 | loss 0.0146 | obs 0.0152 ret 0.0192 | gn 0.04 | dW_q 4.74e-01 dW_k 5.60e-01 dW_v 6.61e-01 | 3.37 s/step | 22 GB
|
| 213 |
+
step 8300 | loss 0.0165 | obs 0.0181 ret 0.0178 | gn 0.04 | dW_q 6.82e-01 dW_k 8.02e-01 dW_v 9.51e-01 | 3.65 s/step | 22 GB
|
| 214 |
+
step 8350 | loss 0.0139 | obs 0.0172 ret 0.0142 | gn 0.04 | dW_q 6.76e-01 dW_k 7.96e-01 dW_v 9.54e-01 | 3.45 s/step | 22 GB
|
| 215 |
+
step 8400 | loss 0.0143 | obs 0.0164 ret 0.0206 | gn 0.04 | dW_q 5.56e-01 dW_k 6.60e-01 dW_v 7.66e-01 | 3.30 s/step | 22 GB
|
| 216 |
+
step 8450 | loss 0.0161 | obs 0.0184 ret 0.0143 | gn 0.04 | dW_q 6.77e-01 dW_k 8.10e-01 dW_v 9.82e-01 | 3.40 s/step | 22 GB
|
| 217 |
+
step 8500 | loss 0.0158 | obs 0.0180 ret 0.0188 | gn 0.04 | dW_q 4.97e-01 dW_k 5.99e-01 dW_v 6.88e-01 | 3.67 s/step | 22 GB
|
| 218 |
+
step 8550 | loss 0.0137 | obs 0.0158 ret 0.0144 | gn 0.04 | dW_q 4.99e-01 dW_k 5.95e-01 dW_v 6.84e-01 | 3.55 s/step | 22 GB
|
| 219 |
+
step 8600 | loss 0.0153 | obs 0.0192 ret 0.0185 | gn 0.04 | dW_q 4.66e-01 dW_k 5.54e-01 dW_v 6.25e-01 | 3.57 s/step | 22 GB
|
| 220 |
+
step 8650 | loss 0.0147 | obs 0.0175 ret 0.0150 | gn 0.05 | dW_q 6.03e-01 dW_k 7.45e-01 dW_v 8.80e-01 | 3.48 s/step | 22 GB
|
| 221 |
+
step 8700 | loss 0.0152 | obs 0.0178 ret 0.0173 | gn 0.04 | dW_q 5.33e-01 dW_k 6.57e-01 dW_v 7.43e-01 | 3.59 s/step | 22 GB
|
| 222 |
+
step 8750 | loss 0.0146 | obs 0.0165 ret 0.0212 | gn 0.04 | dW_q 5.42e-01 dW_k 6.56e-01 dW_v 7.62e-01 | 3.24 s/step | 22 GB
|
| 223 |
+
step 8800 | loss 0.0144 | obs 0.0164 ret 0.0143 | gn 0.04 | dW_q 6.65e-01 dW_k 8.09e-01 dW_v 9.60e-01 | 3.62 s/step | 22 GB
|
| 224 |
+
step 8850 | loss 0.0149 | obs 0.0159 ret 0.0176 | gn 0.04 | dW_q 4.95e-01 dW_k 5.95e-01 dW_v 6.79e-01 | 3.74 s/step | 22 GB
|
| 225 |
+
step 8900 | loss 0.0138 | obs 0.0164 ret 0.0126 | gn 0.04 | dW_q 5.84e-01 dW_k 7.23e-01 dW_v 8.29e-01 | 3.42 s/step | 22 GB
|
| 226 |
+
step 8950 | loss 0.0139 | obs 0.0168 ret 0.0146 | gn 0.04 | dW_q 5.86e-01 dW_k 7.00e-01 dW_v 8.19e-01 | 3.44 s/step | 22 GB
|
| 227 |
+
step 9000 | loss 0.0139 | obs 0.0147 ret 0.0198 | gn 0.04 | dW_q 5.25e-01 dW_k 6.30e-01 dW_v 7.17e-01 | 3.40 s/step | 22 GB
|
| 228 |
+
step 9050 | loss 0.0135 | obs 0.0152 ret 0.0164 | gn 0.04 | dW_q 6.24e-01 dW_k 7.51e-01 dW_v 8.84e-01 | 3.55 s/step | 22 GB
|
| 229 |
+
step 9100 | loss 0.0144 | obs 0.0155 ret 0.0157 | gn 0.04 | dW_q 6.45e-01 dW_k 7.83e-01 dW_v 9.05e-01 | 3.50 s/step | 22 GB
|
| 230 |
+
step 9150 | loss 0.0144 | obs 0.0158 ret 0.0150 | gn 0.04 | dW_q 6.17e-01 dW_k 7.55e-01 dW_v 8.86e-01 | 3.53 s/step | 22 GB
|
| 231 |
+
step 9200 | loss 0.0143 | obs 0.0178 ret 0.0171 | gn 0.04 | dW_q 5.11e-01 dW_k 6.24e-01 dW_v 7.31e-01 | 3.48 s/step | 22 GB
|
| 232 |
+
step 9250 | loss 0.0168 | obs 0.0190 ret 0.0157 | gn 0.04 | dW_q 5.86e-01 dW_k 7.21e-01 dW_v 8.39e-01 | 3.87 s/step | 22 GB
|
| 233 |
+
step 9300 | loss 0.0144 | obs 0.0140 ret 0.0175 | gn 0.04 | dW_q 5.80e-01 dW_k 7.20e-01 dW_v 8.44e-01 | 3.40 s/step | 22 GB
|
| 234 |
+
step 9350 | loss 0.0150 | obs 0.0172 ret 0.0175 | gn 0.04 | dW_q 5.40e-01 dW_k 6.66e-01 dW_v 7.63e-01 | 3.58 s/step | 22 GB
|
| 235 |
+
step 9400 | loss 0.0137 | obs 0.0159 ret 0.0123 | gn 0.04 | dW_q 5.23e-01 dW_k 6.36e-01 dW_v 7.41e-01 | 3.42 s/step | 22 GB
|
| 236 |
+
step 9450 | loss 0.0140 | obs 0.0172 ret 0.0116 | gn 0.04 | dW_q 5.87e-01 dW_k 7.23e-01 dW_v 8.28e-01 | 3.42 s/step | 22 GB
|
| 237 |
+
step 9500 | loss 0.0139 | obs 0.0169 ret 0.0159 | gn 0.04 | dW_q 6.30e-01 dW_k 7.80e-01 dW_v 9.23e-01 | 3.26 s/step | 22 GB
|
| 238 |
+
step 9550 | loss 0.0140 | obs 0.0170 ret 0.0143 | gn 0.04 | dW_q 4.60e-01 dW_k 5.66e-01 dW_v 6.62e-01 | 3.50 s/step | 22 GB
|
| 239 |
+
step 9600 | loss 0.0134 | obs 0.0158 ret 0.0133 | gn 0.04 | dW_q 5.17e-01 dW_k 6.32e-01 dW_v 7.27e-01 | 3.48 s/step | 22 GB
|
| 240 |
+
step 9650 | loss 0.0132 | obs 0.0156 ret 0.0160 | gn 0.04 | dW_q 5.57e-01 dW_k 7.06e-01 dW_v 8.17e-01 | 3.67 s/step | 22 GB
|
| 241 |
+
step 9700 | loss 0.0139 | obs 0.0168 ret 0.0147 | gn 0.04 | dW_q 6.43e-01 dW_k 8.06e-01 dW_v 9.40e-01 | 3.79 s/step | 22 GB
|
| 242 |
+
step 9750 | loss 0.0137 | obs 0.0151 ret 0.0134 | gn 0.04 | dW_q 6.35e-01 dW_k 7.98e-01 dW_v 9.31e-01 | 3.46 s/step | 22 GB
|
| 243 |
+
step 9800 | loss 0.0147 | obs 0.0168 ret 0.0166 | gn 0.04 | dW_q 6.26e-01 dW_k 7.87e-01 dW_v 9.33e-01 | 3.39 s/step | 22 GB
|
| 244 |
+
step 9850 | loss 0.0138 | obs 0.0188 ret 0.0173 | gn 0.04 | dW_q 5.79e-01 dW_k 7.22e-01 dW_v 8.44e-01 | 3.62 s/step | 22 GB
|
| 245 |
+
step 9900 | loss 0.0140 | obs 0.0156 ret 0.0165 | gn 0.06 | dW_q 4.96e-01 dW_k 6.08e-01 dW_v 7.07e-01 | 3.24 s/step | 22 GB
|
| 246 |
+
step 9950 | loss 0.0145 | obs 0.0164 ret 0.0196 | gn 0.04 | dW_q 6.40e-01 dW_k 7.95e-01 dW_v 9.76e-01 | 3.35 s/step | 22 GB
|
| 247 |
+
step 10000 | loss 0.0145 | obs 0.0176 ret 0.0128 | gn 0.04 | dW_q 6.43e-01 dW_k 7.91e-01 dW_v 9.67e-01 | 3.41 s/step | 22 GB
|
| 248 |
+
step 10050 | loss 0.0140 | obs 0.0170 ret 0.0151 | gn 0.05 | dW_q 5.58e-01 dW_k 6.79e-01 dW_v 8.02e-01 | 3.49 s/step | 22 GB
|
| 249 |
+
step 10100 | loss 0.0149 | obs 0.0183 ret 0.0159 | gn 0.04 | dW_q 4.72e-01 dW_k 5.75e-01 dW_v 6.72e-01 | 3.39 s/step | 22 GB
|
| 250 |
+
step 10150 | loss 0.0146 | obs 0.0185 ret 0.0162 | gn 0.04 | dW_q 4.48e-01 dW_k 5.41e-01 dW_v 6.38e-01 | 3.46 s/step | 22 GB
|
| 251 |
+
step 10200 | loss 0.0147 | obs 0.0183 ret 0.0136 | gn 0.04 | dW_q 6.12e-01 dW_k 7.46e-01 dW_v 9.03e-01 | 3.34 s/step | 22 GB
|
| 252 |
+
step 10250 | loss 0.0136 | obs 0.0160 ret 0.0119 | gn 0.04 | dW_q 5.66e-01 dW_k 6.82e-01 dW_v 8.10e-01 | 3.64 s/step | 22 GB
|
| 253 |
+
step 10300 | loss 0.0146 | obs 0.0141 ret 0.0152 | gn 0.04 | dW_q 4.96e-01 dW_k 6.09e-01 dW_v 7.00e-01 | 3.50 s/step | 22 GB
|
| 254 |
+
