trident-ec / g1fs16_st_random_frozen_s0.json
farguney's picture
Upload g1fs16_st_random_frozen_s0.json with huggingface_hub
b5a4de4 verified
Raw
History Blame Contribute Delete
10.4 kB
{
"screen": "G1-FS16",
"arm": "random_frozen",
"seeds": [
0
],
"nonce": 0,
"steps": 24000,
"events": 2,
"m_records": 16,
"d": 64,
"layers": 3,
"carrier": 256,
"train_depth": 10,
"closure": 16,
"lam_mid": 0.0,
"lam_cpu": 0.02,
"lam_a": 0.02,
"mid_warmup": 0.0,
"curriculum": 0.5,
"straight_through": true,
"eps": 0.05,
"device": "cuda",
"runs": [
{
"seed": 0,
"oracle": {
"worst_acc": 0.0,
"worst_depth": 11,
"max_mid_entropy": 1.0215333700180054,
"pass": false
},
"history": [
{
"acc_d11": 0.002,
"exact_d11": 0.0,
"acc_d12": 0.0049,
"exact_d12": 0.0,
"acc_d13": 0.0059,
"exact_d13": 0.0,
"acc_d14": 0.0029,
"exact_d14": 0.0,
"acc_d15": 0.0029,
"exact_d15": 0.0,
"acc_d16": 0.0098,
"exact_d16": 0.0,
"heldout_byte_acc": 0.004720052083333333,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 8.016162316004435,
"heldout_min_exact": 0.0,
"singleton_len1": 0.5,
"distinct_codes": 0.0,
"code_map": "[1, None]",
"empty_word_rate": 1.0,
"word_purity": 0.0,
"word_len_match": 0.0,
"step": 0,
"loss": 4102.23291015625,
"lam_mid": 0.0,
"cur_depth": 1,
"s_per_step": 1.1911,
"wall_s": 1.2
},
{
"acc_d11": 0.0068,
"exact_d11": 0.0,
"acc_d12": 0.0059,
"exact_d12": 0.0,
"acc_d13": 0.0039,
"exact_d13": 0.0,
"acc_d14": 0.0059,
"exact_d14": 0.0,
"acc_d15": 0.0059,
"exact_d15": 0.0,
"acc_d16": 0.002,
"exact_d16": 0.0,
"heldout_byte_acc": 0.005045572916666667,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 10.018467326958975,
"heldout_min_exact": 0.0,
"singleton_len1": 0.0,
"distinct_codes": 0.0,
"code_map": "[None, None]",
"empty_word_rate": 1.0,
"word_purity": 0.0,
"word_len_match": 0.0,
"step": 1000,
"loss": 1058.2784423828125,
"lam_mid": 0.0,
"cur_depth": 1,
"s_per_step": 0.0288,
"wall_s": 28.8
},
{
"acc_d11": 0.0029,
"exact_d11": 0.0,
"acc_d12": 0.0059,
"exact_d12": 0.0,
"acc_d13": 0.0068,
"exact_d13": 0.0,
"acc_d14": 0.0088,
"exact_d14": 0.0,
"acc_d15": 0.0039,
"exact_d15": 0.0,
"acc_d16": 0.001,
"exact_d16": 0.0,
"heldout_byte_acc": 0.0048828125,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 11.416457295417786,
"heldout_min_exact": 0.0,
"singleton_len1": 0.0,
"distinct_codes": 0.0,
"code_map": "[None, None]",
"empty_word_rate": 1.0,
"word_purity": 0.0,
"word_len_match": 0.0,
"step": 2000,
"loss": 1182.09033203125,
"lam_mid": 0.0,
"cur_depth": 2,
"s_per_step": 0.0298,
"wall_s": 59.6
},
{
"acc_d11": 0.0088,
"exact_d11": 0.0,
"acc_d12": 0.0049,
"exact_d12": 0.0,
"acc_d13": 0.0098,
"exact_d13": 0.0,
"acc_d14": 0.0039,
"exact_d14": 0.0,
"acc_d15": 0.0029,
"exact_d15": 0.0,
"acc_d16": 0.001,
"exact_d16": 0.0,
"heldout_byte_acc": 0.005208333333333333,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 10.723142544428507,
"heldout_min_exact": 0.0,
"singleton_len1": 0.0,
