Hugging Face's logo

JohnP1
/
d1a-e2b

Text Classification
PEFT
Safetensors
English
Japanese
d1a
decision-model
calibration
lora
gemma4
typesafe
system-one
pull-requests
on-device
Model card Files
xet
 Community
JohnP1 commited on
Commit
20d6289
·
verified ·
1 Parent(s): 2f1b19e

D1A-E2B v0.2: two epochs (dev acc 0.824 trained / 0.602 new sources)

Browse files
Files changed (9) hide show
  1. README.md +11 -8
  2. adapter_config.json +5 -5
  3. adapter_model.safetensors +1 -1
  4. head.pt +2 -2
  5. provenance.json +28 -48
  6. result.json +0 -0
  7. train.log +395 -190
  8. training_config.json +3 -3
  9. training_metrics.json +7 -7
README.md CHANGED
@@ -11,15 +11,17 @@ tags: [d1a, decision-model, calibration, lora, gemma4, typesafe, system-one]
11
 
12
  **D1A** is a small open decision model in the Jev style: one document and a set of typed questions in, a calibrated probability for every option out, in one forward pass, with no text generation. It speaks the same System One API as Jev, so the TypeSafe SDK works against it unchanged.
13
 
14
- This checkpoint is a LoRA adapter plus a pointer head on `google/gemma-4-E2B` (revision `d29ff6b4`), calibrated with a temperature of 1.48.
15
 
16
- | | D1A-E2B (this checkpoint, 1 epoch) | Jev (TypeSafe, hosted) |
17
- |---|---|---|
18
- | Accuracy, trained sources (dev) | 0.794 | 0.845 |
19
- | Accuracy, new sources (dev) | 0.569 | 0.857 |
20
- | Where it runs | your machine (Apple GPU / NVIDIA), free, private | TypeSafe's cloud API |
 
 
21
 
22
- Measured on the same frozen evaluation sets (development partitions; the locked test was not read). Jev is more accurate today; this is a one-epoch prototype, and a two-epoch model is training. Tag `v0.1-1epoch` keeps this version.
23
 
24
  - Playground (nine live use cases): https://github.com/jonpol01/d1a-playground
25
  - Code: https://github.com/jonpol01/d1a
@@ -27,7 +29,8 @@ Measured on the same frozen evaluation sets (development partitions; the locked
27
  ## Serve it
28
 
29
  ```bash
30
- uv run --extra serve python -m kev.serve --run JohnP1/d1a-e2b --port 8009 # from github.com/jonpol01/kev; D1A's own server (d1a.serve) is on the way
 
31
  ```
32
 
33
  ## License
 
11
 
12
  **D1A** is a small open decision model in the Jev style: one document and a set of typed questions in, a calibrated probability for every option out, in one forward pass, with no text generation. It speaks the same System One API as Jev, so the TypeSafe SDK works against it unchanged.
13
 
14
+ This checkpoint is a LoRA adapter plus a pointer head on `google/gemma-4-E2B` (revision `d29ff6b4`), trained for two epochs and calibrated with a temperature of 1.52.
15
 
16
+ | | D1A-E2B (this checkpoint, 2 epochs) | v0.1 (1 epoch) | Jev (TypeSafe, hosted) |
17
+ |---|---|---|---|
18
+ | Accuracy, trained sources (dev) | 0.824 | 0.794 | 0.845 |
19
+ | Accuracy, new sources (dev) | 0.602 | 0.569 | 0.857 |
20
+ | Log loss (dev) | 0.499 | 0.522 | |
21
+ | Calibration error, ECE (dev) | 0.057 | 0.040 | |
22
+ | Where it runs | your machine (Apple GPU / NVIDIA), free, private | | TypeSafe's cloud API |
23
 
24
+ Measured on the same frozen evaluation sets (development partitions; the locked test was not read). Jev is more accurate today. The second epoch gained 3 points on both sets at a slightly higher calibration error. Tag `v0.1-1epoch` keeps the previous version; the Apple Silicon build [JohnP1/d1a-e2b-mlx-q8](https://huggingface.co/JohnP1/d1a-e2b-mlx-q8) is still made from v0.1.
25
 
26
  - Playground (nine live use cases): https://github.com/jonpol01/d1a-playground
27
  - Code: https://github.com/jonpol01/d1a
 
29
  ## Serve it
30
 
31
  ```bash
32
+ pip install "d1a[serve] @ git+https://github.com/jonpol01/d1a"
33
+ python -m d1a.serve --run JohnP1/d1a-e2b --port 8009
34
  ```
35
 
36
  ## License
adapter_config.json CHANGED
@@ -32,13 +32,13 @@
32
  "rank_pattern": {},
33
  "revision": null,
34
  "target_modules": [
35
- "up_proj",
36
- "q_proj",
37
- "k_proj",
38
- "gate_proj",
39
  "down_proj",
 
 
40
  "v_proj",
41
- "o_proj"
 
 
42
  ],
43
  "target_parameters": null,
44
  "task_type": "FEATURE_EXTRACTION",
 
32
  "rank_pattern": {},
33
  "revision": null,
34
  "target_modules": [
 
 
 
 
35
  "down_proj",
36
+ "o_proj",
37
+ "q_proj",
38
  "v_proj",
39
+ "up_proj",
40
+ "gate_proj",
41
+ "k_proj"
42
  ],
43
  "target_parameters": null,
44
  "task_type": "FEATURE_EXTRACTION",
adapter_model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:93a489ba6fd4eddb766e9f9bbd70fbec04babe33c7ecd3925f5f9973e325410a
3
  size 96684752
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e23dbe51453b6e926da0db70a38d680bc64644caff872cf0defcbbd8b1acbe46
3
  size 96684752
head.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:20353fb397f0e4383f002a921eb1dd431b22b3ad6782ba4b99c74edab5d13429
3
- size 3152383
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dd2d8c5b1ac8bee3f413cd89802d93826674baf181dd4649004868b358605785
3
+ size 3151231
provenance.json CHANGED
@@ -1,18 +1,33 @@
1
  {
2
  "config": {
3
- "base": "google/gemma-4-E2B",
4
- "base_revision": "d29ff6b45f081a49ee2733a859c9c9c2d95d1a6f",
5
  "seed": 0,
6
- "epochs": 1,
7
  "lr": 0.0001,
8
- "batch": 4,
9
  "accum": 2,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
10
  "dtype": "bf16",
11
  "weights_dtype": "bf16",
12
- "checkpointing": 1,
13
- "p_none_pair": 0.25
14
  },
15
- "config_sha256": "355241763405ebfaa97b051898d14a98695c5ef0ed47b6877082fca47adf1506",
16
  "suite_sha256": "a8f50e481b7d90b97da049e0ff6a01cee2f1ed204aed61a8265af0edbb5514d2",
17
  "source_hashes": {
18
  "kev/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
@@ -46,51 +61,16 @@
46
  "uv.lock": "a9922dbb89acdef78299fd2b4a8c3f7f0fa1b2bc08b55595b6926fa785a9c466"
47
  },
48
  "git_commit": "ccd6e8b646068a6689792d76e3cae3698e9a015a",
49
- "platform": "HF Jobs l4x1 (job 6ab426ec52d0dbd7f1d86bac)",
50
  "torch": "2.8.0+cu128",
51
  "device": "cuda",
52
  "gpu": "NVIDIA L4",
53
  "legacy_checkpoint": false,
54
  "measured_checkpoint": {
55
- "requested": "runs/g4-e2b-e1-fp32/00-trial-0/checkpoint",
56
- "resolved": "runs/g4-e2b-e1-fp32/00-trial-0/checkpoint",
57
- "head_sha256": "80bcdddf6f671e9901e3805f8c311492c93d86ac63af0440d519b2d45bb90361",
58
- "adapter_sha256": "93a489ba6fd4eddb766e9f9bbd70fbec04babe33c7ecd3925f5f9973e325410a",
59
  "inference_temperature": 1.0
60
- },
61
- "eval_device": "mps",
62
- "resumed_evaluation": true,
63
- "resumed_source_hashes": {
64
- "kev/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
65
- "kev/anchors.py": "6963eafdb276db5a6c94d939eae675c761555553b8d0f963ffd803448beaabbb",
66
- "kev/api.py": "7bffacfb762c626b8bc2f670f350295af5ccb0883dbe90239c7d2f8e5ef56582",
67
- "kev/autoresearch.py": "f7d7fabb9c0df7065bee3fec4aa4bee7028c5d1565847aefe2d74e3d7d40ed8a",
68
- "kev/benchmark.py": "9c442efd8b6b852c574a646192c61ee91afa1306e3b2978bb6391ae1b91a3fa0",
69
- "kev/calibrate.py": "c1eff4744fabd349e8abca86777a7aa0cbea44904c8223d3b61fca3d43741519",
70
- "kev/checkpoint.py": "9f1dbc5d3d7d3050aca4ee34e0578c9ad76b7a2a7467dfc7a4ad65e3c471db59",
71
- "kev/compare.py": "3606dbf98bb305d66420158498cb837304d2edb8dab01cdaf2ec42964bef3435",
72
- "kev/composition.py": "f335ed17e18e0a544893db5e22b9a059e6ce1b2e14dbb863ac7d7bbf8f3e0536",
73
- "kev/contrastive.py": "cbb979aa5d40265ad0e64695f94b281d91751fa405811ddfa8212fede111edcf",
74
- "kev/data.py": "ef1be396e7218ba5c94f0f67de2ceff23a39656b57199cb3aa3192cde53e8320",
75
- "kev/device.py": "d1677fd98ec0979c7284546306e34e0d09298ca4042fc39eb2b32a74f4e975c4",
76
- "kev/evaluate.py": "999264f2837dcfbf2ec93601aa4745e701698ab850a43ad89324672a6604965c",
77
- "kev/experiment.py": "d12ab7f0765de0f29364f3411b99f06bb80b1cbf120ca9ed2b03316ec766e828",
78
- "kev/jev.py": "acd4cc3f1844e438cc83a8d409c15ef78a5ef64d39ea10f583d75a6c7646b243",
79
- "kev/metrics.py": "dba8d90550999edcd642581fb4da0367bd4b39d08471815184fa732d11e137b9",
80
- "kev/mlx_model.py": "f582428796faf6bf962b772a69a6227ac3259ccbd3215c0c8b55ce28dbbd0909",
81
- "kev/model.py": "c2129026ab56395f78c7d3071410541d5423dd7d97f7371518570572f206c376",
82
- "kev/plot.py": "0874bcff2885d8155a1cceade0de8a275a7163d3c3fd6aa0c294e8ccf5ff7e02",
83
- "kev/predictors.py": "b2a66dd9f0f0a8026bc94603383bacfdd622e75d1c935173997175606626a9b3",
84
- "kev/publish.py": "8abc9bcf4a01365697b05cf3c5ad0013462bd92d005f5616954e40ab1100c7aa",
85
- "kev/serve.py": "c210b1f4598e64b49c399f73603ed372f4314c165ceda50e1340c0466ec1c735",
86
- "kev/study_v3.py": "7891150aa623e4479c7c789d4dc4662f186132b37b7bac1e1f072b9e08ee13e3",
87
- "kev/suite.py": "d18d5f4f02d079344d61c2a31882c524c90012ebfdb660bf9160e145f37e3d42",
88
- "kev/train.py": "e103e99d3a68ab99eb58f135320fbc46b18514fcc2b195d1c6879633836e99b9",
89
- "kev/transfer_v9.py": "0902409742151250a28af1fd8f42258b70fcc161f3deed3ee35fe3f473f3c763",
90
- "modal_app.py": "7a13e9d29b3fc68f8236a0c1e4d67079b40a2a328a296f7ced1022817dff82aa",
91
- "pyproject.toml": "7c17fbe9efcadda3eb488b59adf1b6db5938cd33af4d3bd389ce4359b73fcc7d",
92
- "uv.lock": "a9922dbb89acdef78299fd2b4a8c3f7f0fa1b2bc08b55595b6926fa785a9c466"
93
- },
94
- "resumed_with_changed_non_evaluator_files": [],
95
- "resumed_git_commit": "ccd6e8b646068a6689792d76e3cae3698e9a015a"
96
  }
 