step 10350 | loss 0.0148 | obs 0.0189 ret 0.0138 | gn 0.04 | dW_q 6.52e-01 dW_k 8.03e-01 dW_v 9.65e-01 | 3.57 s/step | 22 GB
|
| 255 |
+
step 10400 | loss 0.0144 | obs 0.0166 ret 0.0141 | gn 0.04 | dW_q 5.68e-01 dW_k 6.94e-01 dW_v 8.19e-01 | 3.40 s/step | 22 GB
|
| 256 |
+
step 10450 | loss 0.0157 | obs 0.0174 ret 0.0249 | gn 0.04 | dW_q 5.92e-01 dW_k 7.32e-01 dW_v 8.72e-01 | 3.64 s/step | 22 GB
|
| 257 |
+
step 10500 | loss 0.0141 | obs 0.0177 ret 0.0125 | gn 0.04 | dW_q 5.85e-01 dW_k 7.30e-01 dW_v 8.76e-01 | 3.47 s/step | 22 GB
|
| 258 |
+
step 10550 | loss 0.0137 | obs 0.0186 ret 0.0123 | gn 0.04 | dW_q 6.36e-01 dW_k 7.83e-01 dW_v 9.31e-01 | 3.67 s/step | 22 GB
|
| 259 |
+
step 10600 | loss 0.0142 | obs 0.0158 ret 0.0161 | gn 0.05 | dW_q 5.49e-01 dW_k 6.79e-01 dW_v 7.99e-01 | 3.55 s/step | 22 GB
|
| 260 |
+
step 10650 | loss 0.0143 | obs 0.0146 ret 0.0204 | gn 0.04 | dW_q 5.24e-01 dW_k 6.36e-01 dW_v 7.41e-01 | 3.53 s/step | 22 GB
|
| 261 |
+
step 10700 | loss 0.0138 | obs 0.0173 ret 0.0145 | gn 0.04 | dW_q 6.22e-01 dW_k 7.52e-01 dW_v 9.02e-01 | 3.44 s/step | 22 GB
|
| 262 |
+
step 10750 | loss 0.0155 | obs 0.0181 ret 0.0187 | gn 0.04 | dW_q 6.06e-01 dW_k 7.36e-01 dW_v 8.86e-01 | 3.52 s/step | 22 GB
|
| 263 |
+
step 10800 | loss 0.0142 | obs 0.0159 ret 0.0160 | gn 0.04 | dW_q 6.04e-01 dW_k 7.48e-01 dW_v 8.97e-01 | 3.67 s/step | 22 GB
|
| 264 |
+
step 10850 | loss 0.0142 | obs 0.0149 ret 0.0129 | gn 0.04 | dW_q 6.14e-01 dW_k 7.69e-01 dW_v 9.17e-01 | 3.52 s/step | 22 GB
|
| 265 |
+
step 10900 | loss 0.0128 | obs 0.0142 ret 0.0172 | gn 0.04 | dW_q 5.90e-01 dW_k 7.40e-01 dW_v 8.64e-01 | 3.39 s/step | 22 GB
|
| 266 |
+
step 10950 | loss 0.0135 | obs 0.0137 ret 0.0161 | gn 0.04 | dW_q 5.27e-01 dW_k 6.67e-01 dW_v 7.72e-01 | 3.38 s/step | 22 GB
|
| 267 |
+
step 11000 | loss 0.0138 | obs 0.0143 ret 0.0157 | gn 0.04 | dW_q 5.48e-01 dW_k 6.88e-01 dW_v 8.06e-01 | 3.42 s/step | 22 GB
|
| 268 |
+
step 11050 | loss 0.0141 | obs 0.0149 ret 0.0140 | gn 0.04 | dW_q 5.64e-01 dW_k 7.31e-01 dW_v 8.68e-01 | 3.55 s/step | 22 GB
|
| 269 |
+
step 11100 | loss 0.0148 | obs 0.0156 ret 0.0130 | gn 0.04 | dW_q 6.05e-01 dW_k 7.82e-01 dW_v 9.32e-01 | 3.56 s/step | 22 GB
|
| 270 |
+
step 11150 | loss 0.0145 | obs 0.0174 ret 0.0191 | gn 0.04 | dW_q 5.20e-01 dW_k 6.58e-01 dW_v 7.84e-01 | 3.57 s/step | 22 GB
|
| 271 |
+
step 11200 | loss 0.0147 | obs 0.0174 ret 0.0110 | gn 0.05 | dW_q 5.11e-01 dW_k 6.41e-01 dW_v 7.44e-01 | 3.08 s/step | 22 GB
|
| 272 |
+
step 11250 | loss 0.0135 | obs 0.0151 ret 0.0132 | gn 0.04 | dW_q 6.00e-01 dW_k 7.47e-01 dW_v 8.86e-01 | 3.55 s/step | 22 GB
|
| 273 |
+
step 11300 | loss 0.0133 | obs 0.0179 ret 0.0131 | gn 0.04 | dW_q 5.23e-01 dW_k 6.54e-01 dW_v 7.59e-01 | 3.65 s/step | 22 GB
|
| 274 |
+
step 11350 | loss 0.0137 | obs 0.0153 ret 0.0187 | gn 0.04 | dW_q 5.09e-01 dW_k 6.41e-01 dW_v 7.44e-01 | 3.41 s/step | 22 GB
|
| 275 |
+
step 11400 | loss 0.0144 | obs 0.0156 ret 0.0163 | gn 0.04 | dW_q 5.56e-01 dW_k 6.85e-01 dW_v 7.83e-01 | 3.58 s/step | 22 GB
|
| 276 |
+
step 11450 | loss 0.0138 | obs 0.0169 ret 0.0157 | gn 0.04 | dW_q 4.52e-01 dW_k 5.56e-01 dW_v 6.60e-01 | 3.35 s/step | 22 GB
|
| 277 |
+
step 11500 | loss 0.0144 | obs 0.0175 ret 0.0117 | gn 0.04 | dW_q 5.15e-01 dW_k 6.32e-01 dW_v 7.39e-01 | 3.43 s/step | 22 GB
|
| 278 |
+
step 11550 | loss 0.0136 | obs 0.0150 ret 0.0203 | gn 0.04 | dW_q 6.15e-01 dW_k 7.65e-01 dW_v 9.07e-01 | 3.32 s/step | 22 GB
|
| 279 |
+
step 11600 | loss 0.0131 | obs 0.0152 ret 0.0173 | gn 0.04 | dW_q 5.63e-01 dW_k 6.99e-01 dW_v 8.23e-01 | 3.42 s/step | 22 GB
|
| 280 |
+
step 11650 | loss 0.0137 | obs 0.0149 ret 0.0122 | gn 0.04 | dW_q 5.44e-01 dW_k 6.70e-01 dW_v 7.83e-01 | 3.57 s/step | 22 GB
|
| 281 |
+
step 11700 | loss 0.0139 | obs 0.0160 ret 0.0102 | gn 0.04 | dW_q 5.59e-01 dW_k 6.89e-01 dW_v 8.16e-01 | 3.28 s/step | 22 GB
|
| 282 |
+
step 11750 | loss 0.0151 | obs 0.0183 ret 0.0154 | gn 0.04 | dW_q 5.01e-01 dW_k 6.27e-01 dW_v 7.27e-01 | 3.55 s/step | 22 GB
|
| 283 |
+
step 11800 | loss 0.0142 | obs 0.0167 ret 0.0137 | gn 0.04 | dW_q 6.50e-01 dW_k 8.09e-01 dW_v 9.66e-01 | 3.55 s/step | 22 GB
|
| 284 |
+
step 11850 | loss 0.0142 | obs 0.0167 ret 0.0179 | gn 0.04 | dW_q 5.83e-01 dW_k 7.29e-01 dW_v 8.53e-01 | 3.45 s/step | 22 GB
|
| 285 |
+
step 11900 | loss 0.0131 | obs 0.0147 ret 0.0124 | gn 0.04 | dW_q 5.05e-01 dW_k 6.31e-01 dW_v 7.32e-01 | 3.33 s/step | 22 GB
|
| 286 |
+
step 11950 | loss 0.0136 | obs 0.0153 ret 0.0136 | gn 0.04 | dW_q 4.73e-01 dW_k 5.88e-01 dW_v 6.97e-01 | 3.36 s/step | 22 GB
|
| 287 |
+
step 12000 | loss 0.0131 | obs 0.0149 ret 0.0145 | gn 0.04 | dW_q 4.69e-01 dW_k 5.86e-01 dW_v 6.90e-01 | 3.28 s/step | 22 GB
|
| 288 |
+
[1;34mwandb[0m:
|
| 289 |
+
[1;34mwandb[0m: 🚀 View run [33mrk_qkv_norm[0m at: [34m[0m
|
| 290 |
+
[1;34mwandb[0m: Find logs at: [1;35m../../../../../../wv/E1_rubik_20261104031751/rk_qkv_norm/wandb/run-20261105_174950-816i4yzf/logs[0m
|
rubik_r1_20261106011120/logs/rk_v.log
ADDED
|
@@ -0,0 +1,274 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
run rk_v | MemConfig(proj='v', ffn=False, layers=(0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11), fast_heads=4, blocks=4, ffn_frac=0.25, ffn_blocks=8, loss='delta', key='raw', lr=0.001, renorm=True) | max_ctx 0 | new params 4.00M | fast weights / sample: writable 1.00M, state 3.98M (15 MB fp32) | start step 0
|
| 2 |
+
wandb: Currently logged in as: xsourse to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
|
| 3 |
+
wandb: creating run
|
| 4 |
+
wandb: Tracking run with wandb version 0.22.0
|
| 5 |
+
wandb: Run data is saved locally in /mnt/localssd/lehany/wv/E1_rubik_20261104031751/rk_v/wandb/run-20261104_071615-w5mv6kf0
|
| 6 |
+
wandb: Run `wandb offline` to turn off syncing.