"distinct_codes": 0.0,
"code_map": "[None, None]",
"empty_word_rate": 1.0,
"word_purity": 0.0,
"word_len_match": 0.0,
"step": 3000,
"loss": 1045.934814453125,
"lam_mid": 0.0,
"cur_depth": 3,
"s_per_step": 0.0311,
"wall_s": 93.4
},
{
"acc_d11": 0.0059,
"exact_d11": 0.0,
"acc_d12": 0.0068,
"exact_d12": 0.0,
"acc_d13": 0.0049,
"exact_d13": 0.0,
"acc_d14": 0.001,
"exact_d14": 0.0,
"acc_d15": 0.0049,
"exact_d15": 0.0,
"acc_d16": 0.0049,
"exact_d16": 0.0,
"heldout_byte_acc": 0.004720052083333333,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 11.959731797377268,
"heldout_min_exact": 0.0,
"singleton_len1": 0.0,
"distinct_codes": 0.0,
"code_map": "[None, None]",
"empty_word_rate": 1.0,
"word_purity": 0.0,
"word_len_match": 0.0,
"step": 4000,
"loss": 1056.711181640625,
"lam_mid": 0.0,
"cur_depth": 4,
"s_per_step": 0.0325,
"wall_s": 130.0
},
{
"acc_d11": 0.0029,
"exact_d11": 0.0,
"acc_d12": 0.002,
"exact_d12": 0.0,
"acc_d13": 0.0039,
"exact_d13": 0.0,
"acc_d14": 0.0068,
"exact_d14": 0.0,
"acc_d15": 0.0039,
"exact_d15": 0.0,
"acc_d16": 0.0059,
"exact_d16": 0.0,
"heldout_byte_acc": 0.004231770833333333,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 8.56697154045105,
"heldout_min_exact": 0.0,
"singleton_len1": 0.0,
"distinct_codes": 0.0,
"code_map": "[None, None]",
"empty_word_rate": 1.0,
"word_purity": 0.0,
"word_len_match": 0.0,
"step": 5000,
"loss": 1062.3609619140625,
"lam_mid": 0.0,
"cur_depth": 4,
"s_per_step": 0.0342,
"wall_s": 171.2
},
{
"acc_d11": 0.0049,
"exact_d11": 0.0,
"acc_d12": 0.0029,
"exact_d12": 0.0,
"acc_d13": 0.0049,
"exact_d13": 0.0,
"acc_d14": 0.0049,
"exact_d14": 0.0,
"acc_d15": 0.0049,
"exact_d15": 0.0,
"acc_d16": 0.0039,
"exact_d16": 0.0,
"heldout_byte_acc": 0.00439453125,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 8.27760124206543,
"heldout_min_exact": 0.0,
"singleton_len1": 0.0,
"distinct_codes": 0.0,
"code_map": "[None, None]",
"empty_word_rate": 1.0,
"word_purity": 0.0,
"word_len_match": 0.0,
"step": 6000,
"loss": 1146.08251953125,
"lam_mid": 0.0,
"cur_depth": 5,
"s_per_step": 0.0359,
"wall_s": 215.7
},
{
"acc_d11": 0.0029,
"exact_d11": 0.0,
"acc_d12": 0.002,
"exact_d12": 0.0,
"acc_d13": 0.001,
"exact_d13": 0.0,
"acc_d14": 0.0029,
"exact_d14": 0.0,
"acc_d15": 0.001,
"exact_d15": 0.0,
"acc_d16": 0.0029,
"exact_d16": 0.0,
"heldout_byte_acc": 0.0021158854166666665,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 8.090721507867178,
"heldout_min_exact": 0.0,
"singleton_len1": 0.0,
"distinct_codes": 0.0,
"code_map": "[None, None]",
"empty_word_rate": 1.0,
"word_purity": 0.0,
"word_len_match": 0.0,
"step": 7000,
"loss": 1060.4466552734375,
"lam_mid": 0.0,
"cur_depth": 6,
"s_per_step": 0.0377,
"wall_s": 263.8
},
{
"acc_d11": 0.0039,
"exact_d11": 0.0,
"acc_d12": 0.002,
"exact_d12": 0.0,
"acc_d13": 0.001,
"exact_d13": 0.0,
"acc_d14": 0.0049,
"exact_d14": 0.0,
"acc_d15": 0.0049,
"exact_d15": 0.0,
"acc_d16": 0.0059,