1
  {
2
  "config": {
3
+ "epochs": 2,
 
4
  "seed": 0,
 
5
  "lr": 0.0001,
6
+ "lora": 16,
7
  "accum": 2,
8
+ "batch": 4,
9
+ "perm_kl": 0.0,
10
+ "perm_frac": 0.3,
11
+ "ord_w": 0.0,
12
+ "p_none": 0.1,
13
+ "p_none_distract": 0.12,
14
+ "p_distract": 0.15,
15
+ "p_none_pair": 0.25,
16
+ "synthetic_repeat": 1,
17
+ "public_frac": 1.0,
18
+ "head_lr": 0.0,
19
+ "weight_decay": 0.01,
20
+ "anchor_w": 0.0,
21
+ "label_smoothing": 0.0,
22
+ "brier_w": 0.0,
23
+ "focal_gamma": 0.0,
24
+ "base": "google/gemma-4-E2B",
25
+ "base_revision": "d29ff6b45f081a49ee2733a859c9c9c2d95d1a6f",
26
  "dtype": "bf16",
27
  "weights_dtype": "bf16",
28
+ "checkpointing": 1
 
29
  },
30
+ "config_sha256": "cb5576f6ebf3fffe10370041387b4b6810c03c2b112b925168c16ded6729c34c",
31
  "suite_sha256": "a8f50e481b7d90b97da049e0ff6a01cee2f1ed204aed61a8265af0edbb5514d2",
32
  "source_hashes": {
33
  "kev/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
 
61
  "uv.lock": "a9922dbb89acdef78299fd2b4a8c3f7f0fa1b2bc08b55595b6926fa785a9c466"
62
  },
63
  "git_commit": "ccd6e8b646068a6689792d76e3cae3698e9a015a",
64
+ "platform": "Linux-6.12.103-127.188.amzn2023.x86_64-x86_64-with-glibc2.36",
65
  "torch": "2.8.0+cu128",
66
  "device": "cuda",
67
  "gpu": "NVIDIA L4",
68
  "legacy_checkpoint": false,
69
  "measured_checkpoint": {
70
+ "requested": "/kev/runs/g4-e2b-e2-l4/00-trial-0/checkpoint",
71
+ "resolved": "/kev/runs/g4-e2b-e2-l4/00-trial-0/checkpoint",
72
+ "head_sha256": "dd2d8c5b1ac8bee3f413cd89802d93826674baf181dd4649004868b358605785",
73
+ "adapter_sha256": "e23dbe51453b6e926da0db70a38d680bc64644caff872cf0defcbbd8b1acbe46",
74
  "inference_temperature": 1.0
75
+ }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
76
  }
result.json CHANGED
The diff for this file is too large to render. See raw diff
 
train.log CHANGED
@@ -1,194 +1,399 @@
1
 
2
-
3
-
4
-
5
-
6
-
7
-
8
-
9
-
10
-
11
-
12
-
13
-
14
-
15
-
16
-
17
-
18
-
19
-
20
-
21
-
22
-
23
-
24
-
25
-
26
-
27
-
28
-
29
-
30
-
31
- device=cuda trainable params=24.9M
32
 