|
| 7 |
+
wandb: Syncing run rk_v
|
| 8 |
+
wandb: ⭐️ View project at https://wandb.ai/xsourse/world-auto-visual
|
| 9 |
+
wandb: 🚀 View run at https://wandb.ai/xsourse/world-auto-visual/runs/w5mv6kf0
|
| 10 |
+
step 50 | loss 0.0619 | obs 0.0596 ret 0.0600 | gn 0.12 | dW_v 2.17e-04 | 2.75 s/step | 17 GB
|
| 11 |
+
step 100 | loss 0.0564 | obs 0.0553 ret 0.0586 | gn 0.04 | dW_v 1.99e-03 | 2.40 s/step | 17 GB
|
| 12 |
+
step 150 | loss 0.0542 | obs 0.0547 ret 0.0536 | gn 0.04 | dW_v 4.70e-03 | 2.49 s/step | 17 GB
|
| 13 |
+
step 200 | loss 0.0552 | obs 0.0555 ret 0.0537 | gn 0.04 | dW_v 7.49e-03 | 2.53 s/step | 17 GB
|
| 14 |
+
step 250 | loss 0.0534 | obs 0.0518 ret 0.0530 | gn 0.04 | dW_v 1.83e-02 | 2.41 s/step | 17 GB
|
| 15 |
+
step 300 | loss 0.0524 | obs 0.0525 ret 0.0533 | gn 0.05 | dW_v 2.39e-02 | 2.58 s/step | 17 GB
|
| 16 |
+
step 350 | loss 0.0537 | obs 0.0516 ret 0.0561 | gn 0.04 | dW_v 2.87e-02 | 2.60 s/step | 17 GB
|
| 17 |
+
step 400 | loss 0.0548 | obs 0.0528 ret 0.0532 | gn 0.04 | dW_v 3.04e-02 | 2.55 s/step | 17 GB
|
| 18 |
+
step 450 | loss 0.0541 | obs 0.0543 ret 0.0513 | gn 0.05 | dW_v 2.94e-02 | 2.50 s/step | 17 GB
|
| 19 |
+
step 500 | loss 0.0532 | obs 0.0514 ret 0.0568 | gn 0.05 | dW_v 4.05e-02 | 2.45 s/step | 17 GB
|
| 20 |
+
step 550 | loss 0.0526 | obs 0.0509 ret 0.0508 | gn 0.05 | dW_v 4.62e-02 | 2.49 s/step | 17 GB
|
| 21 |
+
step 600 | loss 0.0542 | obs 0.0513 ret 0.0532 | gn 0.07 | dW_v 5.38e-02 | 2.54 s/step | 17 GB
|
| 22 |
+
step 650 | loss 0.0521 | obs 0.0474 ret 0.0505 | gn 0.05 | dW_v 5.94e-02 | 2.39 s/step | 17 GB
|
| 23 |
+
step 700 | loss 0.0535 | obs 0.0499 ret 0.0545 | gn 0.05 | dW_v 7.32e-02 | 2.51 s/step | 17 GB
|
| 24 |
+
step 750 | loss 0.0515 | obs 0.0479 ret 0.0548 | gn 0.05 | dW_v 9.46e-02 | 2.64 s/step | 17 GB
|
| 25 |
+
step 800 | loss 0.0501 | obs 0.0424 ret 0.0539 | gn 0.06 | dW_v 1.08e-01 | 2.62 s/step | 17 GB
|
| 26 |
+
step 850 | loss 0.0472 | obs 0.0408 ret 0.0542 | gn 0.08 | dW_v 1.13e-01 | 2.57 s/step | 17 GB
|
| 27 |
+
step 900 | loss 0.0422 | obs 0.0369 ret 0.0565 | gn 0.10 | dW_v 1.39e-01 | 2.49 s/step | 17 GB
|
| 28 |
+
step 950 | loss 0.0378 | obs 0.0327 ret 0.0476 | gn 0.08 | dW_v 1.19e-01 | 2.36 s/step | 17 GB
|
| 29 |
+
step 1000 | loss 0.0320 | obs 0.0309 ret 0.0378 | gn 0.08 | dW_v 1.30e-01 | 2.45 s/step | 17 GB
|
| 30 |
+
step 1050 | loss 0.0272 | obs 0.0246 ret 0.0303 | gn 0.11 | dW_v 1.23e-01 | 2.54 s/step | 17 GB
|
| 31 |
+
step 1100 | loss 0.0245 | obs 0.0234 ret 0.0292 | gn 0.06 | dW_v 1.36e-01 | 2.60 s/step | 17 GB
|
| 32 |
+
step 1150 | loss 0.0212 | obs 0.0194 ret 0.0225 | gn 0.05 | dW_v 1.21e-01 | 2.46 s/step | 17 GB
|
| 33 |
+
step 1200 | loss 0.0194 | obs 0.0192 ret 0.0203 | gn 0.04 | dW_v 1.24e-01 | 2.59 s/step | 17 GB
|
| 34 |
+
step 1250 | loss 0.0203 | obs 0.0205 ret 0.0206 | gn 0.19 | dW_v 1.21e-01 | 2.43 s/step | 17 GB
|
| 35 |
+
step 1300 | loss 0.0195 | obs 0.0186 ret 0.0207 | gn 0.13 | dW_v 1.20e-01 | 2.53 s/step | 17 GB
|
| 36 |
+
step 1350 | loss 0.0201 | obs 0.0187 ret 0.0241 | gn 0.14 | dW_v 1.10e-01 | 2.44 s/step | 17 GB
|
| 37 |
+
step 1400 | loss 0.0175 | obs 0.0177 ret 0.0202 | gn 0.06 | dW_v 1.09e-01 | 2.59 s/step | 17 GB
|
| 38 |
+
step 1450 | loss 0.0168 | obs 0.0185 ret 0.0184 | gn 0.10 | dW_v 1.09e-01 | 2.48 s/step | 17 GB
|
| 39 |
+
step 1500 | loss 0.0175 | obs 0.0174 ret 0.0214 | gn 0.09 | dW_v 1.21e-01 | 2.54 s/step | 17 GB
|
| 40 |
+
step 1550 | loss 0.0183 | obs 0.0194 ret 0.0195 | gn 0.07 | dW_v 1.23e-01 | 2.40 s/step | 17 GB
|
| 41 |
+
step 1600 | loss 0.0163 | obs 0.0173 ret 0.0188 | gn 0.04 | dW_v 1.19e-01 | 2.34 s/step | 17 GB
|
| 42 |
+
step 1650 | loss 0.0154 | obs 0.0157 ret 0.0144 | gn 0.04 | dW_v 1.21e-01 | 2.42 s/step | 17 GB
|
| 43 |
+
step 1700 | loss 0.0172 | obs 0.0168 ret 0.0203 | gn 0.04 | dW_v 1.13e-01 | 2.45 s/step | 17 GB
|
| 44 |
+
step 1750 | loss 0.0171 | obs 0.0154 ret 0.0192 | gn 0.04 | dW_v 1.15e-01 | 2.51 s/step | 17 GB
|
| 45 |
+
step 1800 | loss 0.0160 | obs 0.0182 ret 0.0181 | gn 0.04 | dW_v 1.13e-01 | 2.41 s/step | 17 GB
|
| 46 |
+
step 1850 | loss 0.0154 | obs 0.0142 ret 0.0188 | gn 0.04 | dW_v 1.14e-01 | 2.43 s/step | 17 GB
|
| 47 |
+
step 1900 | loss 0.0175 | obs 0.0172 ret 0.0178 | gn 0.04 | dW_v 1.12e-01 | 2.55 s/step | 17 GB
|
| 48 |
+
step 1950 | loss 0.0167 | obs 0.0182 ret 0.0158 | gn 0.04 | dW_v 1.12e-01 | 2.45 s/step | 17 GB
|
| 49 |
+
step 2000 | loss 0.0165 | obs 0.0162 ret 0.0191 | gn 0.04 | dW_v 1.10e-01 | 2.29 s/step | 17 GB
|
| 50 |
+
step 2050 | loss 0.0163 | obs 0.0155 ret 0.0161 | gn 0.04 | dW_v 1.16e-01 | 2.63 s/step | 17 GB
|
| 51 |
+
run rk_v | MemConfig(proj='v', ffn=False, layers=(0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11), fast_heads=4, blocks=4, ffn_frac=0.25, ffn_blocks=8, loss='delta', key='raw', lr=0.001, renorm=True) | max_ctx 0 | new params 4.00M | fast weights / sample: writable 1.00M, state 3.98M (15 MB fp32) | start step 2000
|
| 52 |
+
wandb: Currently logged in as: xsourse to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
|
| 53 |
+
wandb: creating run
|
| 54 |
+
wandb: Tracking run with wandb version 0.22.0
|
| 55 |
+
wandb: Run data is saved locally in /mnt/localssd/lehany/wv/E1_rubik_20261104031751/rk_v/wandb/run-20261105_121205-w5mv6kf0
|
| 56 |
+
wandb: Run `wandb offline` to turn off syncing.
|
| 57 |
+
wandb: Resuming run rk_v
|
| 58 |
+
wandb: ⭐️ View project at https://wandb.ai/xsourse/world-auto-visual
|
| 59 |
+
wandb: 🚀 View run at https://wandb.ai/xsourse/world-auto-visual/runs/w5mv6kf0
|
| 60 |
+
step 2050 | loss 0.0156 | obs 0.0155 ret 0.0139 | gn 0.04 | dW_v 1.17e-01 | 2.58 s/step | 19 GB
|
| 61 |
+
step 2100 | loss 0.0160 | obs 0.0175 ret 0.0152 | gn 0.04 | dW_v 1.11e-01 | 2.43 s/step | 19 GB
|
| 62 |
+
step 2150 | loss 0.0159 | obs 0.0167 ret 0.0204 | gn 0.04 | dW_v 1.14e-01 | 2.45 s/step | 19 GB
|
| 63 |
+
step 2200 | loss 0.0160 | obs 0.0166 ret 0.0147 | gn 0.04 | dW_v 1.21e-01 | 2.63 s/step | 19 GB
|
| 64 |
+
step 2250 | loss 0.0148 | obs 0.0156 ret 0.0128 | gn 0.04 | dW_v 1.20e-01 | 2.56 s/step | 19 GB
|
| 65 |
+
step 2300 | loss 0.0149 | obs 0.0179 ret 0.0158 | gn 0.04 | dW_v 1.22e-01 | 2.49 s/step | 19 GB
|
| 66 |
+
step 2350 | loss 0.0156 | obs 0.0159 ret 0.0169 | gn 0.04 | dW_v 1.23e-01 | 2.51 s/step | 19 GB
|
| 67 |
+
step 2400 | loss 0.0174 | obs 0.0164 ret 0.0252 | gn 0.04 | dW_v 1.20e-01 | 2.58 s/step | 19 GB
|
| 68 |
+
step 2450 | loss 0.0156 | obs 0.0171 ret 0.0165 | gn 0.04 | dW_v 1.22e-01 | 2.42 s/step | 19 GB
|
| 69 |
+
step 2500 | loss 0.0150 | obs 0.0155 ret 0.0195 | gn 0.04 | dW_v 1.19e-01 | 2.44 s/step | 19 GB
|
| 70 |
+
step 2550 | loss 0.0144 | obs 0.0148 ret 0.0145 | gn 0.04 | dW_v 1.31e-01 | 2.58 s/step | 19 GB
|
| 71 |
+
step 2600 | loss 0.0171 | obs 0.0174 ret 0.0202 | gn 0.05 | dW_v 1.24e-01 | 2.44 s/step | 19 GB
|
| 72 |
+
step 2650 | loss 0.0155 | obs 0.0171 ret 0.0157 | gn 0.04 | dW_v 1.29e-01 | 2.48 s/step | 19 GB
|
| 73 |
+