"exact_d16": 0.0,
"heldout_byte_acc": 0.0037434895833333335,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 8.100298275550207,
"heldout_min_exact": 0.0,
"singleton_len1": 0.0,
"distinct_codes": 0.0,
"code_map": "[None, None]",
"empty_word_rate": 1.0,
"word_purity": 0.0,
"word_len_match": 0.0,
"step": 8000,
"loss": 1059.7364501953125,
"lam_mid": 0.0,
"cur_depth": 7,
"s_per_step": 0.0395,
"wall_s": 315.9
},
{
"acc_d11": 0.0068,
"exact_d11": 0.0,
"acc_d12": 0.0039,
"exact_d12": 0.0,
"acc_d13": 0.001,
"exact_d13": 0.0,
"acc_d14": 0.002,
"exact_d14": 0.0,
"acc_d15": 0.0117,
"exact_d15": 0.0,
"acc_d16": 0.0049,
"exact_d16": 0.0,
"heldout_byte_acc": 0.005045572916666667,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 8.048147857189178,
"heldout_min_exact": 0.0,
"singleton_len1": 0.0,
"distinct_codes": 0.0,
"code_map": "[None, None]",
"empty_word_rate": 1.0,
"word_purity": 0.0,
"word_len_match": 0.0,
"step": 9000,
"loss": 1060.303955078125,
"lam_mid": 0.0,
"cur_depth": 7,
"s_per_step": 0.042,
"wall_s": 377.6
},
{
"acc_d11": 0.0029,
"exact_d11": 0.0,
"acc_d12": 0.0107,
"exact_d12": 0.0,
"acc_d13": 0.002,
"exact_d13": 0.0,
"acc_d14": 0.0039,
"exact_d14": 0.0,
"acc_d15": 0.0049,
"exact_d15": 0.0,
"acc_d16": 0.0068,
"exact_d16": 0.0,
"heldout_byte_acc": 0.005208333333333333,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 8.048940201600393,
"heldout_min_exact": 0.0,
"singleton_len1": 0.0,
"distinct_codes": 0.0,
"code_map": "[None, None]",
"empty_word_rate": 1.0,
"word_purity": 0.0,
"word_len_match": 0.0,
"step": 10000,
"loss": 1078.638671875,
"lam_mid": 0.0,
"cur_depth": 8,
"s_per_step": 0.0457,
"wall_s": 457.2
},
{
"acc_d11": 0.0059,
"exact_d11": 0.0,
"acc_d12": 0.0039,
"exact_d12": 0.0,
"acc_d13": 0.0078,
"exact_d13": 0.0,
"acc_d14": 0.002,
"exact_d14": 0.0,
"acc_d15": 0.0029,
"exact_d15": 0.0,
"acc_d16": 0.0039,
"exact_d16": 0.0,
"heldout_byte_acc": 0.00439453125,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 8.024273504813513,
"heldout_min_exact": 0.0,
"singleton_len1": 0.0,
"distinct_codes": 0.0,
"code_map": "[None, None]",
"empty_word_rate": 1.0,
"word_purity": 0.0,
"word_len_match": 0.0,
"step": 11000,
"loss": 1063.150634765625,
"lam_mid": 0.0,
"cur_depth": 9,
"s_per_step": 0.0517,
"wall_s": 568.4
}
],
"final": {
"acc_d11": 0.0059,
"exact_d11": 0.0,
"acc_d12": 0.0039,
"exact_d12": 0.0,
"acc_d13": 0.0078,
"exact_d13": 0.0,
"acc_d14": 0.002,
"exact_d14": 0.0,
"acc_d15": 0.0029,
"exact_d15": 0.0,
"acc_d16": 0.0039,
"exact_d16": 0.0,
"heldout_byte_acc": 0.00439453125,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 8.024273504813513,
"heldout_min_exact": 0.0,
"singleton_len1": 0.0,
"distinct_codes": 0.0,
"code_map": "[None, None]",
"empty_word_rate": 1.0,
"word_purity": 0.0,
"word_len_match": 0.0,
"step": 11000,
"loss": 1063.150634765625,
"lam_mid": 0.0,
"cur_depth": 9,
"s_per_step": 0.0517,
"wall_s": 568.4
}
}
],
"partial": true
}