33
-
34
-
35
- fetched v7/decision-v7/train.jsonl from jaredpalmer/kev-suites@a88f56db53
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
36
  dropped 70 of 12576 records that exceed the training context (384 state / 960 branch / 2048 packed tokens)
37
  12506 training requests (holdout=[]), questions by type {'score': 3448, 'noul': 5224, 'choice': 6834}
38
- ep0 step 10/1564 loss 2.494 kl 0.000 anchor 0.000 0.444s/rec
39
- ep0 step 20/1564 loss 2.089 kl 0.000 anchor 0.000 0.383s/rec
40
- ep0 step 30/1564 loss 1.886 kl 0.000 anchor 0.000 0.358s/rec
41
- ep0 step 40/1564 loss 2.217 kl 0.000 anchor 0.000 0.383s/rec
42
- ep0 step 50/1564 loss 1.778 kl 0.000 anchor 0.000 0.382s/rec
43
- ep0 step 60/1564 loss 1.879 kl 0.000 anchor 0.000 0.377s/rec
44
- ep0 step 70/1564 loss 1.540 kl 0.000 anchor 0.000 0.356s/rec
45
- ep0 step 80/1564 loss 1.700 kl 0.000 anchor 0.000 0.363s/rec
46
- ep0 step 90/1564 loss 1.608 kl 0.000 anchor 0.000 0.357s/rec
47
- ep0 step 100/1564 loss 1.627 kl 0.000 anchor 0.000 0.355s/rec
48
- ep0 step 110/1564 loss 1.473 kl 0.000 anchor 0.000 0.348s/rec
49
- ep0 step 120/1564 loss 1.457 kl 0.000 anchor 0.000 0.347s/rec
50
- ep0 step 130/1564 loss 1.059 kl 0.000 anchor 0.000 0.350s/rec
51
- ep0 step 140/1564 loss 1.110 kl 0.000 anchor 0.000 0.352s/rec
52
- ep0 step 150/1564 loss 1.593 kl 0.000 anchor 0.000 0.350s/rec
53
- ep0 step 160/1564 loss 0.870 kl 0.000 anchor 0.000 0.346s/rec
54
- ep0 step 170/1564 loss 1.391 kl 0.000 anchor 0.000 0.346s/rec
55
- ep0 step 180/1564 loss 1.205 kl 0.000 anchor 0.000 0.349s/rec
56
- ep0 step 190/1564 loss 0.786 kl 0.000 anchor 0.000 0.345s/rec
57
- ep0 step 200/1564 loss 0.964 kl 0.000 anchor 0.000 0.338s/rec
58
- ep0 step 210/1564 loss 1.072 kl 0.000 anchor 0.000 0.334s/rec
59
- ep0 step 220/1564 loss 0.788 kl 0.000 anchor 0.000 0.338s/rec
60
- ep0 step 230/1564 loss 1.440 kl 0.000 anchor 0.000 0.344s/rec
61
- ep0 step 240/1564 loss 1.064 kl 0.000 anchor 0.000 0.342s/rec
62
- ep0 step 250/1564 loss 0.909 kl 0.000 anchor 0.000 0.341s/rec
63
- ep0 step 260/1564 loss 0.880 kl 0.000 anchor 0.000 0.341s/rec
64
- ep0 step 270/1564 loss 0.853 kl 0.000 anchor 0.000 0.342s/rec
65
- ep0 step 280/1564 loss 0.939 kl 0.000 anchor 0.000 0.344s/rec
66
- ep0 step 290/1564 loss 0.841 kl 0.000 anchor 0.000 0.343s/rec
67
- ep0 step 300/1564 loss 0.791 kl 0.000 anchor 0.000 0.345s/rec
68
- ep0 step 310/1564 loss 1.095 kl 0.000 anchor 0.000 0.346s/rec
69
- ep0 step 320/1564 loss 1.164 kl 0.000 anchor 0.000 0.348s/rec
70
- ep0 step 330/1564 loss 0.823 kl 0.000 anchor 0.000 0.344s/rec
71
- ep0 step 340/1564 loss 0.710 kl 0.000 anchor 0.000 0.343s/rec
72
- ep0 step 350/1564 loss 0.876 kl 0.000 anchor 0.000 0.343s/rec
73
- ep0 step 360/1564 loss 0.983 kl 0.000 anchor 0.000 0.341s/rec
74
- ep0 step 370/1564 loss 0.673 kl 0.000 anchor 0.000 0.337s/rec
75
- ep0 step 380/1564 loss 1.054 kl 0.000 anchor 0.000 0.335s/rec
76
- ep0 step 390/1564 loss 0.603 kl 0.000 anchor 0.000 0.335s/rec
77
- ep0 step 400/1564 loss 0.934 kl 0.000 anchor 0.000 0.334s/rec
78
- ep0 step 410/1564 loss 0.793 kl 0.000 anchor 0.000 0.332s/rec
79
- ep0 step 420/1564 loss 0.735 kl 0.000 anchor 0.000 0.331s/rec
80
- ep0 step 430/1564 loss 0.823 kl 0.000 anchor 0.000 0.334s/rec
81
- ep0 step 440/1564 loss 0.741 kl 0.000 anchor 0.000 0.334s/rec
82
- ep0 step 450/1564 loss 0.651 kl 0.000 anchor 0.000 0.334s/rec
83
- ep0 step 460/1564 loss 0.821 kl 0.000 anchor 0.000 0.333s/rec
84
- ep0 step 470/1564 loss 0.634 kl 0.000 anchor 0.000 0.331s/rec
85
- ep0 step 480/1564 loss 0.620 kl 0.000 anchor 0.000 0.330s/rec
86
- ep0 step 490/1564 loss 0.596 kl 0.000 anchor 0.000 0.331s/rec
87
- ep0 step 500/1564 loss 1.017 kl 0.000 anchor 0.000 0.328s/rec
88
- ep0 step 510/1564 loss 0.903 kl 0.000 anchor 0.000 0.328s/rec
89
- ep0 step 520/1564 loss 0.672 kl 0.000 anchor 0.000 0.327s/rec
90
- ep0 step 530/1564 loss 0.966 kl 0.000 anchor 0.000 0.326s/rec
91
- ep0 step 540/1564 loss 0.609 kl 0.000 anchor 0.000 0.324s/rec
92
- ep0 step 550/1564 loss 0.519 kl 0.000 anchor 0.000 0.324s/rec
93
- ep0 step 560/1564 loss 0.658 kl 0.000 anchor 0.000 0.322s/rec
94
- ep0 step 570/1564 loss 0.776 kl 0.000 anchor 0.000 0.322s/rec
95
- ep0 step 580/1564 loss 0.752 kl 0.000 anchor 0.000 0.321s/rec
96
- ep0 step 590/1564 loss 0.576 kl 0.000 anchor 0.000 0.321s/rec
97
- ep0 step 600/1564 loss 0.713 kl 0.000 anchor 0.000 0.321s/rec
98
- ep0 step 610/1564 loss 0.660 kl 0.000 anchor 0.000 0.320s/rec
99
- ep0 step 620/1564 loss 0.724 kl 0.000 anchor 0.000 0.321s/rec
100
- ep0 step 630/1564 loss 0.746 kl 0.000 anchor 0.000 0.323s/rec
101
- ep0 step 640/1564 loss 0.483 kl 0.000 anchor 0.000 0.323s/rec
102
- ep0 step 650/1564 loss 0.813 kl 0.000 anchor 0.000 0.323s/rec
103
- ep0 step 660/1564 loss 0.664 kl 0.000 anchor 0.000 0.324s/rec
104
- ep0 step 670/1564 loss 0.578 kl 0.000 anchor 0.000 0.324s/rec
105
- ep0 step 680/1564 loss 0.602 kl 0.000 anchor 0.000 0.325s/rec
106
- ep0 step 690/1564 loss 0.622 kl 0.000 anchor 0.000 0.324s/rec
107
- ep0 step 700/1564 loss 0.738 kl 0.000 anchor 0.000 0.325s/rec
108
- ep0 step 710/1564 loss 0.598 kl 0.000 anchor 0.000 0.326s/rec
109
- ep0 step 720/1564 loss 0.842 kl 0.000 anchor 0.000 0.326s/rec
110
- ep0 step 730/1564 loss 0.680 kl 0.000 anchor 0.000 0.329s/rec
111
- ep0 step 740/1564 loss 0.602 kl 0.000 anchor 0.000 0.330s/rec
112
- ep0 step 750/1564 loss 0.381 kl 0.000 anchor 0.000 0.329s/rec
113
- ep0 step 760/1564 loss 0.818 kl 0.000 anchor 0.000 0.329s/rec
114
- ep0 step 770/1564 loss 0.586 kl 0.000 anchor 0.000 0.328s/rec
115
- ep0 step 780/1564 loss 0.696 kl 0.000 anchor 0.000 0.327s/rec
116
- ep0 step 790/1564 loss 0.460 kl 0.000 anchor 0.000 0.327s/rec
117
- ep0 step 800/1564 loss 0.778 kl 0.000 anchor 0.000 0.329s/rec
118
- ep0 step 810/1564 loss 0.440 kl 0.000 anchor 0.000 0.328s/rec
119
- ep0 step 820/1564 loss 0.575 kl 0.000 anchor 0.000 0.329s/rec
120
- ep0 step 830/1564 loss 0.609 kl 0.000 anchor 0.000 0.328s/rec
121
- ep0 step 840/1564 loss 0.666 kl 0.000 anchor 0.000 0.329s/rec
122
- ep0 step 850/1564 loss 0.605 kl 0.000 anchor 0.000 0.328s/rec
123
- ep0 step 860/1564 loss 0.432 kl 0.000 anchor 0.000 0.327s/rec
124
- ep0 step 870/1564 loss 0.615 kl 0.000 anchor 0.000 0.327s/rec
125
- ep0 step 880/1564 loss 0.750 kl 0.000 anchor 0.000 0.328s/rec
126
- ep0 step 890/1564 loss 0.348 kl 0.000 anchor 0.000 0.328s/rec
127
- ep0 step 900/1564 loss 0.652 kl 0.000 anchor 0.000 0.328s/rec
128
- ep0 step 910/1564 loss 0.504 kl 0.000 anchor 0.000 0.328s/rec
129
- ep0 step 920/1564 loss 0.657 kl 0.000 anchor 0.000 0.328s/rec
130
- ep0 step 930/1564 loss 0.748 kl 0.000 anchor 0.000 0.329s/rec
131
- ep0 step 940/1564 loss 0.553 kl 0.000 anchor 0.000 0.330s/rec
132
- ep0 step 950/1564 loss 0.634 kl 0.000 anchor 0.000 0.328s/rec
133
- ep0 step 960/1564 loss 0.499 kl 0.000 anchor 0.000 0.328s/rec
134
- ep0 step 970/1564 loss 0.509 kl 0.000 anchor 0.000 0.327s/rec
135
- ep0 step 980/1564 loss 0.697 kl 0.000 anchor 0.000 0.326s/rec
136
- ep0 step 990/1564 loss 0.688 kl 0.000 anchor 0.000 0.325s/rec
137
- ep0 step 1000/1564 loss 0.660 kl 0.000 anchor 0.000 0.325s/rec
138
- ep0 step 1010/1564 loss 0.695 kl 0.000 anchor 0.000 0.326s/rec
139
- ep0 step 1020/1564 loss 0.691 kl 0.000 anchor 0.000 0.327s/rec
140
- ep0 step 1030/1564 loss 0.569 kl 0.000 anchor 0.000 0.326s/rec
141
- ep0 step 1040/1564 loss 0.532 kl 0.000 anchor 0.000 0.325s/rec
142
- ep0 step 1050/1564 loss 0.553 kl 0.000 anchor 0.000 0.324s/rec
143
- ep0 step 1060/1564 loss 0.766 kl 0.000 anchor 0.000 0.324s/rec
144
- ep0 step 1070/1564 loss 0.463 kl 0.000 anchor 0.000 0.323s/rec
145
- ep0 step 1080/1564 loss 0.436 kl 0.000 anchor 0.000 0.324s/rec
146
- ep0 step 1090/1564 loss 0.715 kl 0.000 anchor 0.000 0.323s/rec
147
- ep0 step 1100/1564 loss 0.501 kl 0.000 anchor 0.000 0.322s/rec
148
- ep0 step 1110/1564 loss 0.495 kl 0.000 anchor 0.000 0.322s/rec
149
- ep0 step 1120/1564 loss 0.607 kl 0.000 anchor 0.000 0.322s/rec
150
- ep0 step 1130/1564 loss 0.642 kl 0.000 anchor 0.000 0.323s/rec
151
- ep0 step 1140/1564 loss 0.716 kl 0.000 anchor 0.000 0.323s/rec
152
- ep0 step 1150/1564 loss 0.327 kl 0.000 anchor 0.000 0.323s/rec
153
- ep0 step 1160/1564 loss 0.847 kl 0.000 anchor 0.000 0.323s/rec
154
- ep0 step 1170/1564 loss 0.885 kl 0.000 anchor 0.000 0.324s/rec
155
- ep0 step 1180/1564 loss 0.411 kl 0.000 anchor 0.000 0.323s/rec
156
- ep0 step 1190/1564 loss 0.498 kl 0.000 anchor 0.000 0.323s/rec
157
- ep0 step 1200/1564 loss 0.659 kl 0.000 anchor 0.000 0.323s/rec
158
- ep0 step 1210/1564 loss 0.316 kl 0.000 anchor 0.000 0.324s/rec
159
- ep0 step 1220/1564 loss 0.395 kl 0.000 anchor 0.000 0.324s/rec
160
- ep0 step 1230/1564 loss 0.505 kl 0.000 anchor 0.000 0.324s/rec
161
- ep0 step 1240/1564 loss 0.628 kl 0.000 anchor 0.000 0.324s/rec
162
- ep0 step 1250/1564 loss 0.645 kl 0.000 anchor 0.000 0.325s/rec
163
- ep0 step 1260/1564 loss 0.628 kl 0.000 anchor 0.000 0.325s/rec
164
- ep0 step 1270/1564 loss 0.576 kl 0.000 anchor 0.000 0.325s/rec
165
- ep0 step 1280/1564 loss 0.712 kl 0.000 anchor 0.000 0.326s/rec
166
- ep0 step 1290/1564 loss 0.604 kl 0.000 anchor 0.000 0.325s/rec
167
- ep0 step 1300/1564 loss 0.398 kl 0.000 anchor 0.000 0.324s/rec
168
- ep0 step 1310/1564 loss 0.543 kl 0.000 anchor 0.000 0.324s/rec
169
- ep0 step 1320/1564 loss 0.427 kl 0.000 anchor 0.000 0.324s/rec
170
- ep0 step 1330/1564 loss 0.734 kl 0.000 anchor 0.000 0.324s/rec
171
- ep0 step 1340/1564 loss 0.468 kl 0.000 anchor 0.000 0.324s/rec
172
- ep0 step 1350/1564 loss 0.377 kl 0.000 anchor 0.000 0.324s/rec
173
- ep0 step 1360/1564 loss 0.454 kl 0.000 anchor 0.000 0.325s/rec
174
- ep0 step 1370/1564 loss 0.639 kl 0.000 anchor 0.000 0.324s/rec
175
- ep0 step 1380/1564 loss 0.494 kl 0.000 anchor 0.000 0.324s/rec
176
- ep0 step 1390/1564 loss 0.553 kl 0.000 anchor 0.000 0.324s/rec
177
- ep0 step 1400/1564 loss 0.371 kl 0.000 anchor 0.000 0.323s/rec
178
- ep0 step 1410/1564 loss 0.453 kl 0.000 anchor 0.000 0.323s/rec
179
- ep0 step 1420/1564 loss 0.393 kl 0.000 anchor 0.000 0.322s/rec
180
- ep0 step 1430/1564 loss 0.590 kl 0.000 anchor 0.000 0.322s/rec
181
- ep0 step 1440/1564 loss 0.457 kl 0.000 anchor 0.000 0.322s/rec
182
- ep0 step 1450/1564 loss 0.762 kl 0.000 anchor 0.000 0.322s/rec
183
- ep0 step 1460/1564 loss 0.683 kl 0.000 anchor 0.000 0.321s/rec
184
- ep0 step 1470/1564 loss 0.459 kl 0.000 anchor 0.000 0.321s/rec
185
- ep0 step 1480/1564 loss 0.696 kl 0.000 anchor 0.000 0.321s/rec
186
- ep0 step 1490/1564 loss 0.246 kl 0.000 anchor 0.000 0.321s/rec
187
- ep0 step 1500/1564 loss 0.592 kl 0.000 anchor 0.000 0.320s/rec
188
- ep0 step 1510/1564 loss 0.554 kl 0.000 anchor 0.000 0.320s/rec
189
- ep0 step 1520/1564 loss 0.434 kl 0.000 anchor 0.000 0.320s/rec
190
- ep0 step 1530/1564 loss 0.455 kl 0.000 anchor 0.000 0.320s/rec
191
- ep0 step 1540/1564 loss 0.457 kl 0.000 anchor 0.000 0.320s/rec
192
- ep0 step 1550/1564 loss 0.505 kl 0.000 anchor 0.000 0.321s/rec
193
- ep0 step 1560/1564 loss 0.613 kl 0.000 anchor 0.000 0.322s/rec
194
- saved runs/g4-e2b-e1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2
 