step 2700 | loss 0.0160 | obs 0.0171 ret 0.0178 | gn 0.04 | dW_v 1.35e-01 | 2.34 s/step | 19 GB
|
| 74 |
+
step 2750 | loss 0.0165 | obs 0.0182 ret 0.0170 | gn 0.04 | dW_v 1.36e-01 | 2.50 s/step | 19 GB
|
| 75 |
+
step 2800 | loss 0.0150 | obs 0.0171 ret 0.0156 | gn 0.04 | dW_v 1.41e-01 | 2.58 s/step | 19 GB
|
| 76 |
+
step 2850 | loss 0.0152 | obs 0.0161 ret 0.0177 | gn 0.04 | dW_v 1.38e-01 | 2.74 s/step | 19 GB
|
| 77 |
+
step 2900 | loss 0.0154 | obs 0.0155 ret 0.0175 | gn 0.04 | dW_v 1.36e-01 | 2.38 s/step | 19 GB
|
| 78 |
+
step 2950 | loss 0.0154 | obs 0.0172 ret 0.0173 | gn 0.04 | dW_v 1.45e-01 | 2.31 s/step | 19 GB
|
| 79 |
+
step 3000 | loss 0.0151 | obs 0.0164 ret 0.0143 | gn 0.04 | dW_v 1.39e-01 | 2.39 s/step | 19 GB
|
| 80 |
+
step 3050 | loss 0.0151 | obs 0.0152 ret 0.0184 | gn 0.04 | dW_v 1.33e-01 | 2.51 s/step | 19 GB
|
| 81 |
+
step 3100 | loss 0.0148 | obs 0.0186 ret 0.0178 | gn 0.04 | dW_v 1.37e-01 | 2.42 s/step | 19 GB
|
| 82 |
+
step 3150 | loss 0.0143 | obs 0.0143 ret 0.0165 | gn 0.04 | dW_v 1.44e-01 | 2.33 s/step | 19 GB
|
| 83 |
+
step 3200 | loss 0.0139 | obs 0.0158 ret 0.0141 | gn 0.04 | dW_v 1.50e-01 | 2.25 s/step | 19 GB
|
| 84 |
+
step 3250 | loss 0.0143 | obs 0.0185 ret 0.0153 | gn 0.04 | dW_v 1.39e-01 | 2.54 s/step | 19 GB
|
| 85 |
+
step 3300 | loss 0.0157 | obs 0.0190 ret 0.0178 | gn 0.04 | dW_v 1.38e-01 | 2.42 s/step | 19 GB
|
| 86 |
+
step 3350 | loss 0.0142 | obs 0.0177 ret 0.0152 | gn 0.04 | dW_v 1.39e-01 | 2.45 s/step | 19 GB
|
| 87 |
+
step 3400 | loss 0.0146 | obs 0.0169 ret 0.0148 | gn 0.04 | dW_v 1.45e-01 | 2.37 s/step | 19 GB
|
| 88 |
+
step 3450 | loss 0.0144 | obs 0.0161 ret 0.0200 | gn 0.04 | dW_v 1.43e-01 | 2.43 s/step | 19 GB
|
| 89 |
+
step 3500 | loss 0.0129 | obs 0.0145 ret 0.0122 | gn 0.04 | dW_v 1.52e-01 | 2.55 s/step | 19 GB
|
| 90 |
+
step 3550 | loss 0.0152 | obs 0.0174 ret 0.0148 | gn 0.04 | dW_v 1.46e-01 | 2.67 s/step | 19 GB
|
| 91 |
+
step 3600 | loss 0.0130 | obs 0.0141 ret 0.0152 | gn 0.04 | dW_v 1.46e-01 | 2.45 s/step | 19 GB
|
| 92 |
+
step 3650 | loss 0.0145 | obs 0.0170 ret 0.0162 | gn 0.04 | dW_v 1.39e-01 | 2.57 s/step | 19 GB
|
| 93 |
+
step 3700 | loss 0.0133 | obs 0.0146 ret 0.0157 | gn 0.04 | dW_v 1.46e-01 | 2.57 s/step | 19 GB
|
| 94 |
+
step 3750 | loss 0.0136 | obs 0.0154 ret 0.0110 | gn 0.04 | dW_v 1.50e-01 | 2.37 s/step | 19 GB
|
| 95 |
+
step 3800 | loss 0.0152 | obs 0.0168 ret 0.0149 | gn 0.05 | dW_v 1.42e-01 | 2.18 s/step | 19 GB
|
| 96 |
+
step 3850 | loss 0.0137 | obs 0.0148 ret 0.0146 | gn 0.04 | dW_v 1.49e-01 | 2.51 s/step | 19 GB
|
| 97 |
+
step 3900 | loss 0.0136 | obs 0.0167 ret 0.0119 | gn 0.04 | dW_v 1.45e-01 | 2.38 s/step | 19 GB
|
| 98 |
+
step 3950 | loss 0.0158 | obs 0.0181 ret 0.0221 | gn 0.04 | dW_v 1.46e-01 | 2.43 s/step | 19 GB
|
| 99 |
+
step 4000 | loss 0.0142 | obs 0.0154 ret 0.0120 | gn 0.05 | dW_v 1.44e-01 | 2.46 s/step | 19 GB
|
| 100 |
+
step 4050 | loss 0.0146 | obs 0.0168 ret 0.0124 | gn 0.04 | dW_v 1.49e-01 | 2.66 s/step | 19 GB
|
| 101 |
+
step 4100 | loss 0.0136 | obs 0.0156 ret 0.0171 | gn 0.05 | dW_v 1.40e-01 | 2.62 s/step | 19 GB
|
| 102 |
+
step 4150 | loss 0.0148 | obs 0.0177 ret 0.0138 | gn 0.04 | dW_v 1.43e-01 | 2.52 s/step | 19 GB
|
| 103 |
+
step 4200 | loss 0.0148 | obs 0.0175 ret 0.0150 | gn 0.04 | dW_v 1.45e-01 | 2.64 s/step | 19 GB
|
| 104 |
+
step 4250 | loss 0.0136 | obs 0.0160 ret 0.0178 | gn 0.04 | dW_v 1.50e-01 | 2.52 s/step | 19 GB
|
| 105 |
+
step 4300 | loss 0.0145 | obs 0.0172 ret 0.0160 | gn 0.04 | dW_v 1.49e-01 | 2.39 s/step | 19 GB
|
| 106 |
+
step 4350 | loss 0.0147 | obs 0.0164 ret 0.0139 | gn 0.05 | dW_v 1.47e-01 | 2.71 s/step | 19 GB
|
| 107 |
+
step 4400 | loss 0.0131 | obs 0.0146 ret 0.0131 | gn 0.04 | dW_v 1.45e-01 | 2.45 s/step | 19 GB
|
| 108 |
+
step 4450 | loss 0.0135 | obs 0.0166 ret 0.0123 | gn 0.04 | dW_v 1.44e-01 | 2.46 s/step | 19 GB
|
| 109 |
+
step 4500 | loss 0.0135 | obs 0.0157 ret 0.0163 | gn 0.04 | dW_v 1.51e-01 | 2.50 s/step | 19 GB
|
| 110 |
+
step 4550 | loss 0.0125 | obs 0.0151 ret 0.0135 | gn 0.04 | dW_v 1.50e-01 | 2.35 s/step | 19 GB
|
| 111 |
+
step 4600 | loss 0.0140 | obs 0.0179 ret 0.0159 | gn 0.04 | dW_v 1.49e-01 | 2.35 s/step | 19 GB
|
| 112 |
+
step 4650 | loss 0.0141 | obs 0.0162 ret 0.0149 | gn 0.04 | dW_v 1.45e-01 | 2.67 s/step | 19 GB
|
| 113 |
+
step 4700 | loss 0.0144 | obs 0.0163 ret 0.0166 | gn 0.04 | dW_v 1.48e-01 | 2.61 s/step | 19 GB
|
| 114 |
+
step 4750 | loss 0.0153 | obs 0.0155 ret 0.0189 | gn 0.04 | dW_v 1.44e-01 | 2.61 s/step | 19 GB
|
| 115 |
+
step 4800 | loss 0.0136 | obs 0.0155 ret 0.0126 | gn 0.04 | dW_v 1.55e-01 | 2.43 s/step | 19 GB
|
| 116 |
+
step 4850 | loss 0.0144 | obs 0.0177 ret 0.0163 | gn 0.04 | dW_v 1.52e-01 | 2.49 s/step | 19 GB
|
| 117 |
+
step 4900 | loss 0.0135 | obs 0.0163 ret 0.0157 | gn 0.04 | dW_v 1.52e-01 | 2.42 s/step | 19 GB
|
| 118 |
+
step 4950 | loss 0.0133 | obs 0.0154 ret 0.0129 | gn 0.04 | dW_v 1.55e-01 | 2.74 s/step | 19 GB
|
| 119 |
+
step 5000 | loss 0.0132 | obs 0.0157 ret 0.0127 | gn 0.04 | dW_v 1.52e-01 | 2.42 s/step | 19 GB
|
| 120 |
+
step 5050 | loss 0.0129 | obs 0.0155 ret 0.0155 | gn 0.04 | dW_v 1.48e-01 | 2.50 s/step | 19 GB
|
| 121 |
+
step 5100 | loss 0.0133 | obs 0.0166 ret 0.0127 | gn 0.04 | dW_v 1.46e-01 | 2.46 s/step | 19 GB
|
| 122 |
+
step 5150 | loss 0.0126 | obs 0.0136 ret 0.0114 | gn 0.04 | dW_v 1.55e-01 | 2.30 s/step | 19 GB
|
| 123 |
+
step 5200 | loss 0.0134 | obs 0.0154 ret 0.0130 | gn 0.04 | dW_v 1.59e-01 | 2.50 s/step | 19 GB
|
| 124 |
+
step 5250 | loss 0.0129 | obs 0.0152 ret 0.0156 | gn 0.04 | dW_v 1.60e-01 | 2.44 s/step | 19 GB
|
| 125 |
+
step 5300 | loss 0.0144 | obs 0.0152 ret 0.0202 | gn 0.04 | dW_v 1.39e-01 | 2.46 s/step | 19 GB
|
| 126 |
+
step 5350 | loss 0.0135 | obs 0.0158 ret 0.0160 | gn 0.04 | dW_v 1.57e-01 | 2.51 s/step | 19 GB
|
| 127 |
+
step 5400 | loss 0.0136 | obs 0.0165 ret 0.0127 | gn 0.04 | dW_v 1.54e-01 | 2.50 s/step | 19 GB
|
| 128 |
+
step 5450 | loss 0.0132 | obs 0.0178 ret 0.0118 | gn 0.04 | dW_v 1.51e-01 | 2.39 s/step | 19 GB
|
| 129 |
+
step 5500 | loss 0.0140 | obs 0.0159 ret 0.0133 | gn 0.04 | dW_v 1.49e-01 | 2.43 s/step | 19 GB
|
| 130 |
+
step 5550 | loss 0.0133 | obs 0.0159 ret 0.0109 | gn 0.04 | dW_v 1.59e-01 | 2.44 s/step | 19 GB
|
| 131 |
+
step 5600 | loss 0.0139 | obs 0.0163 ret 0.0166 | gn 0.04 | dW_v 1.41e-01 | 2.49 s/step | 19 GB
|
| 132 |
+
step 5650 | loss 0.0143 | obs 0.0154 ret 0.0147 | gn 0.04 | dW_v 1.57e-01 | 2.49 s/step | 19 GB
|
| 133 |
+
step 5700 | loss 0.0139 | obs 0.0143 ret 0.0168 | gn 0.04 | dW_v 1.57e-01 | 2.42 s/step | 19 GB
|
| 134 |
+
step 5750 | loss 0.0132 | obs 0.0155 ret 0.0146 | gn 0.04 | dW_v 1.44e-01 | 2.55 s/step | 19 GB
|
| 135 |
+
step 5800 | loss 0.0144 | obs 0.0171 ret 0.0136 | gn 0.04 | dW_v 1.61e-01 | 2.64 s/step | 19 GB
|
| 136 |
+
step 5850 | loss 0.0132 | obs 0.0150 ret 0.0134 | gn 0.04 | dW_v 1.57e-01 | 2.55 s/step | 19 GB
|
| 137 |
+
