Upload g1fs16_st_nosched_noop.json with huggingface_hub
Browse files- g1fs16_st_nosched_noop.json +80 -0
g1fs16_st_nosched_noop.json
ADDED
|
@@ -0,0 +1,80 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"screen": "G1-FS16",
|
| 3 |
+
"arm": "no_opfield",
|
| 4 |
+
"seeds": [
|
| 5 |
+
0
|
| 6 |
+
],
|
| 7 |
+
"nonce": 0,
|
| 8 |
+
"steps": 24000,
|
| 9 |
+
"events": 2,
|
| 10 |
+
"m_records": 16,
|
| 11 |
+
"d": 64,
|
| 12 |
+
"layers": 3,
|
| 13 |
+
"carrier": 256,
|
| 14 |
+
"train_depth": 10,
|
| 15 |
+
"closure": 16,
|
| 16 |
+
"lam_mid": 0.0,
|
| 17 |
+
"lam_cpu": 0.02,
|
| 18 |
+
"lam_a": 0.02,
|
| 19 |
+
"mid_warmup": 0.0,
|
| 20 |
+
"curriculum": 0.0,
|
| 21 |
+
"straight_through": true,
|
| 22 |
+
"eps": 0.05,
|
| 23 |
+
"device": "cuda",
|
| 24 |
+
"runs": [
|
| 25 |
+
{
|
| 26 |
+
"seed": 0,
|
| 27 |
+
"history": [
|
| 28 |
+
{
|
| 29 |
+
"acc_d11": 0.002,
|
| 30 |
+
"exact_d11": 0.0,
|
| 31 |
+
"acc_d12": 0.0049,
|
| 32 |
+
"exact_d12": 0.0,
|
| 33 |
+
"acc_d13": 0.0059,
|
| 34 |
+
"exact_d13": 0.0,
|
| 35 |
+
"acc_d14": 0.0029,
|
| 36 |
+
"exact_d14": 0.0,
|
| 37 |
+
"acc_d15": 0.0029,
|
| 38 |
+
"exact_d15": 0.0,
|
| 39 |
+
"acc_d16": 0.0068,
|
| 40 |
+
"exact_d16": 0.0,
|
| 41 |
+
"heldout_byte_acc": 0.004231770833333333,
|
| 42 |
+
"heldout_record_exact": 0.0,
|
| 43 |
+
"heldout_answer_bpb": 8.016158431768417,
|
| 44 |
+
"heldout_min_exact": 0.0,
|
| 45 |
+
"step": 0,
|
| 46 |
+
"loss": 4100.154296875,
|
| 47 |
+
"lam_mid": 0.0,
|
| 48 |
+
"cur_depth": 10,
|
| 49 |
+
"s_per_step": 0.8158,
|
| 50 |
+
"wall_s": 0.8
|
| 51 |
+
}
|
| 52 |
+
],
|
| 53 |
+
"final": {
|
| 54 |
+
"acc_d11": 0.002,
|
| 55 |
+
"exact_d11": 0.0,
|
| 56 |
+
"acc_d12": 0.0049,
|
| 57 |
+
"exact_d12": 0.0,
|
| 58 |
+
"acc_d13": 0.0059,
|
| 59 |
+
"exact_d13": 0.0,
|
| 60 |
+
"acc_d14": 0.0029,
|
| 61 |
+
"exact_d14": 0.0,
|
| 62 |
+
"acc_d15": 0.0029,
|
| 63 |
+
"exact_d15": 0.0,
|
| 64 |
+
"acc_d16": 0.0068,
|
| 65 |
+
"exact_d16": 0.0,
|
| 66 |
+
"heldout_byte_acc": 0.004231770833333333,
|
| 67 |
+
"heldout_record_exact": 0.0,
|
| 68 |
+
"heldout_answer_bpb": 8.016158431768417,
|
| 69 |
+
"heldout_min_exact": 0.0,
|
| 70 |
+
"step": 0,
|
| 71 |
+
"loss": 4100.154296875,
|
| 72 |
+
"lam_mid": 0.0,
|
| 73 |
+
"cur_depth": 10,
|
| 74 |
+
"s_per_step": 0.8158,
|
| 75 |
+
"wall_s": 0.8
|
| 76 |
+
}
|
| 77 |
+
}
|
| 78 |
+
],
|
| 79 |
+
"partial": true
|
| 80 |
+
}
|