3
+ config.json: 0%| | 0.00/4.91k [00:00<?, ?B/s]
4
+ config.json: 100%|██████████| 4.91k/4.91k [00:00<00:00, 27.8MB/s]
5
+
6
+
7
+ tokenizer_config.json: 0%| | 0.00/906 [00:00<?, ?B/s]
8
+ tokenizer_config.json: 100%|██████████| 906/906 [00:00<00:00, 5.99MB/s]
9
+
10
+
11
+ tokenizer.json: reconstructing file: 0%| | 0.00B / 32.2MB 
12
+ tokenizer.json: downloading bytes: | 0.00B
13
+ tokenizer.json: downloading bytes: ██████████| 8.75MB, 872kB/s
14
+ tokenizer.json: downloading bytes: ██████████| 8.75MB, 872kB/s
15
+
16
+ tokenizer.json: reconstructing file: 100%|██████████| 32.2MB / 32.2MB, 3.21MB/s
17
+
18
+
19
+ model.safetensors: reconstructing file: 0%| | 0.00B / 10.2GB 
20
+ model.safetensors: downloading bytes: | 0.00B
21
+
22
+ model.safetensors: reconstructing file: 3%|▎ | 302MB / 10.2GB, 21.0MB/s 
23
+ model.safetensors: downloading bytes: ▏ | 159MB, 12.7MB/s
24
+
25
+ model.safetensors: reconstructing file: 6%|▌ | 610MB / 10.2GB, 50.1MB/s 
26
+ model.safetensors: downloading bytes: ▍ | 387MB, 28.2MB/s
27
+
28
+ model.safetensors: reconstructing file: 10%|█ | 1.05GB / 10.2GB, 66.2MB/s 
29
+ model.safetensors: downloading bytes: ▉ | 926MB, 76.8MB/s
30
+
31
+ model.safetensors: reconstructing file: 17%|█▋ | 1.78GB / 10.2GB, 142MB/s 
32
+ model.safetensors: downloading bytes: █▋ | 1.69GB, 134MB/s
33
+ model.safetensors: downloading bytes: ██▍ | 2.53GB, 193MB/s
34
+
35
+ model.safetensors: reconstructing file: 27%|██▋ | 2.79GB / 10.2GB, 192MB/s 
36
+ model.safetensors: downloading bytes: ███▎ | 3.46GB, 254MB/s
37
+
38
+ model.safetensors: reconstructing file: 38%|███▊ | 3.93GB / 10.2GB, 237MB/s 
39
+ model.safetensors: downloading bytes: ████▍ | 4.58GB, 324MB/s
40
+ model.safetensors: downloading bytes: █████▎ | 5.49GB, 375MB/s
41
+
42
+ model.safetensors: reconstructing file: 51%|█████▏ | 5.27GB / 10.2GB, 309MB/s 
43
+ model.safetensors: downloading bytes: ██████▏ | 6.37GB, 329MB/s
44
+
45
+ model.safetensors: reconstructing file: 59%|█████▉ | 6.08GB / 10.2GB, 290MB/s 
46
+ model.safetensors: downloading bytes: ██████▊ | 7.04GB, 298MB/s
47
+
48
+ model.safetensors: reconstructing file: 66%|██████▌ | 6.76GB / 10.2GB, 271MB/s 
49
+ model.safetensors: downloading bytes: ███████▍ | 7.63GB, 277MB/s
50
+
51
+ model.safetensors: reconstructing file: 70%|██████▉ | 7.16GB / 10.2GB, 262MB/s 
52
+ model.safetensors: downloading bytes: ███████▉ | 8.10GB, 277MB/s
53
+
54
+ model.safetensors: reconstructing file: 74%|███████▍ | 7.62GB / 10.2GB, 255MB/s 
55
+ model.safetensors: downloading bytes: ████████▎ | 8.55GB, 261MB/s
56
+
57
+ model.safetensors: reconstructing file: 78%|███████▊ | 8.03GB / 10.2GB, 250MB/s 
58
+
59
+ model.safetensors: reconstructing file: 82%|████████▏ | 8.42GB / 10.2GB, 246MB/s 
60
+
61
+ model.safetensors: reconstructing file: 85%|████████▍ | 8.69GB / 10.2GB, 244MB/s 
62
+
63
+ model.safetensors: reconstructing file: 88%|████████▊ | 9.01GB / 10.2GB, 240MB/s 
64
+
65
+ model.safetensors: reconstructing file: 90%|█████████ | 9.26GB / 10.2GB, 239MB/s 
66
+
67
+ model.safetensors: reconstructing file: 94%|█████████▎| 9.58GB / 10.2GB, 237MB/s 
68
+
69
+ model.safetensors: reconstructing file: 96%|█████████▌| 9.86GB / 10.2GB, 235MB/s 
70
+
71
+ model.safetensors: reconstructing file: 98%|█████████▊| 10.1GB / 10.2GB, 235MB/s 
72
+ model.safetensors: downloading bytes: ██████████| 8.66GB, 261MB/s
73
+ model.safetensors: downloading bytes: ██████████| 8.66GB, 261MB/s
74
+
75
+ model.safetensors: reconstructing file: 100%|██████████| 10.2GB / 10.2GB, 235MB/s
76
+
77
+
78
+ Loading weights: 0%| | 0/1951 [00:00<?, ?it/s]
79
+ Loading weights: 100%|██████████| 1951/1951 [00:00<00:00, 5307.06it/s]
80
+
81
+
82
+ generation_config.json: 0%| | 0.00/181 [00:00<?, ?B/s]
83
+ generation_config.json: 100%|██████████| 181/181 [00:00<00:00, 1.09MB/s]
84
+ device=cuda trainable params=24.9M
85
  dropped 70 of 12576 records that exceed the training context (384 state / 960 branch / 2048 packed tokens)
86
  12506 training requests (holdout=[]), questions by type {'score': 3448, 'noul': 5224, 'choice': 6834}
87
+ ep0 step 10/3128 loss 2.506 kl 0.000 anchor 0.000 0.443s/rec
88
+ ep0 step 20/3128 loss 2.095 kl 0.000 anchor 0.000 0.380s/rec
89
+ ep0 step 30/3128 loss 1.953 kl 0.000 anchor 0.000 0.355s/rec
90
+ ep0 step 40/3128 loss 2.202 kl 0.000 anchor 0.000 0.379s/rec
91
+ ep0 step 50/3128 loss 1.863 kl 0.000 anchor 0.000 0.377s/rec
92
+ ep0 step 60/3128 loss 1.980 kl 0.000 anchor 0.000 0.373s/rec
93
+ ep0 step 70/3128 loss 1.595 kl 0.000 anchor 0.000 0.352s/rec
94
+ ep0 step 80/3128 loss 1.736 kl 0.000 anchor 0.000 0.360s/rec
95
+ ep0 step 90/3128 loss 1.607 kl 0.000 anchor 0.000 0.353s/rec
96
+ ep0 step 100/3128 loss 1.654 kl 0.000 anchor 0.000 0.351s/rec
97
+ ep0 step 110/3128 loss 1.521 kl 0.000 anchor 0.000 0.345s/rec
98
+ ep0 step 120/3128 loss 1.439 kl 0.000 anchor 0.000 0.344s/rec
99
+ ep0 step 130/3128 loss 1.560 kl 0.000 anchor 0.000 0.348s/rec
100
+ ep0 step 140/3128 loss 1.407 kl 0.000 anchor 0.000 0.349s/rec
101
+ ep0 step 150/3128 loss 1.608 kl 0.000 anchor 0.000 0.347s/rec
102
+ ep0 step 160/3128 loss 1.093 kl 0.000 anchor 0.000 0.343s/rec
103
+ ep0 step 170/3128 loss 1.345 kl 0.000 anchor 0.000 0.344s/rec
104
+ ep0 step 180/3128 loss 1.338 kl 0.000 anchor 0.000 0.347s/rec
105
+ ep0 step 190/3128 loss 1.099 kl 0.000 anchor 0.000 0.343s/rec
106
+ ep0 step 200/3128 loss 1.322 kl 0.000 anchor 0.000 0.336s/rec
107
+ ep0 step 210/3128 loss 1.262 kl 0.000 anchor 0.000 0.332s/rec
108
+ ep0 step 220/3128 loss 1.285 kl 0.000 anchor 0.000 0.336s/rec
109
+ ep0 step 230/3128 loss 1.338 kl 0.000 anchor 0.000 0.342s/rec
110
+ ep0 step 240/3128 loss 1.044 kl 0.000 anchor 0.000 0.340s/rec
111
+ ep0 step 250/3128 loss 0.938 kl 0.000 anchor 0.000 0.339s/rec
112
+ ep0 step 260/3128 loss 1.045 kl 0.000 anchor 0.000 0.339s/rec
113
+ ep0 step 270/3128 loss 1.045 kl 0.000 anchor 0.000 0.340s/rec
114
+ ep0 step 280/3128 loss 1.016 kl 0.000 anchor 0.000 0.342s/rec
115
+ ep0 step 290/3128 loss 0.797 kl 0.000 anchor 0.000 0.341s/rec
116
+ ep0 step 300/3128 loss 1.012 kl 0.000 anchor 0.000 0.343s/rec
117
+ ep0 step 310/3128 loss 0.952 kl 0.000 anchor 0.000 0.345s/rec
118
+ ep0 step 320/3128 loss 1.101 kl 0.000 anchor 0.000 0.346s/rec
119
+ ep0 step 330/3128 loss 1.034 kl 0.000 anchor 0.000 0.343s/rec
120
+ ep0 step 340/3128 loss 0.711 kl 0.000 anchor 0.000 0.342s/rec
121
+ ep0 step 350/3128 loss 0.927 kl 0.000 anchor 0.000 0.342s/rec
122
+ ep0 step 360/3128 loss 1.065 kl 0.000 anchor 0.000 0.340s/rec
123