step 5900 | loss 0.0139 | obs 0.0175 ret 0.0118 | gn 0.04 | dW_v 1.54e-01 | 2.63 s/step | 19 GB
|
| 138 |
+
step 5950 | loss 0.0145 | obs 0.0170 ret 0.0159 | gn 0.04 | dW_v 1.54e-01 | 2.65 s/step | 19 GB
|
| 139 |
+
step 6000 | loss 0.0138 | obs 0.0158 ret 0.0160 | gn 0.04 | dW_v 1.61e-01 | 2.83 s/step | 19 GB
|
| 140 |
+
step 6050 | loss 0.0139 | obs 0.0149 ret 0.0179 | gn 0.04 | dW_v 1.57e-01 | 2.66 s/step | 19 GB
|
| 141 |
+
step 6100 | loss 0.0135 | obs 0.0150 ret 0.0134 | gn 0.04 | dW_v 1.63e-01 | 2.62 s/step | 19 GB
|
| 142 |
+
step 6150 | loss 0.0138 | obs 0.0148 ret 0.0211 | gn 0.04 | dW_v 1.57e-01 | 2.59 s/step | 19 GB
|
| 143 |
+
step 6200 | loss 0.0135 | obs 0.0152 ret 0.0136 | gn 0.04 | dW_v 1.62e-01 | 2.50 s/step | 19 GB
|
| 144 |
+
step 6250 | loss 0.0145 | obs 0.0175 ret 0.0160 | gn 0.04 | dW_v 1.55e-01 | 2.63 s/step | 19 GB
|
| 145 |
+
step 6300 | loss 0.0144 | obs 0.0168 ret 0.0157 | gn 0.04 | dW_v 1.61e-01 | 2.69 s/step | 19 GB
|
| 146 |
+
step 6350 | loss 0.0128 | obs 0.0170 ret 0.0170 | gn 0.03 | dW_v 1.57e-01 | 2.42 s/step | 19 GB
|
| 147 |
+
step 6400 | loss 0.0127 | obs 0.0144 ret 0.0145 | gn 0.04 | dW_v 1.62e-01 | 2.43 s/step | 19 GB
|
| 148 |
+
step 6450 | loss 0.0129 | obs 0.0139 ret 0.0149 | gn 0.04 | dW_v 1.63e-01 | 2.50 s/step | 19 GB
|
| 149 |
+
step 6500 | loss 0.0136 | obs 0.0148 ret 0.0117 | gn 0.04 | dW_v 1.63e-01 | 2.53 s/step | 19 GB
|
| 150 |
+
step 6550 | loss 0.0126 | obs 0.0148 ret 0.0187 | gn 0.04 | dW_v 1.56e-01 | 2.35 s/step | 19 GB
|
| 151 |
+
step 6600 | loss 0.0143 | obs 0.0172 ret 0.0152 | gn 0.04 | dW_v 1.57e-01 | 2.51 s/step | 19 GB
|
| 152 |
+
step 6650 | loss 0.0142 | obs 0.0149 ret 0.0182 | gn 0.04 | dW_v 1.59e-01 | 2.62 s/step | 19 GB
|
| 153 |
+
step 6700 | loss 0.0138 | obs 0.0165 ret 0.0170 | gn 0.04 | dW_v 1.64e-01 | 2.74 s/step | 19 GB
|
| 154 |
+
step 6750 | loss 0.0131 | obs 0.0153 ret 0.0158 | gn 0.04 | dW_v 1.64e-01 | 2.50 s/step | 19 GB
|
| 155 |
+
step 6800 | loss 0.0122 | obs 0.0161 ret 0.0108 | gn 0.04 | dW_v 1.61e-01 | 2.40 s/step | 19 GB
|
| 156 |
+
step 6850 | loss 0.0140 | obs 0.0162 ret 0.0157 | gn 0.04 | dW_v 1.45e-01 | 2.33 s/step | 19 GB
|
| 157 |
+
step 6900 | loss 0.0129 | obs 0.0154 ret 0.0137 | gn 0.04 | dW_v 1.60e-01 | 2.63 s/step | 19 GB
|
| 158 |
+
step 6950 | loss 0.0138 | obs 0.0158 ret 0.0137 | gn 0.04 | dW_v 1.65e-01 | 2.47 s/step | 19 GB
|
| 159 |
+
step 7000 | loss 0.0132 | obs 0.0176 ret 0.0105 | gn 0.04 | dW_v 1.64e-01 | 2.34 s/step | 19 GB
|
| 160 |
+
step 7050 | loss 0.0132 | obs 0.0164 ret 0.0127 | gn 0.07 | dW_v 1.59e-01 | 2.40 s/step | 19 GB
|
| 161 |
+
step 7100 | loss 0.0118 | obs 0.0147 ret 0.0106 | gn 0.04 | dW_v 1.56e-01 | 2.25 s/step | 19 GB
|
| 162 |
+
step 7150 | loss 0.0137 | obs 0.0157 ret 0.0153 | gn 0.04 | dW_v 1.58e-01 | 2.83 s/step | 19 GB
|
| 163 |
+
step 7200 | loss 0.0134 | obs 0.0159 ret 0.0205 | gn 0.04 | dW_v 1.66e-01 | 2.59 s/step | 19 GB
|
| 164 |
+
step 7250 | loss 0.0146 | obs 0.0153 ret 0.0186 | gn 0.04 | dW_v 1.62e-01 | 2.46 s/step | 19 GB
|
| 165 |
+
step 7300 | loss 0.0136 | obs 0.0165 ret 0.0124 | gn 0.04 | dW_v 1.59e-01 | 2.47 s/step | 19 GB
|
| 166 |
+
step 7350 | loss 0.0126 | obs 0.0148 ret 0.0133 | gn 0.04 | dW_v 1.50e-01 | 2.45 s/step | 19 GB
|
| 167 |
+
step 7400 | loss 0.0124 | obs 0.0162 ret 0.0124 | gn 0.04 | dW_v 1.71e-01 | 2.46 s/step | 19 GB
|
| 168 |
+
step 7450 | loss 0.0121 | obs 0.0147 ret 0.0155 | gn 0.04 | dW_v 1.49e-01 | 2.33 s/step | 19 GB
|
| 169 |
+
step 7500 | loss 0.0135 | obs 0.0150 ret 0.0172 | gn 0.04 | dW_v 1.60e-01 | 2.51 s/step | 19 GB
|
| 170 |
+
step 7550 | loss 0.0123 | obs 0.0160 ret 0.0164 | gn 0.04 | dW_v 1.74e-01 | 2.58 s/step | 19 GB
|
| 171 |
+
step 7600 | loss 0.0128 | obs 0.0172 ret 0.0143 | gn 0.04 | dW_v 1.60e-01 | 2.33 s/step | 19 GB
|
| 172 |
+
step 7650 | loss 0.0125 | obs 0.0135 ret 0.0122 | gn 0.04 | dW_v 1.66e-01 | 2.45 s/step | 19 GB
|
| 173 |
+
step 7700 | loss 0.0144 | obs 0.0153 ret 0.0204 | gn 0.04 | dW_v 1.77e-01 | 2.28 s/step | 19 GB
|
| 174 |
+
step 7750 | loss 0.0134 | obs 0.0171 ret 0.0104 | gn 0.05 | dW_v 1.64e-01 | 2.26 s/step | 19 GB
|
| 175 |
+
step 7800 | loss 0.0129 | obs 0.0147 ret 0.0110 | gn 0.04 | dW_v 1.80e-01 | 2.42 s/step | 19 GB
|
| 176 |
+
step 7850 | loss 0.0125 | obs 0.0150 ret 0.0145 | gn 0.04 | dW_v 1.67e-01 | 2.58 s/step | 19 GB
|
| 177 |
+
step 7900 | loss 0.0115 | obs 0.0163 ret 0.0086 | gn 0.04 | dW_v 1.73e-01 | 2.41 s/step | 19 GB
|
| 178 |
+
step 7950 | loss 0.0118 | obs 0.0157 ret 0.0100 | gn 0.03 | dW_v 1.84e-01 | 2.36 s/step | 19 GB
|
| 179 |
+
step 8000 | loss 0.0123 | obs 0.0149 ret 0.0146 | gn 0.04 | dW_v 1.74e-01 | 2.28 s/step | 19 GB
|
| 180 |
+
step 8050 | loss 0.0124 | obs 0.0137 ret 0.0135 | gn 0.04 | dW_v 1.76e-01 | 2.53 s/step | 19 GB
|
| 181 |
+
step 8100 | loss 0.0140 | obs 0.0169 ret 0.0150 | gn 0.04 | dW_v 1.76e-01 | 2.35 s/step | 19 GB
|
| 182 |
+
step 8150 | loss 0.0133 | obs 0.0158 ret 0.0131 | gn 0.04 | dW_v 1.67e-01 | 2.57 s/step | 19 GB
|
| 183 |
+
step 8200 | loss 0.0119 | obs 0.0142 ret 0.0122 | gn 0.04 | dW_v 1.65e-01 | 2.40 s/step | 19 GB
|
| 184 |
+
step 8250 | loss 0.0115 | obs 0.0145 ret 0.0142 | gn 0.03 | dW_v 1.76e-01 | 2.46 s/step | 19 GB
|
| 185 |
+
step 8300 | loss 0.0121 | obs 0.0173 ret 0.0112 | gn 0.04 | dW_v 1.78e-01 | 2.44 s/step | 19 GB
|
| 186 |
+
step 8350 | loss 0.0125 | obs 0.0147 ret 0.0110 | gn 0.04 | dW_v 1.80e-01 | 2.35 s/step | 19 GB
|
| 187 |
+
step 8400 | loss 0.0132 | obs 0.0156 ret 0.0169 | gn 0.04 | dW_v 1.81e-01 | 2.52 s/step | 19 GB
|
| 188 |
+
step 8450 | loss 0.0135 | obs 0.0153 ret 0.0152 | gn 0.04 | dW_v 1.65e-01 | 2.32 s/step | 19 GB
|
| 189 |
+
step 8500 | loss 0.0135 | obs 0.0168 ret 0.0141 | gn 0.04 | dW_v 1.77e-01 | 2.38 s/step | 19 GB
|
| 190 |
+
step 8550 | loss 0.0126 | obs 0.0159 ret 0.0111 | gn 0.04 | dW_v 1.69e-01 | 2.45 s/step | 19 GB
|
| 191 |
+
step 8600 | loss 0.0121 | obs 0.0158 ret 0.0143 | gn 0.04 | dW_v 1.80e-01 | 2.25 s/step | 19 GB
|
| 192 |
+
step 8650 | loss 0.0122 | obs 0.0155 ret 0.0108 | gn 0.04 | dW_v 1.68e-01 | 2.25 s/step | 19 GB
|
| 193 |
+
step 8700 | loss 0.0117 | obs 0.0153 ret 0.0141 | gn 0.03 | dW_v 1.82e-01 | 2.48 s/step | 19 GB
|
| 194 |
+
step 8750 | loss 0.0128 | obs 0.0154 ret 0.0119 | gn 0.04 | dW_v 1.82e-01 | 2.49 s/step | 19 GB
|
| 195 |
+
step 8800 | loss 0.0116 | obs 0.0155 ret 0.0093 | gn 0.03 | dW_v 1.81e-01 | 2.37 s/step | 19 GB
|
| 196 |
+
step 8850 | loss 0.0119 | obs 0.0136 ret 0.0177 | gn 0.03 | dW_v 1.81e-01 | 2.41 s/step | 19 GB
|
| 197 |
+
step 8900 | loss 0.0123 | obs 0.0152 ret 0.0122 | gn 0.04 | dW_v 1.79e-01 | 2.38 s/step | 19 GB
|
| 198 |
+
step 8950 | loss 0.0115 | obs 0.0154 ret 0.0133 | gn 0.04 | dW_v 1.77e-01 | 2.43 s/step | 19 GB
|
| 199 |
+
step 9000 | loss 0.0127 | obs 0.0177 ret 0.0146 | gn 0.04 | dW_v 1.58e-01 | 2.44 s/step | 19 GB
|
| 200 |
+
step 9050 | loss 0.0130 | obs 0.0156 ret 0.0159 | gn 0.04 | dW_v 1.85e-01 | 2.42 s/step | 19 GB
|
| 201 |
+
