trident-ec / g1fs16_field.json
farguney's picture
Upload g1fs16_field.json with huggingface_hub
8e37198 verified
Raw
History Blame Contribute Delete
5.45 kB
{
"screen": "G1-FS16",
"arm": "field",
"seeds": [
1
],
"nonce": 0,
"steps": 6000,
"events": 2,
"m_records": 8,
"d": 64,
"layers": 3,
"carrier": 192,
"train_depth": 6,
"closure": 8,
"lam_mid": 0.0,
"lam_cpu": 0.02,
"lam_a": 0.02,
"mid_warmup": 0.0,
"curriculum": 0.0,
"straight_through": true,
"st_after": 0.0,
"act_init": 4.0,
"sampled": true,
"eps": 0.05,
"device": "cpu",
"runs": [
{
"seed": 1,
"oracle": {
"worst_acc": 1.0,
"worst_depth": -1,
"max_mid_entropy": 0.001570447813719511,
"pass": true
},
"history": [
{
"acc_d7": 0.0013,
"exact_d7": 0.0,
"acc_d8": 0.0039,
"exact_d8": 0.0,
"heldout_byte_acc": 0.0026041666666666665,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 8.097914894421896,
"heldout_min_exact": 0.0,
"singleton_len1": 0.0,
"distinct_codes": 0.0,
"code_map": "[None, None]",
"empty_word_rate": 0.0,
"word_purity": 0.0,
"word_len_match": 0.0,
"step": 0,
"loss": 2093.167724609375,
"train_bits": 2088.73,
"lam_mid": 0.0,
"cur_depth": 6,
"st": 1,
"s_per_step": 1.5649,
"wall_s": 1.6
},
{
"acc_d7": 0.0033,
"exact_d7": 0.0,
"acc_d8": 0.0026,
"exact_d8": 0.0,
"heldout_byte_acc": 0.0029296875,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 8.2235999306043,
"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": 500,
"loss": 1025.944091796875,
"train_bits": 1025.77,
"lam_mid": 0.0,
"cur_depth": 6,
"st": 1,
"s_per_step": 0.0951,
"wall_s": 47.5
},
{
"acc_d7": 0.0046,
"exact_d7": 0.0,
"acc_d8": 0.0072,
"exact_d8": 0.0,
"heldout_byte_acc": 0.005859375,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 8.812768797079721,
"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": 1104.336181640625,
"train_bits": 1104.3,
"lam_mid": 0.0,
"cur_depth": 6,
"st": 1,
"s_per_step": 0.1003,
"wall_s": 100.3
},
{
"acc_d7": 0.0033,
"exact_d7": 0.0,
"acc_d8": 0.0046,
"exact_d8": 0.0,
"heldout_byte_acc": 0.00390625,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 9.563717067241669,
"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": 1500,
"loss": 728.7249145507812,
"train_bits": 728.79,
"lam_mid": 0.0,
"cur_depth": 6,
"st": 1,
"s_per_step": 0.1086,
"wall_s": 162.8
},
{
"acc_d7": 0.0072,
"exact_d7": 0.0,
"acc_d8": 0.0026,
"exact_d8": 0.0,
"heldout_byte_acc": 0.0048828125,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 9.90467651685079,
"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": 918.9149169921875,
"train_bits": 918.96,
"lam_mid": 0.0,
"cur_depth": 6,
"st": 1,
"s_per_step": 0.1198,
"wall_s": 239.7
},
{
"acc_d7": 0.0046,
"exact_d7": 0.0,
"acc_d8": 0.0052,
"exact_d8": 0.0,
"heldout_byte_acc": 0.0048828125,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 10.116246104240417,
"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": 2500,
"loss": 638.6556396484375,
"train_bits": 638.73,
"lam_mid": 0.0,
"cur_depth": 6,
"st": 1,
"s_per_step": 0.1201,
"wall_s": 300.3
},
{
"acc_d7": 0.0039,
"exact_d7": 0.0,
"acc_d8": 0.0007,
"exact_d8": 0.0,
"heldout_byte_acc": 0.0022786458333333335,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 9.915315906206766,
"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": 878.8968505859375,
"train_bits": 878.96,
"lam_mid": 0.0,
"cur_depth": 6,
"st": 1,
"s_per_step": 0.1219,
"wall_s": 365.7
}
],
"final": {
"acc_d7": 0.0039,
"exact_d7": 0.0,
"acc_d8": 0.0007,
"exact_d8": 0.0,
"heldout_byte_acc": 0.0022786458333333335,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 9.915315906206766,
"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": 878.8968505859375,
"train_bits": 878.96,
"lam_mid": 0.0,
"cur_depth": 6,
"st": 1,
"s_per_step": 0.1219,
"wall_s": 365.7
}
}
],
"partial": true
}