+ ep0 step 370/3128 loss 0.650 kl 0.000 anchor 0.000 0.335s/rec
124
+ ep0 step 380/3128 loss 1.009 kl 0.000 anchor 0.000 0.333s/rec
125
+ ep0 step 390/3128 loss 0.725 kl 0.000 anchor 0.000 0.333s/rec
126
+ ep0 step 400/3128 loss 0.935 kl 0.000 anchor 0.000 0.332s/rec
127
+ ep0 step 410/3128 loss 0.804 kl 0.000 anchor 0.000 0.330s/rec
128
+ ep0 step 420/3128 loss 0.770 kl 0.000 anchor 0.000 0.330s/rec
129
+ ep0 step 430/3128 loss 1.309 kl 0.000 anchor 0.000 0.332s/rec
130
+ ep0 step 440/3128 loss 0.784 kl 0.000 anchor 0.000 0.333s/rec
131
+ ep0 step 450/3128 loss 0.690 kl 0.000 anchor 0.000 0.333s/rec
132
+ ep0 step 460/3128 loss 0.826 kl 0.000 anchor 0.000 0.331s/rec
133
+ ep0 step 470/3128 loss 0.709 kl 0.000 anchor 0.000 0.330s/rec
134
+ ep0 step 480/3128 loss 0.733 kl 0.000 anchor 0.000 0.328s/rec
135
+ ep0 step 490/3128 loss 0.727 kl 0.000 anchor 0.000 0.330s/rec
136
+ ep0 step 500/3128 loss 0.804 kl 0.000 anchor 0.000 0.327s/rec
137
+ ep0 step 510/3128 loss 0.785 kl 0.000 anchor 0.000 0.327s/rec
138
+ ep0 step 520/3128 loss 0.532 kl 0.000 anchor 0.000 0.326s/rec
139
+ ep0 step 530/3128 loss 1.143 kl 0.000 anchor 0.000 0.325s/rec
140
+ ep0 step 540/3128 loss 0.629 kl 0.000 anchor 0.000 0.323s/rec
141
+ ep0 step 550/3128 loss 0.622 kl 0.000 anchor 0.000 0.323s/rec
142
+ ep0 step 560/3128 loss 0.718 kl 0.000 anchor 0.000 0.321s/rec
143
+ ep0 step 570/3128 loss 0.757 kl 0.000 anchor 0.000 0.321s/rec
144
+ ep0 step 580/3128 loss 0.903 kl 0.000 anchor 0.000 0.320s/rec
145
+ ep0 step 590/3128 loss 0.621 kl 0.000 anchor 0.000 0.320s/rec
146
+ ep0 step 600/3128 loss 0.820 kl 0.000 anchor 0.000 0.320s/rec
147
+ ep0 step 610/3128 loss 0.698 kl 0.000 anchor 0.000 0.319s/rec
148
+ ep0 step 620/3128 loss 0.643 kl 0.000 anchor 0.000 0.320s/rec
149
+ ep0 step 630/3128 loss 0.705 kl 0.000 anchor 0.000 0.322s/rec
150
+ ep0 step 640/3128 loss 0.523 kl 0.000 anchor 0.000 0.322s/rec
151
+ ep0 step 650/3128 loss 0.872 kl 0.000 anchor 0.000 0.322s/rec
152
+ ep0 step 660/3128 loss 0.779 kl 0.000 anchor 0.000 0.323s/rec
153
+ ep0 step 670/3128 loss 0.564 kl 0.000 anchor 0.000 0.323s/rec
154
+ ep0 step 680/3128 loss 0.648 kl 0.000 anchor 0.000 0.323s/rec
155
+ ep0 step 690/3128 loss 0.617 kl 0.000 anchor 0.000 0.323s/rec
156
+ ep0 step 700/3128 loss 0.789 kl 0.000 anchor 0.000 0.324s/rec
157
+ ep0 step 710/3128 loss 0.603 kl 0.000 anchor 0.000 0.325s/rec
158
+ ep0 step 720/3128 loss 0.971 kl 0.000 anchor 0.000 0.325s/rec
159
+ ep0 step 730/3128 loss 0.770 kl 0.000 anchor 0.000 0.328s/rec
160
+ ep0 step 740/3128 loss 0.568 kl 0.000 anchor 0.000 0.329s/rec
161
+ ep0 step 750/3128 loss 0.430 kl 0.000 anchor 0.000 0.328s/rec
162
+ ep0 step 760/3128 loss 0.736 kl 0.000 anchor 0.000 0.328s/rec
163
+ ep0 step 770/3128 loss 0.627 kl 0.000 anchor 0.000 0.327s/rec
164
+ ep0 step 780/3128 loss 0.821 kl 0.000 anchor 0.000 0.326s/rec
165
+ ep0 step 790/3128 loss 0.509 kl 0.000 anchor 0.000 0.326s/rec
166
+ ep0 step 800/3128 loss 0.867 kl 0.000 anchor 0.000 0.327s/rec
167
+ ep0 step 810/3128 loss 0.503 kl 0.000 anchor 0.000 0.327s/rec
168
+ ep0 step 820/3128 loss 0.779 kl 0.000 anchor 0.000 0.328s/rec
169
+ ep0 step 830/3128 loss 0.676 kl 0.000 anchor 0.000 0.327s/rec
170
+ ep0 step 840/3128 loss 0.670 kl 0.000 anchor 0.000 0.327s/rec
171
+ ep0 step 850/3128 loss 0.716 kl 0.000 anchor 0.000 0.327s/rec
172
+ ep0 step 860/3128 loss 0.394 kl 0.000 anchor 0.000 0.326s/rec
173
+ ep0 step 870/3128 loss 0.826 kl 0.000 anchor 0.000 0.325s/rec
174
+ ep0 step 880/3128 loss 0.819 kl 0.000 anchor 0.000 0.327s/rec
175
+ ep0 step 890/3128 loss 0.446 kl 0.000 anchor 0.000 0.327s/rec
176
+ ep0 step 900/3128 loss 0.698 kl 0.000 anchor 0.000 0.327s/rec
177
+ ep0 step 910/3128 loss 0.585 kl 0.000 anchor 0.000 0.327s/rec
178
+ ep0 step 920/3128 loss 0.767 kl 0.000 anchor 0.000 0.327s/rec
179
+ ep0 step 930/3128 loss 0.761 kl 0.000 anchor 0.000 0.328s/rec
180
+ ep0 step 940/3128 loss 0.635 kl 0.000 anchor 0.000 0.328s/rec
181
+ ep0 step 950/3128 loss 0.702 kl 0.000 anchor 0.000 0.327s/rec
182
+ ep0 step 960/3128 loss 0.477 kl 0.000 anchor 0.000 0.326s/rec
183
+ ep0 step 970/3128 loss 0.663 kl 0.000 anchor 0.000 0.325s/rec
184
+ ep0 step 980/3128 loss 0.740 kl 0.000 anchor 0.000 0.324s/rec
185
+ ep0 step 990/3128 loss 0.646 kl 0.000 anchor 0.000 0.323s/rec
186
+ ep0 step 1000/3128 loss 0.598 kl 0.000 anchor 0.000 0.323s/rec
187
+ ep0 step 1010/3128 loss 0.645 kl 0.000 anchor 0.000 0.324s/rec
188
+ ep0 step 1020/3128 loss 0.749 kl 0.000 anchor 0.000 0.325s/rec
189
+ ep0 step 1030/3128 loss 0.760 kl 0.000 anchor 0.000 0.324s/rec
190
+ ep0 step 1040/3128 loss 0.591 kl 0.000 anchor 0.000 0.324s/rec
191
+ ep0 step 1050/3128 loss 0.685 kl 0.000 anchor 0.000 0.323s/rec
192
+ ep0 step 1060/3128 loss 0.735 kl 0.000 anchor 0.000 0.322s/rec
193
+ ep0 step 1070/3128 loss 0.524 kl 0.000 anchor 0.000 0.321s/rec
194
+ ep0 step 1080/3128 loss 0.463 kl 0.000 anchor 0.000 0.322s/rec
195
+ ep0 step 1090/3128 loss 0.852 kl 0.000 anchor 0.000 0.322s/rec
196
+ ep0 step 1100/3128 loss 0.567 kl 0.000 anchor 0.000 0.320s/rec
197
+ ep0 step 1110/3128 loss 0.523 kl 0.000 anchor 0.000 0.320s/rec
198
+ ep0 step 1120/3128 loss 0.743 kl 0.000 anchor 0.000 0.320s/rec
199
+ ep0 step 1130/3128 loss 0.754 kl 0.000 anchor 0.000 0.322s/rec
200
+ ep0 step 1140/3128 loss 0.769 kl 0.000 anchor 0.000 0.321s/rec
201
+ ep0 step 1150/3128 loss 0.452 kl 0.000 anchor 0.000 0.322s/rec
202
+ ep0 step 1160/3128 loss 0.827 kl 0.000 anchor 0.000 0.322s/rec
203
+ ep0 step 1170/3128 loss 0.836 kl 0.000 anchor 0.000 0.323s/rec
204
+ ep0 step 1180/3128 loss 0.521 kl 0.000 anchor 0.000 0.322s/rec
205
+ ep0 step 1190/3128 loss 0.517 kl 0.000 anchor 0.000 0.321s/rec
206
+ ep0 step 1200/3128 loss 0.767 kl 0.000 anchor 0.000 0.321s/rec
207
+ ep0 step 1210/3128 loss 0.430 kl 0.000 anchor 0.000 0.322s/rec
208
+ ep0 step 1220/3128 loss 0.527 kl 0.000 anchor 0.000 0.322s/rec
209
+ ep0 step 1230/3128 loss 0.552 kl 0.000 anchor 0.000 0.323s/rec
210
+ ep0 step 1240/3128 loss 0.676 kl 0.000 anchor 0.000 0.323s/rec
211
+ ep0 step 1250/3128 loss 0.701 kl 0.000 anchor 0.000 0.323s/rec
212
+ ep0 step 1260/3128 loss 0.643 kl 0.000 anchor 0.000 0.323s/rec
213
+ ep0 step 1270/3128 loss 0.650 kl 0.000 anchor 0.000 0.323s/rec
214
+ ep0 step 1280/3128 loss 0.888 kl 0.000 anchor 0.000 0.324s/rec
215
+ ep0 step 1290/3128 loss 0.596 kl 0.000 anchor 0.000 0.323s/rec
216
+ ep0 step 1300/3128 loss 0.449 kl 0.000 anchor 0.000 0.323s/rec
217
+ ep0 step 1310/3128 loss 0.714 kl 0.000 anchor 0.000 0.322s/rec
218
+ ep0 step 1320/3128 loss 0.392 kl 0.000 anchor 0.000 0.323s/rec