step 9100 | loss 0.0139 | obs 0.0177 ret 0.0129 | gn 0.04 | dW_v 1.75e-01 | 2.34 s/step | 19 GB
|
| 202 |
+
step 9150 | loss 0.0119 | obs 0.0141 ret 0.0143 | gn 0.04 | dW_v 1.89e-01 | 2.39 s/step | 19 GB
|
| 203 |
+
step 9200 | loss 0.0118 | obs 0.0130 ret 0.0106 | gn 0.04 | dW_v 1.85e-01 | 2.53 s/step | 19 GB
|
| 204 |
+
step 9250 | loss 0.0130 | obs 0.0168 ret 0.0128 | gn 0.04 | dW_v 1.52e-01 | 2.36 s/step | 19 GB
|
| 205 |
+
step 9300 | loss 0.0129 | obs 0.0164 ret 0.0130 | gn 0.04 | dW_v 1.81e-01 | 2.30 s/step | 19 GB
|
| 206 |
+
step 9350 | loss 0.0142 | obs 0.0175 ret 0.0149 | gn 0.04 | dW_v 1.73e-01 | 2.56 s/step | 19 GB
|
| 207 |
+
step 9400 | loss 0.0136 | obs 0.0175 ret 0.0154 | gn 0.04 | dW_v 1.81e-01 | 2.29 s/step | 19 GB
|
| 208 |
+
step 9450 | loss 0.0124 | obs 0.0165 ret 0.0143 | gn 0.04 | dW_v 1.83e-01 | 2.34 s/step | 19 GB
|
| 209 |
+
step 9500 | loss 0.0121 | obs 0.0156 ret 0.0124 | gn 0.04 | dW_v 1.63e-01 | 2.47 s/step | 19 GB
|
| 210 |
+
step 9550 | loss 0.0126 | obs 0.0170 ret 0.0122 | gn 0.04 | dW_v 1.80e-01 | 2.57 s/step | 19 GB
|
| 211 |
+
step 9600 | loss 0.0125 | obs 0.0162 ret 0.0155 | gn 0.04 | dW_v 1.65e-01 | 2.41 s/step | 19 GB
|
| 212 |
+
step 9650 | loss 0.0124 | obs 0.0180 ret 0.0096 | gn 0.04 | dW_v 1.80e-01 | 2.26 s/step | 19 GB
|
| 213 |
+
step 9700 | loss 0.0123 | obs 0.0157 ret 0.0107 | gn 0.04 | dW_v 1.61e-01 | 2.25 s/step | 19 GB
|
| 214 |
+
step 9750 | loss 0.0114 | obs 0.0151 ret 0.0146 | gn 0.04 | dW_v 1.76e-01 | 2.45 s/step | 19 GB
|
| 215 |
+
step 9800 | loss 0.0134 | obs 0.0177 ret 0.0149 | gn 0.04 | dW_v 1.72e-01 | 2.32 s/step | 19 GB
|
| 216 |
+
step 9850 | loss 0.0113 | obs 0.0140 ret 0.0100 | gn 0.04 | dW_v 1.74e-01 | 2.30 s/step | 19 GB
|
| 217 |
+
step 9900 | loss 0.0128 | obs 0.0160 ret 0.0130 | gn 0.04 | dW_v 1.86e-01 | 2.32 s/step | 19 GB
|
| 218 |
+
step 9950 | loss 0.0112 | obs 0.0141 ret 0.0123 | gn 0.04 | dW_v 1.50e-01 | 2.46 s/step | 19 GB
|
| 219 |
+
step 10000 | loss 0.0128 | obs 0.0173 ret 0.0117 | gn 0.04 | dW_v 1.79e-01 | 2.27 s/step | 19 GB
|
| 220 |
+
step 10050 | loss 0.0116 | obs 0.0148 ret 0.0134 | gn 0.04 | dW_v 1.76e-01 | 2.63 s/step | 19 GB
|
| 221 |
+
step 10100 | loss 0.0118 | obs 0.0144 ret 0.0129 | gn 0.04 | dW_v 1.88e-01 | 2.41 s/step | 19 GB
|
| 222 |
+
run rk_v | MemConfig(proj='v', ffn=False, layers=(0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11), fast_heads=4, blocks=4, ffn_frac=0.25, ffn_blocks=8, loss='delta', key='raw', lr=0.001, renorm=True) | max_ctx 0 | new params 4.00M | fast weights / sample: writable 1.00M, state 3.98M (15 MB fp32) | start step 10000
|
| 223 |
+
wandb: Currently logged in as: xsourse to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
|
| 224 |
+
wandb: creating run
|
| 225 |
+
wandb: Tracking run with wandb version 0.22.0
|
| 226 |
+
wandb: Run data is saved locally in /mnt/localssd/lehany/wv/E1_rubik_20261104031751/rk_v/wandb/run-20261105_174950-w5mv6kf0
|
| 227 |
+
wandb: Run `wandb offline` to turn off syncing.
|
| 228 |
+
wandb: Resuming run rk_v
|
| 229 |
+
wandb: ⭐️ View project at https://wandb.ai/xsourse/world-auto-visual
|
| 230 |
+
wandb: 🚀 View run at https://wandb.ai/xsourse/world-auto-visual/runs/w5mv6kf0
|
| 231 |
+
step 10050 | loss 0.0117 | obs 0.0149 ret 0.0127 | gn 0.03 | dW_v 1.79e-01 | 2.76 s/step | 19 GB
|
| 232 |
+
wandb: WARNING Tried to log to step 10050 that is less than the current step 10051. Steps must be monotonically increasing, so this data will be ignored. See https://wandb.me/define-metric to log data out of order.
|
| 233 |
+
step 10100 | loss 0.0124 | obs 0.0159 ret 0.0121 | gn 0.03 | dW_v 1.76e-01 | 2.61 s/step | 19 GB
|
| 234 |
+
step 10150 | loss 0.0117 | obs 0.0157 ret 0.0099 | gn 0.03 | dW_v 1.80e-01 | 2.41 s/step | 19 GB
|
| 235 |
+
step 10200 | loss 0.0125 | obs 0.0162 ret 0.0142 | gn 0.04 | dW_v 1.52e-01 | 2.43 s/step | 19 GB
|
| 236 |
+
step 10250 | loss 0.0110 | obs 0.0144 ret 0.0145 | gn 0.03 | dW_v 1.88e-01 | 2.46 s/step | 19 GB
|
| 237 |
+
step 10300 | loss 0.0106 | obs 0.0146 ret 0.0109 | gn 0.03 | dW_v 1.91e-01 | 2.39 s/step | 19 GB
|
| 238 |
+
step 10350 | loss 0.0114 | obs 0.0138 ret 0.0106 | gn 0.03 | dW_v 1.89e-01 | 2.45 s/step | 19 GB
|
| 239 |
+
step 10400 | loss 0.0134 | obs 0.0193 ret 0.0154 | gn 0.04 | dW_v 1.89e-01 | 2.62 s/step | 19 GB
|
| 240 |
+
step 10450 | loss 0.0116 | obs 0.0161 ret 0.0088 | gn 0.04 | dW_v 1.81e-01 | 2.49 s/step | 19 GB
|
| 241 |
+
step 10500 | loss 0.0120 | obs 0.0156 ret 0.0157 | gn 0.04 | dW_v 1.75e-01 | 2.55 s/step | 19 GB
|
| 242 |
+
step 10550 | loss 0.0109 | obs 0.0162 ret 0.0101 | gn 0.04 | dW_v 1.74e-01 | 2.60 s/step | 19 GB
|
| 243 |
+
step 10600 | loss 0.0133 | obs 0.0182 ret 0.0119 | gn 0.04 | dW_v 1.83e-01 | 2.41 s/step | 19 GB
|
| 244 |
+
step 10650 | loss 0.0116 | obs 0.0143 ret 0.0148 | gn 0.04 | dW_v 1.56e-01 | 2.44 s/step | 19 GB
|
| 245 |
+
step 10700 | loss 0.0124 | obs 0.0159 ret 0.0102 | gn 0.04 | dW_v 1.90e-01 | 2.42 s/step | 19 GB
|
| 246 |
+
step 10750 | loss 0.0129 | obs 0.0151 ret 0.0170 | gn 0.04 | dW_v 1.56e-01 | 2.38 s/step | 19 GB
|
| 247 |
+
step 10800 | loss 0.0119 | obs 0.0149 ret 0.0105 | gn 0.04 | dW_v 1.83e-01 | 2.44 s/step | 19 GB
|
| 248 |
+
step 10850 | loss 0.0114 | obs 0.0146 ret 0.0102 | gn 0.03 | dW_v 1.87e-01 | 2.49 s/step | 19 GB
|
| 249 |
+
step 10900 | loss 0.0123 | obs 0.0159 ret 0.0107 | gn 0.04 | dW_v 1.84e-01 | 2.44 s/step | 19 GB
|
| 250 |
+
step 10950 | loss 0.0125 | obs 0.0158 ret 0.0126 | gn 0.03 | dW_v 1.89e-01 | 2.52 s/step | 19 GB
|
| 251 |
+
step 11000 | loss 0.0124 | obs 0.0145 ret 0.0171 | gn 0.03 | dW_v 1.95e-01 | 2.56 s/step | 19 GB
|
| 252 |
+
step 11050 | loss 0.0118 | obs 0.0158 ret 0.0109 | gn 0.04 | dW_v 1.51e-01 | 2.53 s/step | 19 GB
|
| 253 |
+
step 11100 | loss 0.0114 | obs 0.0149 ret 0.0106 | gn 0.04 | dW_v 1.96e-01 | 2.47 s/step | 19 GB
|
| 254 |
+
step 11150 | loss 0.0116 | obs 0.0145 ret 0.0135 | gn 0.04 | dW_v 1.80e-01 | 2.44 s/step | 19 GB
|
| 255 |
+
step 11200 | loss 0.0106 | obs 0.0170 ret 0.0081 | gn 0.03 | dW_v 1.54e-01 | 2.50 s/step | 19 GB
|
| 256 |
+
step 11250 | loss 0.0115 | obs 0.0135 ret 0.0124 | gn 0.04 | dW_v 1.76e-01 | 2.41 s/step | 19 GB
|
| 257 |
+
step 11300 | loss 0.0123 | obs 0.0162 ret 0.0124 | gn 0.04 | dW_v 1.52e-01 | 2.47 s/step | 19 GB
|
| 258 |
+
step 11350 | loss 0.0118 | obs 0.0159 ret 0.0157 | gn 0.04 | dW_v 1.82e-01 | 2.45 s/step | 19 GB
|
| 259 |
+
step 11400 | loss 0.0120 | obs 0.0162 ret 0.0124 | gn 0.04 | dW_v 1.52e-01 | 2.52 s/step | 19 GB
|
| 260 |
+
step 11450 | loss 0.0114 | obs 0.0145 ret 0.0111 | gn 0.04 | dW_v 1.79e-01 | 2.53 s/step | 19 GB
|
| 261 |
+
step 11500 | loss 0.0105 | obs 0.0164 ret 0.0101 | gn 0.03 | dW_v 1.54e-01 | 2.48 s/step | 19 GB
|
| 262 |
+
step 11550 | loss 0.0127 | obs 0.0169 ret 0.0140 | gn 0.04 | dW_v 1.93e-01 | 2.61 s/step | 19 GB
|
| 263 |
+
step 11600 | loss 0.0113 | obs 0.0175 ret 0.0103 | gn 0.03 | dW_v 1.92e-01 | 2.65 s/step | 19 GB
|
| 264 |
+
step 11650 | loss 0.0116 | obs 0.0167 ret 0.0108 | gn 0.04 | dW_v 1.84e-01 | 2.45 s/step | 19 GB
|
| 265 |
+
step 11700 | loss 0.0116 | obs 0.0148 ret 0.0181 | gn 0.04 | dW_v 1.57e-01 | 2.52 s/step | 19 GB
|
| 266 |
+
step 11750 | loss 0.0121 | obs 0.0183 ret 0.0100 | gn 0.03 | dW_v 1.95e-01 | 2.59 s/step | 19 GB
|
| 267 |
+
step 11800 | loss 0.0121 | obs 0.0164 ret 0.0080 | gn 0.03 | dW_v 1.84e-01 | 2.41 s/step | 19 GB
|
| 268 |
+
step 11850 | loss 0.0104 | obs 0.0152 ret 0.0069 | gn 0.03 | dW_v 1.81e-01 | 2.57 s/step | 19 GB
|
| 269 |
+
step 11900 | loss 0.0126 | obs 0.0170 ret 0.0113 | gn 0.04 | dW_v 1.85e-01 | 2.52 s/step | 19 GB
|
| 270 |
+
step 11950 | loss 0.0122 | obs 0.0169 ret 0.0099 | gn 0.04 | dW_v 1.53e-01 | 2.42 s/step | 19 GB
|
| 271 |
+
step 12000 | loss 0.0120 | obs 0.0175 ret 0.0105 | gn 0.03 | dW_v 1.91e-01 | 2.49 s/step | 19 GB
|
| 272 |
+
[1;34mwandb[0m:
|
| 273 |
+
[1;34mwandb[0m: 🚀 View run [33mrk_v[0m at: [34m[0m
|
| 274 |
+
[1;34mwandb[0m: Find logs at: [1;35m../../../../../../wv/E1_rubik_20261104031751/rk_v/wandb/run-20261105_174950-w5mv6kf0/logs[0m
|
rubik_r1_20261106011120/rk_ctrl/info.json
ADDED
|
@@ -0,0 +1,62 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"args": {
|
| 3 |
+
"run": "rk_ctrl",
|
| 4 |
+
"out": "/mnt/localssd/lehany/wv/E1_rubik_20261104031751",
|
| 5 |
+
"task": "rubik",
|
| 6 |
+
"base": "/mnt/localssd/lehany/wv/E1_rubik_20261104031751/rubik_base/ema_040000.pt",
|
| 7 |
+
"base_gate": 0.0,
|
| 8 |
+
"size": "200m",
|
| 9 |
+
"max_ctx": 0,
|
| 10 |
+
"mem_proj": "",
|
| 11 |
+
"mem_ffn": 0,
|
| 12 |
+
"mem_layers": "all",
|
| 13 |
+
"mem_fast_heads": 4,
|
| 14 |
+
"mem_blocks": 4,
|
| 15 |
+
"mem_ffn_frac": 0.25,
|
| 16 |
+
"mem_ffn_blocks": 8,
|
| 17 |
+
"mem_loss": "delta",
|
| 18 |
+
"mem_key": "raw",
|
| 19 |
+
"mem_lr": 0.001,
|
| 20 |
+
"mem_renorm": 1,
|
| 21 |
+
"mem_lr_lr": 0.001,
|
| 22 |
+
"gaps": [
|
| 23 |
+
0,
|
| 24 |
+
1,
|
| 25 |
+
2,
|
| 26 |
+
3,
|
| 27 |
+
4,
|
| 28 |
+
5,
|
| 29 |
+
6,
|
| 30 |
+
7,
|
| 31 |
+
8
|
| 32 |
+
],
|
| 33 |
+
"batch": 8,
|
| 34 |
+
"accum": 1,
|
| 35 |
+
"lr": 0.0001,
|
| 36 |
+
"steps": 12000,
|
| 37 |
+
"warmup": 200,
|
| 38 |
+
"ema": 0.999,
|
| 39 |
+
"grad_ckpt": 1,
|
| 40 |
+
"eval_every": 0,
|
| 41 |
+
"eval_gaps": [
|
| 42 |
+
0,
|
| 43 |
+
4,
|
| 44 |
+
8,
|
| 45 |
+
16
|
| 46 |
+
],
|
| 47 |
+
"eval_num": 32,
|
| 48 |
+
"final_eval": 0,
|
| 49 |
+
"save_every": 500,
|
| 50 |
+
"keep_every": 4000,
|
| 51 |
+
"log_every": 50,
|
| 52 |
+
"seed": 0,
|
| 53 |
+
"gpu": 0,
|
| 54 |
+
"workers": 6,
|
| 55 |
+
"wandb": 1
|
| 56 |
+
},
|
| 57 |
+
"mem_new_params": 0,
|
| 58 |
+
"mem_writable_per_sample": 0,
|
| 59 |
+
"mem_state_per_sample": 0,
|
| 60 |
+
"mem_state_mb_fp32": 0.0,
|
| 61 |
+
"gpu": "NVIDIA H100 80GB HBM3"
|
| 62 |
+
}
|
rubik_r1_20261106011120/rk_ctrl/log.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
rubik_r1_20261106011120/rk_ffn/info.json
ADDED
|
@@ -0,0 +1,62 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"args": {
|
| 3 |
+
"run": "rk_ffn",
|
| 4 |
+
"out": "/mnt/localssd/lehany/wv/E1_rubik_20261104031751",
|
| 5 |
+
"task": "rubik",
|
| 6 |
+
"base": "/mnt/localssd/lehany/wv/E1_rubik_20261104031751/rubik_base/ema_040000.pt",
|
| 7 |
+
"base_gate": 0.0,
|
| 8 |
+
"size": "200m",
|
| 9 |
+
"max_ctx": 0,
|
| 10 |
+
"mem_proj": "",
|
| 11 |
+
"mem_ffn": 1,
|
| 12 |
+
"mem_layers": "all",
|
| 13 |
+
"mem_fast_heads": 4,
|
| 14 |
+
"mem_blocks": 4,
|
| 15 |
+
"mem_ffn_frac": 0.25,
|
| 16 |
+
"mem_ffn_blocks": 8,
|
| 17 |
+
"mem_loss": "delta",
|
| 18 |
+
"mem_key": "raw",
|
| 19 |
+
"mem_lr": 0.001,
|
| 20 |
+
"mem_renorm": 1,
|
| 21 |
+
"mem_lr_lr": 0.001,
|
| 22 |
+
"gaps": [
|
| 23 |
+
0,
|
| 24 |
+
1,
|
| 25 |
+
2,
|
| 26 |
+
3,
|
| 27 |
+
4,
|
| 28 |
+
5,
|
| 29 |
+
6,
|
| 30 |
+
7,
|
| 31 |
+
8
|
| 32 |
+
],
|
| 33 |
+
"batch": 8,
|
| 34 |
+
"accum": 1,
|
| 35 |
+
"lr": 0.0001,
|
| 36 |
+
"steps": 12000,
|
| 37 |
+
"warmup": 200,
|
| 38 |
+
"ema": 0.999,
|
| 39 |
+
"grad_ckpt": 1,
|
| 40 |
+
"eval_every": 0,
|
| 41 |
+
"eval_gaps": [
|
| 42 |
+
0,
|
| 43 |
+
4,
|
| 44 |
+
8,
|
| 45 |
+
16
|
| 46 |
+
],
|
| 47 |
+
"eval_num": 32,
|
| 48 |
+
"final_eval": 0,
|
| 49 |
+
"save_every": 500,
|
| 50 |
+
"keep_every": 4000,
|
| 51 |
+
"log_every": 50,
|
| 52 |
+
"seed": 0,
|
| 53 |
+
"gpu": 0,
|
| 54 |
+
"workers": 6,
|
| 55 |
+
"wandb": 1
|
| 56 |
+
},
|
| 57 |
+
"mem_new_params": 15939084,
|
| 58 |
+
"mem_writable_per_sample": 1990656,
|
| 59 |
+
"mem_state_per_sample": 15925248,
|
| 60 |
+
"mem_state_mb_fp32": 60.75,
|
| 61 |
+
"gpu": "NVIDIA H100 80GB HBM3"
|
| 62 |
+
}
|
rubik_r1_20261106011120/rk_ffn/log.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
rubik_r1_20261106011120/rk_q/info.json
ADDED
|
@@ -0,0 +1,62 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"args": {
|
| 3 |
+
"run": "rk_q",
|
| 4 |
+
"out": "/mnt/localssd/lehany/wv/E1_rubik_20261104031751",
|
| 5 |
+
"task": "rubik",
|
| 6 |
+
"base": "/mnt/localssd/lehany/wv/E1_rubik_20261104031751/rubik_base/ema_040000.pt",
|
| 7 |
+
"base_gate": 0.0,
|
| 8 |
+
"size": "200m",
|
| 9 |
+
"max_ctx": 0,
|
| 10 |
+
"mem_proj": "q",
|
| 11 |
+
"mem_ffn": 0,
|
| 12 |
+
"mem_layers": "all",
|
| 13 |
+
"mem_fast_heads": 4,
|
| 14 |
+
"mem_blocks": 4,
|
| 15 |
+
"mem_ffn_frac": 0.25,
|
| 16 |
+
"mem_ffn_blocks": 8,
|
| 17 |
+
"mem_loss": "delta",
|
| 18 |
+
"mem_key": "raw",
|
| 19 |
+
"mem_lr": 0.001,
|
| 20 |
+
"mem_renorm": 1,
|
| 21 |
+
"mem_lr_lr": 0.001,
|
| 22 |
+
"gaps": [
|
| 23 |
+
0,
|
| 24 |
+
1,
|
| 25 |
+
2,
|
| 26 |
+
3,
|
| 27 |
+
4,
|
| 28 |
+
5,
|
| 29 |
+
6,
|
| 30 |
+
7,
|
| 31 |
+
8
|
| 32 |
+
],
|
| 33 |
+
"batch": 8,
|
| 34 |
+
"accum": 1,
|
| 35 |
+
"lr": 0.0001,
|
| 36 |
+
"steps": 12000,
|
| 37 |
+
"warmup": 200,
|
| 38 |
+
"ema": 0.999,
|
| 39 |
+
"grad_ckpt": 1,
|
| 40 |
+
"eval_every": 0,
|
| 41 |
+
"eval_gaps": [
|
| 42 |
+
0,
|
| 43 |
+
4,
|
| 44 |
+
8,
|
| 45 |
+
16
|
| 46 |
+
],
|
| 47 |
+
"eval_num": 32,
|
| 48 |
+
"final_eval": 0,
|
| 49 |
+
"save_every": 500,
|
| 50 |
+
"keep_every": 4000,
|
| 51 |
+
"log_every": 50,
|
| 52 |
+
"seed": 0,
|
| 53 |
+
"gpu": 0,
|
| 54 |
+
"workers": 6,
|
| 55 |
+
"wandb": 1
|
| 56 |
+
},
|
| 57 |
+
"mem_new_params": 3995148,
|
| 58 |
+
"mem_writable_per_sample": 995328,
|
| 59 |
+
"mem_state_per_sample": 3981312,
|
| 60 |
+
"mem_state_mb_fp32": 15.1875,
|
| 61 |
+
"gpu": "NVIDIA H100 80GB HBM3"
|
| 62 |
+
}
|
rubik_r1_20261106011120/rk_q/log.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
rubik_r1_20261106011120/rk_qkv/info.json
ADDED
|
@@ -0,0 +1,62 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"args": {
|
| 3 |
+
"run": "rk_qkv",
|
| 4 |
+
"out": "/mnt/localssd/lehany/wv/E1_rubik_20261104031751",
|
| 5 |
+
"task": "rubik",
|
| 6 |
+
"base": "/mnt/localssd/lehany/wv/E1_rubik_20261104031751/rubik_base/ema_040000.pt",
|
| 7 |
+
"base_gate": 0.0,
|
| 8 |
+
"size": "200m",
|
| 9 |
+
"max_ctx": 0,