219
+ ep0 step 1330/3128 loss 0.667 kl 0.000 anchor 0.000 0.323s/rec
220
+ ep0 step 1340/3128 loss 0.519 kl 0.000 anchor 0.000 0.323s/rec
221
+ ep0 step 1350/3128 loss 0.407 kl 0.000 anchor 0.000 0.323s/rec
222
+ ep0 step 1360/3128 loss 0.536 kl 0.000 anchor 0.000 0.323s/rec
223
+ ep0 step 1370/3128 loss 0.681 kl 0.000 anchor 0.000 0.322s/rec
224
+ ep0 step 1380/3128 loss 0.558 kl 0.000 anchor 0.000 0.323s/rec
225
+ ep0 step 1390/3128 loss 0.540 kl 0.000 anchor 0.000 0.323s/rec
226
+ ep0 step 1400/3128 loss 0.410 kl 0.000 anchor 0.000 0.322s/rec
227
+ ep0 step 1410/3128 loss 0.467 kl 0.000 anchor 0.000 0.321s/rec
228
+ ep0 step 1420/3128 loss 0.475 kl 0.000 anchor 0.000 0.321s/rec
229
+ ep0 step 1430/3128 loss 0.598 kl 0.000 anchor 0.000 0.320s/rec
230
+ ep0 step 1440/3128 loss 0.492 kl 0.000 anchor 0.000 0.320s/rec
231
+ ep0 step 1450/3128 loss 0.755 kl 0.000 anchor 0.000 0.320s/rec
232
+ ep0 step 1460/3128 loss 0.819 kl 0.000 anchor 0.000 0.320s/rec
233
+ ep0 step 1470/3128 loss 0.506 kl 0.000 anchor 0.000 0.320s/rec
234
+ ep0 step 1480/3128 loss 0.685 kl 0.000 anchor 0.000 0.320s/rec
235
+ ep0 step 1490/3128 loss 0.317 kl 0.000 anchor 0.000 0.320s/rec
236
+ ep0 step 1500/3128 loss 0.662 kl 0.000 anchor 0.000 0.319s/rec
237
+ ep0 step 1510/3128 loss 0.737 kl 0.000 anchor 0.000 0.318s/rec
238
+ ep0 step 1520/3128 loss 0.453 kl 0.000 anchor 0.000 0.319s/rec
239
+ ep0 step 1530/3128 loss 0.459 kl 0.000 anchor 0.000 0.318s/rec
240
+ ep0 step 1540/3128 loss 0.446 kl 0.000 anchor 0.000 0.318s/rec
241
+ ep0 step 1550/3128 loss 0.648 kl 0.000 anchor 0.000 0.319s/rec
242
+ ep0 step 1560/3128 loss 0.693 kl 0.000 anchor 0.000 0.320s/rec
243
+ ep1 step 1570/3128 loss 0.488 kl 0.000 anchor 0.000 0.319s/rec
244
+ ep1 step 1580/3128 loss 0.460 kl 0.000 anchor 0.000 0.320s/rec
245
+ ep1 step 1590/3128 loss 0.577 kl 0.000 anchor 0.000 0.320s/rec
246
+ ep1 step 1600/3128 loss 0.321 kl 0.000 anchor 0.000 0.320s/rec
247
+ ep1 step 1610/3128 loss 0.645 kl 0.000 anchor 0.000 0.320s/rec
248
+ ep1 step 1620/3128 loss 0.368 kl 0.000 anchor 0.000 0.320s/rec
249
+ ep1 step 1630/3128 loss 0.377 kl 0.000 anchor 0.000 0.320s/rec
250
+ ep1 step 1640/3128 loss 0.730 kl 0.000 anchor 0.000 0.320s/rec
251
+ ep1 step 1650/3128 loss 0.523 kl 0.000 anchor 0.000 0.320s/rec
252
+ ep1 step 1660/3128 loss 0.361 kl 0.000 anchor 0.000 0.320s/rec
253
+ ep1 step 1670/3128 loss 0.357 kl 0.000 anchor 0.000 0.320s/rec
254
+ ep1 step 1680/3128 loss 0.373 kl 0.000 anchor 0.000 0.320s/rec
255
+ ep1 step 1690/3128 loss 1.005 kl 0.000 anchor 0.000 0.319s/rec
256
+ ep1 step 1700/3128 loss 0.404 kl 0.000 anchor 0.000 0.320s/rec
257
+ ep1 step 1710/3128 loss 0.366 kl 0.000 anchor 0.000 0.319s/rec
258
+ ep1 step 1720/3128 loss 0.387 kl 0.000 anchor 0.000 0.319s/rec
259
+ ep1 step 1730/3128 loss 0.608 kl 0.000 anchor 0.000 0.320s/rec
260
+ ep1 step 1740/3128 loss 0.553 kl 0.000 anchor 0.000 0.319s/rec
261
+ ep1 step 1750/3128 loss 0.367 kl 0.000 anchor 0.000 0.319s/rec
262
+ ep1 step 1760/3128 loss 0.671 kl 0.000 anchor 0.000 0.319s/rec
263
+ ep1 step 1770/3128 loss 0.347 kl 0.000 anchor 0.000 0.320s/rec
264
+ ep1 step 1780/3128 loss 0.362 kl 0.000 anchor 0.000 0.320s/rec
265
+ ep1 step 1790/3128 loss 0.598 kl 0.000 anchor 0.000 0.319s/rec
266
+ ep1 step 1800/3128 loss 0.612 kl 0.000 anchor 0.000 0.320s/rec
267
+ ep1 step 1810/3128 loss 0.366 kl 0.000 anchor 0.000 0.320s/rec
268
+ ep1 step 1820/3128 loss 0.536 kl 0.000 anchor 0.000 0.319s/rec
269
+ ep1 step 1830/3128 loss 0.359 kl 0.000 anchor 0.000 0.319s/rec
270
+ ep1 step 1840/3128 loss 0.398 kl 0.000 anchor 0.000 0.318s/rec
271
+ ep1 step 1850/3128 loss 0.523 kl 0.000 anchor 0.000 0.318s/rec
272
+ ep1 step 1860/3128 loss 0.309 kl 0.000 anchor 0.000 0.318s/rec
273
+ ep1 step 1870/3128 loss 0.455 kl 0.000 anchor 0.000 0.318s/rec
274
+ ep1 step 1880/3128 loss 0.500 kl 0.000 anchor 0.000 0.318s/rec
275
+ ep1 step 1890/3128 loss 0.462 kl 0.000 anchor 0.000 0.318s/rec
276
+ ep1 step 1900/3128 loss 0.419 kl 0.000 anchor 0.000 0.318s/rec
277
+ ep1 step 1910/3128 loss 0.459 kl 0.000 anchor 0.000 0.318s/rec
278
+ ep1 step 1920/3128 loss 0.391 kl 0.000 anchor 0.000 0.318s/rec
279
+ ep1 step 1930/3128 loss 0.489 kl 0.000 anchor 0.000 0.318s/rec
280
+ ep1 step 1940/3128 loss 0.385 kl 0.000 anchor 0.000 0.318s/rec
281
+ ep1 step 1950/3128 loss 0.263 kl 0.000 anchor 0.000 0.318s/rec
282
+ ep1 step 1960/3128 loss 0.417 kl 0.000 anchor 0.000 0.318s/rec
283
+ ep1 step 1970/3128 loss 0.258 kl 0.000 anchor 0.000 0.318s/rec
284
+ ep1 step 1980/3128 loss 0.618 kl 0.000 anchor 0.000 0.318s/rec
285
+ ep1 step 1990/3128 loss 0.751 kl 0.000 anchor 0.000 0.319s/rec
286
+ ep1 step 2000/3128 loss 0.277 kl 0.000 anchor 0.000 0.319s/rec
287
+ ep1 step 2010/3128 loss 0.334 kl 0.000 anchor 0.000 0.318s/rec
288
+ ep1 step 2020/3128 loss 0.484 kl 0.000 anchor 0.000 0.319s/rec
289
+ ep1 step 2030/3128 loss 0.390 kl 0.000 anchor 0.000 0.319s/rec
290
+ ep1 step 2040/3128 loss 0.337 kl 0.000 anchor 0.000 0.318s/rec
291
+ ep1 step 2050/3128 loss 0.333 kl 0.000 anchor 0.000 0.318s/rec
292
+ ep1 step 2060/3128 loss 0.383 kl 0.000 anchor 0.000 0.318s/rec
293
+ ep1 step 2070/3128 loss 0.540 kl 0.000 anchor 0.000 0.318s/rec
294
+ ep1 step 2080/3128 loss 0.404 kl 0.000 anchor 0.000 0.318s/rec
295
+ ep1 step 2090/3128 loss 0.353 kl 0.000 anchor 0.000 0.318s/rec
296
+ ep1 step 2100/3128 loss 0.603 kl 0.000 anchor 0.000 0.318s/rec
297
+ ep1 step 2110/3128 loss 0.476 kl 0.000 anchor 0.000 0.318s/rec
298
+ ep1 step 2120/3128 loss 0.461 kl 0.000 anchor 0.000 0.318s/rec
299
+ ep1 step 2130/3128 loss 0.291 kl 0.000 anchor 0.000 0.318s/rec
300
+ ep1 step 2140/3128 loss 0.549 kl 0.000 anchor 0.000 0.318s/rec
301
+ ep1 step 2150/3128 loss 0.468 kl 0.000 anchor 0.000 0.318s/rec
302
+ ep1 step 2160/3128 loss 0.422 kl 0.000 anchor 0.000 0.318s/rec
303
+ ep1 step 2170/3128 loss 0.381 kl 0.000 anchor 0.000 0.318s/rec
304
+ ep1 step 2180/3128 loss 0.513 kl 0.000 anchor 0.000 0.318s/rec
305
+ ep1 step 2190/3128 loss 0.327 kl 0.000 anchor 0.000 0.319s/rec
306
+ ep1 step 2200/3128 loss 0.488 kl 0.000 anchor 0.000 0.319s/rec
307
+ ep1 step 2210/3128 loss 0.696 kl 0.000 anchor 0.000 0.319s/rec
308
+ ep1 step 2220/3128 loss 0.320 kl 0.000 anchor 0.000 0.320s/rec
309
+ ep1 step 2230/3128 loss 0.342 kl 0.000 anchor 0.000 0.320s/rec
310
+ ep1 step 2240/3128 loss 0.268 kl 0.000 anchor 0.000 0.320s/rec
311
+ ep1 step 2250/3128 loss 0.466 kl 0.000 anchor 0.000 0.320s/rec
312
+ ep1 step 2260/3128 loss 0.370 kl 0.000 anchor 0.000 0.319s/rec
313
+ ep1 step 2270/3128 loss 0.300 kl 0.000 anchor 0.000 0.320s/rec
314
+ ep1 step 2280/3128 loss 0.409 kl 0.000 anchor 0.000 0.320s/rec
315
+ ep1 step 2290/3128 loss 0.594 kl 0.000 anchor 0.000 0.320s/rec
316
+ ep1 step 2300/3128 loss 0.333 kl 0.000 anchor 0.000 0.320s/rec
317
+ ep1 step 2310/3128 loss 0.355 kl 0.000 anchor 0.000 0.320s/rec
318
+ ep1 step 2320/3128 loss 0.394 kl 0.000 anchor 0.000 0.320s/rec
319
+ ep1 step 2330/3128 loss 0.459 kl 0.000 anchor 0.000 0.321s/rec
320
+ ep1 step 2340/3128 loss 0.214 kl 0.000 anchor 0.000 0.320s/rec
321
+ ep1 step 2350/3128 loss 0.425 kl 0.000 anchor 0.000 0.320s/rec
322
+ ep1 step 2360/3128 loss 0.320 kl 0.000 anchor 0.000 0.320s/rec
323
+ ep1 step 2370/3128 loss 0.623 kl 0.000 anchor 0.000 0.321s/rec
324
+ ep1 step 2380/3128 loss 0.335 kl 0.000 anchor 0.000 0.321s/rec
325
+ ep1 step 2390/3128 loss 0.463 kl 0.000 anchor 0.000 0.320s/rec
326
+ ep1 step 2400/3128 loss 0.478 kl 0.000 anchor 0.000 0.321s/rec
327
+ ep1 step 2410/3128 loss 0.253 kl 0.000 anchor 0.000 0.321s/rec
328
+ ep1 step 2420/3128 loss 0.229 kl 0.000 anchor 0.000 0.321s/rec
329
+ ep1 step 2430/3128 loss 0.308 kl 0.000 anchor 0.000 0.321s/rec
330
+ ep1 step 2440/3128 loss 0.313 kl 0.000 anchor 0.000 0.320s/rec
331
+ ep1 step 2450/3128 loss 0.457 kl 0.000 anchor 0.000 0.320s/rec
332
+ ep1 step 2460/3128 loss 0.748 kl 0.000 anchor 0.000 0.321s/rec
333
+ ep1 step 2470/3128 loss 0.483 kl 0.000 anchor 0.000 0.320s/rec
334
+ ep1 step 2480/3128 loss 0.343 kl 0.000 anchor 0.000 0.321s/rec
335
+ ep1 step 2490/3128 loss 0.459 kl 0.000 anchor 0.000 0.321s/rec
336
+ ep1 step 2500/3128 loss 0.300 kl 0.000 anchor 0.000 0.321s/rec
337
+ ep1 step 2510/3128 loss 0.376 kl 0.000 anchor 0.000 0.321s/rec
338
+ ep1 step 2520/3128 loss 0.413 kl 0.000 anchor 0.000 0.321s/rec
339
+ ep1 step 2530/3128 loss 0.247 kl 0.000 anchor 0.000 0.321s/rec
340
+ ep1 step 2540/3128 loss 0.390 kl 0.000 anchor 0.000 0.321s/rec
341
+ ep1 step 2550/3128 loss 0.362 kl 0.000 anchor 0.000 0.321s/rec
342
+ ep1 step 2560/3128 loss 0.196 kl 0.000 anchor 0.000 0.321s/rec
343
+ ep1 step 2570/3128 loss 0.443 kl 0.000 anchor 0.000 0.322s/rec
344
+ ep1 step 2580/3128 loss 0.241 kl 0.000 anchor 0.000 0.322s/rec
345
+ ep1 step 2590/3128 loss 0.683 kl 0.000 anchor 0.000 0.322s/rec
346
+ ep1 step 2600/3128 loss 0.332 kl 0.000 anchor 0.000 0.323s/rec
347
+ ep1 step 2610/3128 loss 0.506 kl 0.000 anchor 0.000 0.322s/rec
348
+ ep1 step 2620/3128 loss 0.410 kl 0.000 anchor 0.000 0.323s/rec
349
+ ep1 step 2630/3128 loss 0.305 kl 0.000 anchor 0.000 0.322s/rec
350
+ ep1 step 2640/3128 loss 0.252 kl 0.000 anchor 0.000 0.322s/rec
351
+ ep1 step 2650/3128 loss 0.439 kl 0.000 anchor 0.000 0.322s/rec
352
+ ep1 step 2660/3128 loss 0.341 kl 0.000 anchor 0.000 0.322s/rec
353
+ ep1 step 2670/3128 loss 0.477 kl 0.000 anchor 0.000 0.322s/rec
354
+ ep1 step 2680/3128 loss 0.525 kl 0.000 anchor 0.000 0.322s/rec
355
+ ep1 step 2690/3128 loss 0.625 kl 0.000 anchor 0.000 0.322s/rec
356
+ ep1 step 2700/3128 loss 0.552 kl 0.000 anchor 0.000 0.322s/rec
357
+ ep1 step 2710/3128 loss 0.611 kl 0.000 anchor 0.000 0.322s/rec
358
+ ep1 step 2720/3128 loss 0.217 kl 0.000 anchor 0.000 0.322s/rec
359
+ ep1 step 2730/3128 loss 0.168 kl 0.000 anchor 0.000 0.322s/rec
360
+ ep1 step 2740/3128 loss 0.428 kl 0.000 anchor 0.000 0.322s/rec
361
+ ep1 step 2750/3128 loss 0.354 kl 0.000 anchor 0.000 0.322s/rec
362
+ ep1 step 2760/3128 loss 0.385 kl 0.000 anchor 0.000 0.322s/rec
363
+ ep1 step 2770/3128 loss 0.292 kl 0.000 anchor 0.000 0.322s/rec
364
+ ep1 step 2780/3128 loss 0.405 kl 0.000 anchor 0.000 0.322s/rec
365
+ ep1 step 2790/3128 loss 0.648 kl 0.000 anchor 0.000 0.322s/rec
366
+ ep1 step 2800/3128 loss 0.378 kl 0.000 anchor 0.000 0.322s/rec
367
+ ep1 step 2810/3128 loss 0.331 kl 0.000 anchor 0.000 0.321s/rec
368
+ ep1 step 2820/3128 loss 0.479 kl 0.000 anchor 0.000 0.322s/rec
369
+ ep1 step 2830/3128 loss 0.369 kl 0.000 anchor 0.000 0.321s/rec
370
+ ep1 step 2840/3128 loss 0.837 kl 0.000 anchor 0.000 0.321s/rec
371
+ ep1 step 2850/3128 loss 0.425 kl 0.000 anchor 0.000 0.321s/rec
372
+ ep1 step 2860/3128 loss 0.558 kl 0.000 anchor 0.000 0.321s/rec
373
+ ep1 step 2870/3128 loss 0.429 kl 0.000 anchor 0.000 0.321s/rec
374
+ ep1 step 2880/3128 loss 0.363 kl 0.000 anchor 0.000 0.321s/rec
375
+ ep1 step 2890/3128 loss 0.345 kl 0.000 anchor 0.000 0.321s/rec
376
+ ep1 step 2900/3128 loss 0.489 kl 0.000 anchor 0.000 0.320s/rec
377
+ ep1 step 2910/3128 loss 0.283 kl 0.000 anchor 0.000 0.320s/rec
378
+ ep1 step 2920/3128 loss 0.285 kl 0.000 anchor 0.000 0.321s/rec
379
+ ep1 step 2930/3128 loss 0.440 kl 0.000 anchor 0.000 0.321s/rec
380
+ ep1 step 2940/3128 loss 0.306 kl 0.000 anchor 0.000 0.320s/rec
381
+ ep1 step 2950/3128 loss 0.457 kl 0.000 anchor 0.000 0.320s/rec
382
+ ep1 step 2960/3128 loss 0.355 kl 0.000 anchor 0.000 0.320s/rec
383
+ ep1 step 2970/3128 loss 0.181 kl 0.000 anchor 0.000 0.321s/rec
384
+ ep1 step 2980/3128 loss 0.365 kl 0.000 anchor 0.000 0.321s/rec
385
+ ep1 step 2990/3128 loss 0.361 kl 0.000 anchor 0.000 0.320s/rec
386
+ ep1 step 3000/3128 loss 0.643 kl 0.000 anchor 0.000 0.321s/rec
387
+ ep1 step 3010/3128 loss 0.380 kl 0.000 anchor 0.000 0.321s/rec
388
+ ep1 step 3020/3128 loss 0.341 kl 0.000 anchor 0.000 0.321s/rec
389
+ ep1 step 3030/3128 loss 0.227 kl 0.000 anchor 0.000 0.321s/rec
390
+ ep1 step 3040/3128 loss 0.231 kl 0.000 anchor 0.000 0.321s/rec
391
+ ep1 step 3050/3128 loss 0.436 kl 0.000 anchor 0.000 0.321s/rec
392
+ ep1 step 3060/3128 loss 0.260 kl 0.000 anchor 0.000 0.321s/rec
393
+ ep1 step 3070/3128 loss 0.513 kl 0.000 anchor 0.000 0.321s/rec
394
+ ep1 step 3080/3128 loss 0.240 kl 0.000 anchor 0.000 0.321s/rec
395
+ ep1 step 3090/3128 loss 0.446 kl 0.000 anchor 0.000 0.321s/rec
396
+ ep1 step 3100/3128 loss 0.373 kl 0.000 anchor 0.000 0.321s/rec
397
+ ep1 step 3110/3128 loss 0.521 kl 0.000 anchor 0.000 0.321s/rec
398
+ ep1 step 3120/3128 loss 0.683 kl 0.000 anchor 0.000 0.322s/rec
399
+ saved /kev/runs/g4-e2b-e2-l4/00-trial-0/checkpoint
training_config.json CHANGED
@@ -2,7 +2,7 @@
2
  "args": {
3
  "base": "google/gemma-4-E2B",
4
  "n_per_source": 1000,
5
- "epochs": 1,
6
  "lr": 0.0001,
7
  "head_lr": 0.0,
8
  "weight_decay": 0.01,
@@ -15,7 +15,7 @@
15
  "label_smoothing": 0.0,
16
  "brier_w": 0.0,
17
  "focal_gamma": 0.0,
18
- "suite": "evals/v7/decision-v7",
19
  "train_sources": "",
20
  "device": "cuda",
21
  "batch": 4,
@@ -36,7 +36,7 @@
36
  "anchor": "",
37
  "anchor_w": 0.0,
38
  "anchor_sources": "",
39
- "out": "runs/g4-e2b-e1",
40
  "data": "",
41
  "max_state": 384,
42
  "replay": 0,
 