|
| 10 |
+
"mem_proj": "qkv",
|
| 11 |
+
"mem_ffn": 0,
|
| 12 |
+
"mem_layers": "all",
|
| 13 |
+
"mem_fast_heads": 4,
|
| 14 |
+
"mem_blocks": 4,
|
| 15 |
+
"mem_ffn_frac": 0.25,
|
| 16 |
+
"mem_ffn_blocks": 8,
|
| 17 |
+
"mem_loss": "delta",
|
| 18 |
+
"mem_key": "raw",
|
| 19 |
+
"mem_lr": 0.001,
|
| 20 |
+
"mem_renorm": 1,
|
| 21 |
+
"mem_lr_lr": 0.001,
|
| 22 |
+
"gaps": [
|
| 23 |
+
0,
|
| 24 |
+
1,
|
| 25 |
+
2,
|
| 26 |
+
3,
|
| 27 |
+
4,
|
| 28 |
+
5,
|
| 29 |
+
6,
|
| 30 |
+
7,
|
| 31 |
+
8
|
| 32 |
+
],
|
| 33 |
+
"batch": 8,
|
| 34 |
+
"accum": 1,
|
| 35 |
+
"lr": 0.0001,
|
| 36 |
+
"steps": 12000,
|
| 37 |
+
"warmup": 200,
|
| 38 |
+
"ema": 0.999,
|
| 39 |
+
"grad_ckpt": 1,
|
| 40 |
+
"eval_every": 0,
|
| 41 |
+
"eval_gaps": [
|
| 42 |
+
0,
|
| 43 |
+
4,
|
| 44 |
+
8,
|
| 45 |
+
16
|
| 46 |
+
],
|
| 47 |
+
"eval_num": 32,
|
| 48 |
+
"final_eval": 0,
|
| 49 |
+
"save_every": 500,
|
| 50 |
+
"keep_every": 4000,
|
| 51 |
+
"log_every": 50,
|
| 52 |
+
"seed": 0,
|
| 53 |
+
"gpu": 0,
|
| 54 |
+
"workers": 6,
|
| 55 |
+
"wandb": 1
|
| 56 |
+
},
|
| 57 |
+
"mem_new_params": 11985444,
|
| 58 |
+
"mem_writable_per_sample": 2985984,
|
| 59 |
+
"mem_state_per_sample": 11943936,
|
| 60 |
+
"mem_state_mb_fp32": 45.5625,
|
| 61 |
+
"gpu": "NVIDIA H100 80GB HBM3"
|
| 62 |
+
}
|
rubik_r1_20261106011120/rk_qkv/log.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
rubik_r1_20261106011120/rk_qkv_norm/info.json
ADDED
|
@@ -0,0 +1,62 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"args": {
|
| 3 |
+
"run": "rk_qkv_norm",
|
| 4 |
+
"out": "/mnt/localssd/lehany/wv/E1_rubik_20261104031751",
|
| 5 |
+
"task": "rubik",
|
| 6 |
+
"base": "/mnt/localssd/lehany/wv/E1_rubik_20261104031751/rubik_base/ema_040000.pt",
|
| 7 |
+
"base_gate": 0.0,
|
| 8 |
+
"size": "200m",
|
| 9 |
+
"max_ctx": 0,
|
| 10 |
+
"mem_proj": "qkv",
|
| 11 |
+
"mem_ffn": 0,
|
| 12 |
+
"mem_layers": "all",
|
| 13 |
+
"mem_fast_heads": 4,
|
| 14 |
+
"mem_blocks": 4,
|
| 15 |
+
"mem_ffn_frac": 0.25,
|
| 16 |
+
"mem_ffn_blocks": 8,
|
| 17 |
+
"mem_loss": "delta",
|
| 18 |
+
"mem_key": "norm",
|
| 19 |
+
"mem_lr": 0.5,
|
| 20 |
+
"mem_renorm": 1,
|
| 21 |
+
"mem_lr_lr": 0.001,
|
| 22 |
+
"gaps": [
|
| 23 |
+
0,
|
| 24 |
+
1,
|
| 25 |
+
2,
|
| 26 |
+
3,
|
| 27 |
+
4,
|
| 28 |
+
5,
|
| 29 |
+
6,
|
| 30 |
+
7,
|
| 31 |
+
8
|
| 32 |
+
],
|
| 33 |
+
"batch": 8,
|
| 34 |
+
"accum": 1,
|
| 35 |
+
"lr": 0.0001,
|
| 36 |
+
"steps": 12000,
|
| 37 |
+
"warmup": 200,
|
| 38 |
+
"ema": 0.999,
|
| 39 |
+
"grad_ckpt": 1,
|
| 40 |
+
"eval_every": 0,
|
| 41 |
+
"eval_gaps": [
|
| 42 |
+
0,
|
| 43 |
+
4,
|
| 44 |
+
8,
|
| 45 |
+
16
|
| 46 |
+
],
|
| 47 |
+
"eval_num": 32,
|
| 48 |
+
"final_eval": 0,
|
| 49 |
+
"save_every": 500,
|
| 50 |
+
"keep_every": 4000,
|
| 51 |
+
"log_every": 50,
|
| 52 |
+
"seed": 0,
|
| 53 |
+
"gpu": 0,
|
| 54 |
+
"workers": 6,
|
| 55 |
+
"wandb": 1
|
| 56 |
+
},
|
| 57 |
+
"mem_new_params": 11985444,
|
| 58 |
+
"mem_writable_per_sample": 2985984,
|
| 59 |
+
"mem_state_per_sample": 11943936,
|
| 60 |
+
"mem_state_mb_fp32": 45.5625,
|
| 61 |
+
"gpu": "NVIDIA H100 80GB HBM3"
|
| 62 |
+
}
|
rubik_r1_20261106011120/rk_qkv_norm/log.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
rubik_r1_20261106011120/rk_v/info.json
ADDED
|
@@ -0,0 +1,62 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"args": {
|
| 3 |
+
"run": "rk_v",
|
| 4 |
+
"out": "/mnt/localssd/lehany/wv/E1_rubik_20261104031751",
|
| 5 |
+
"task": "rubik",
|
| 6 |
+
"base": "/mnt/localssd/lehany/wv/E1_rubik_20261104031751/rubik_base/ema_040000.pt",
|
| 7 |
+
"base_gate": 0.0,
|
| 8 |
+
"size": "200m",
|
| 9 |
+
"max_ctx": 0,
|
| 10 |
+
"mem_proj": "v",
|
| 11 |
+
"mem_ffn": 0,
|
| 12 |
+
"mem_layers": "all",
|
| 13 |
+
"mem_fast_heads": 4,
|
| 14 |
+
"mem_blocks": 4,
|
| 15 |
+
"mem_ffn_frac": 0.25,
|
| 16 |
+
"mem_ffn_blocks": 8,
|
| 17 |
+
"mem_loss": "delta",
|
| 18 |
+
"mem_key": "raw",
|
| 19 |
+
"mem_lr": 0.001,
|
| 20 |
+
"mem_renorm": 1,
|
| 21 |
+
"mem_lr_lr": 0.001,
|
| 22 |
+
"gaps": [
|
| 23 |
+
0,
|
| 24 |
+
1,
|
| 25 |
+
2,
|
| 26 |
+
3,
|
| 27 |
+
4,
|
| 28 |
+
5,
|
| 29 |
+
6,
|
| 30 |
+
7,
|
| 31 |
+
8
|
| 32 |
+
],
|
| 33 |
+
"batch": 8,
|
| 34 |
+
"accum": 1,
|
| 35 |
+
"lr": 0.0001,
|
| 36 |
+
"steps": 12000,
|
| 37 |
+
"warmup": 200,
|
| 38 |
+
"ema": 0.999,
|
| 39 |
+
"grad_ckpt": 1,
|
| 40 |
+
"eval_every": 0,
|
| 41 |
+
"eval_gaps": [
|
| 42 |
+
0,
|
| 43 |
+
4,
|
| 44 |
+
8,
|
| 45 |
+
16
|
| 46 |
+
],
|
| 47 |
+
"eval_num": 32,
|
| 48 |
+
"final_eval": 0,
|
| 49 |
+
"save_every": 500,
|
| 50 |
+
"keep_every": 4000,
|
| 51 |
+
"log_every": 50,
|
| 52 |
+
"seed": 0,
|
| 53 |
+
"gpu": 0,
|
| 54 |
+
"workers": 6,
|
| 55 |
+
"wandb": 1
|
| 56 |
+
},
|
| 57 |
+
"mem_new_params": 3995148,
|
| 58 |
+
"mem_writable_per_sample": 995328,
|
| 59 |
+
"mem_state_per_sample": 3981312,
|
| 60 |
+
"mem_state_mb_fp32": 15.1875,
|
| 61 |
+
"gpu": "NVIDIA H100 80GB HBM3"
|
| 62 |
+
}
|
rubik_r1_20261106011120/rk_v/log.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
rubik_r1_20261106011120/rk_v_norm/info.json
ADDED
|
@@ -0,0 +1,62 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"args": {
|
| 3 |
+
"run": "rk_v_norm",
|
| 4 |
+
"out": "/mnt/localssd/lehany/wv/E1_rubik_20261104031751",
|
| 5 |
+
"task": "rubik",
|
| 6 |
+
"base": "/mnt/localssd/lehany/wv/E1_rubik_20261104031751/rubik_base/ema_040000.pt",
|
| 7 |
+
"base_gate": 0.0,
|
| 8 |
+
"size": "200m",
|
| 9 |
+
"max_ctx": 0,
|
| 10 |
+
"mem_proj": "v",
|
| 11 |
+
"mem_ffn": 0,
|
| 12 |
+
"mem_layers": "all",
|
| 13 |
+
"mem_fast_heads": 4,
|
| 14 |
+
"mem_blocks": 4,
|
| 15 |
+
"mem_ffn_frac": 0.25,
|
| 16 |
+
"mem_ffn_blocks": 8,
|
| 17 |
+
"mem_loss": "delta",
|
| 18 |
+
"mem_key": "norm",
|
| 19 |
+
"mem_lr": 0.5,
|
| 20 |
+
"mem_renorm": 1,
|
| 21 |
+
"mem_lr_lr": 0.001,
|
| 22 |
+
"gaps": [
|
| 23 |
+
0,
|
| 24 |
+
1,
|
| 25 |
+
2,
|
| 26 |
+
3,
|
| 27 |
+
4,
|
| 28 |
+
5,
|
| 29 |
+
6,
|
| 30 |
+
7,
|
| 31 |
+
8
|
| 32 |
+
],
|
| 33 |
+
"batch": 8,
|
| 34 |
+
"accum": 1,
|
| 35 |
+
"lr": 0.0001,
|
| 36 |
+
"steps": 12000,
|
| 37 |
+
"warmup": 200,
|
| 38 |
+
"ema": 0.999,
|
| 39 |
+
"grad_ckpt": 1,
|
| 40 |
+
"eval_every": 0,
|
| 41 |
+
"eval_gaps": [
|
| 42 |
+
0,
|
| 43 |
+
4,
|
| 44 |
+
8,
|
| 45 |
+
16
|
| 46 |
+
],
|
| 47 |
+
"eval_num": 32,
|
| 48 |
+
"final_eval": 0,
|
| 49 |
+
"save_every": 500,
|
| 50 |
+
"keep_every": 4000,
|
| 51 |
+
"log_every": 50,
|
| 52 |
+
"seed": 0,
|
| 53 |
+
"gpu": 0,
|
| 54 |
+
"workers": 6,
|
| 55 |
+
"wandb": 1
|
| 56 |
+
},
|
| 57 |
+
"mem_new_params": 3995148,
|
| 58 |
+
"mem_writable_per_sample": 995328,
|
| 59 |
+
"mem_state_per_sample": 3981312,
|
| 60 |
+
"mem_state_mb_fp32": 15.1875,
|
| 61 |
+
"gpu": "NVIDIA H100 80GB HBM3"
|
| 62 |
+
}
|
rubik_r1_20261106011120/rk_v_norm/log.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
rubik_r1_20261106011120/rubik_base/log.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|