2
  "args": {
3
  "base": "google/gemma-4-E2B",
4
  "n_per_source": 1000,
5
+ "epochs": 2,
6
  "lr": 0.0001,
7
  "head_lr": 0.0,
8
  "weight_decay": 0.01,
 
15
  "label_smoothing": 0.0,
16
  "brier_w": 0.0,
17
  "focal_gamma": 0.0,
18
+ "suite": "/kev/evals/v7/decision-v7",
19
  "train_sources": "",
20
  "device": "cuda",
21
  "batch": 4,
 
36
  "anchor": "",
37
  "anchor_w": 0.0,
38
  "anchor_sources": "",
39
+ "out": "/kev/runs/g4-e2b-e2-l4/00-trial-0/checkpoint",
40
  "data": "",
41
  "max_state": 384,
42
  "replay": 0,
training_metrics.json CHANGED
@@ -1,14 +1,14 @@
1
  {
2
- "wall_seconds": 4865.690968990326,
3
- "records_seen": 15142,
4
- "requested_records": 12506,
5
  "truncated_records": 0,
6
  "rejected_records": 0,
7
- "optimizer_steps": 1564,
8
- "forward_tokens": 3079220,
9
- "peak_device_bytes": 14623992320,
10
  "device": "cuda",
11
  "dtype": "bf16",
12
  "batch": 4,
13
- "peak_rss_bytes": 11216687104
14
  }
 
1
  {
2
+ "wall_seconds": 9709.689131975174,
3
+ "records_seen": 30212,
4
+ "requested_records": 25012,
5
  "truncated_records": 0,
6
  "rejected_records": 0,
7
+ "optimizer_steps": 3128,
8
+ "forward_tokens": 6112066,
9
+ "peak_device_bytes": 14624471040,
10
  "device": "cuda",
11
  "dtype": "bf16",
12
  "batch": 4,
13
+ "peak_rss_bytes": 11347705856
14
  }