trident-ec / scratch /g1rep_random_frozen_s1.json
farguney's picture
Upload scratch/g1rep_random_frozen_s1.json with huggingface_hub
94c9cc4 verified
Raw
History Blame Contribute Delete
9.46 kB
{
"screen": "G1-FS16",
"arm": "random_frozen",
"seeds": [
1
],
"nonce": 0,
"steps": 4000,
"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": 0.0,
"straight_through": true,
"st_after": 0.0,
"act_init": 4.0,
"sampled": true,
"eps": 0.05,
"device": "cuda",
"runs": [
{
"seed": 1,
"oracle": {
"worst_acc": 0.0,
"worst_depth": 8,
"max_mid_entropy": 1.0200042724609375,
"pass": false
},
"history": [
{
"acc_d11": 0.0,
"exact_d11": 0.0,
"acc_d12": 0.002,
"exact_d12": 0.0,
"acc_d13": 0.0039,
"exact_d13": 0.0,
"acc_d14": 0.0039,
"exact_d14": 0.0,
"acc_d15": 0.0,
"exact_d15": 0.0,
"acc_d16": 0.0059,
"exact_d16": 0.0,
"heldout_byte_acc": 0.0026041666666666665,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 8.074793547391891,
"heldout_min_exact": 0.0,
"singleton_len1": 1.0,
"distinct_codes": 1.0,
"code_map": "[1, 0]",
"empty_word_rate": 1.0,
"word_purity": 1.0,
"word_len_match": 1.0,
"purity_by_depth": "[1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0]",
"purity_trained": 1.0,
"purity_closure": 1.0,
"clip_compiler": 1.729,
"clip_spine": 4385.614,
"step": 0,
"loss": 4226.7734375,
"train_bits": 4175.31,
"lam_mid": 0.0,
"cur_depth": 10,
"st": 1,
"s_per_step": 2.2521,
"wall_s": 2.3
},
{
"acc_d11": 0.002,
"exact_d11": 0.0,
"acc_d12": 0.0049,
"exact_d12": 0.0,
"acc_d13": 0.0049,
"exact_d13": 0.0,
"acc_d14": 0.0049,
"exact_d14": 0.0,
"acc_d15": 0.0059,
"exact_d15": 0.0,
"acc_d16": 0.002,
"exact_d16": 0.0,
"heldout_byte_acc": 0.004069010416666667,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 8.410822649796804,
"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,
"clip_compiler": 0.404,
"clip_spine": 1130.573,
"step": 500,
"loss": 2237.34765625,
"train_bits": 2235.19,
"lam_mid": 0.0,
"cur_depth": 10,
"st": 1,
"s_per_step": 0.3176,
"wall_s": 158.8
},
{
"acc_d11": 0.0049,
"exact_d11": 0.0,
"acc_d12": 0.0039,
"exact_d12": 0.0,
"acc_d13": 0.0059,
"exact_d13": 0.0,
"acc_d14": 0.0029,
"exact_d14": 0.0,
"acc_d15": 0.0059,
"exact_d15": 0.0,
"acc_d16": 0.001,
"exact_d16": 0.0,
"heldout_byte_acc": 0.004069010416666667,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 8.67982147137324,
"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,
"clip_compiler": 0.598,
"clip_spine": 899.861,
"step": 1000,
"loss": 2357.125244140625,
"train_bits": 2354.51,
"lam_mid": 0.0,
"cur_depth": 10,
"st": 1,
"s_per_step": 0.3163,
"wall_s": 316.3
},
{
"acc_d11": 0.002,
"exact_d11": 0.0,
"acc_d12": 0.0029,
"exact_d12": 0.0,
"acc_d13": 0.002,
"exact_d13": 0.0,
"acc_d14": 0.0059,
"exact_d14": 0.0,
"acc_d15": 0.0049,
"exact_d15": 0.0,
"acc_d16": 0.002,
"exact_d16": 0.0,
"heldout_byte_acc": 0.0032552083333333335,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 8.675356189409891,
"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,
"clip_compiler": 0.282,
"clip_spine": 1247.017,
"step": 1500,
"loss": 2164.9091796875,
"train_bits": 2164.52,
"lam_mid": 0.0,
"cur_depth": 10,
"st": 1,
"s_per_step": 0.3158,
"wall_s": 473.7
},
{
"acc_d11": 0.002,
"exact_d11": 0.0,
"acc_d12": 0.0,
"exact_d12": 0.0,
"acc_d13": 0.0029,
"exact_d13": 0.0,
"acc_d14": 0.0029,
"exact_d14": 0.0,
"acc_d15": 0.0078,
"exact_d15": 0.0,
"acc_d16": 0.0029,
"exact_d16": 0.0,
"heldout_byte_acc": 0.0030924479166666665,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 8.857601881027222,
"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,
"clip_compiler": 0.433,
"clip_spine": 775.993,
"step": 2000,
"loss": 2357.899169921875,
"train_bits": 2357.09,
"lam_mid": 0.0,
"cur_depth": 10,
"st": 1,
"s_per_step": 0.3155,
"wall_s": 630.9
},
{
"acc_d11": 0.002,
"exact_d11": 0.0,
"acc_d12": 0.0039,
"exact_d12": 0.0,
"acc_d13": 0.002,
"exact_d13": 0.0,
"acc_d14": 0.002,
"exact_d14": 0.0,
"acc_d15": 0.0068,
"exact_d15": 0.0,
"acc_d16": 0.002,
"exact_d16": 0.0,
"heldout_byte_acc": 0.0030924479166666665,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 8.969808419545492,
"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,
"clip_compiler": 0.527,
"clip_spine": 597.457,
"step": 2500,
"loss": 1923.6553955078125,
"train_bits": 1922.9,
"lam_mid": 0.0,
"cur_depth": 10,
"st": 1,
"s_per_step": 0.3153,
"wall_s": 788.2
},
{
"acc_d11": 0.0,
"exact_d11": 0.0,
"acc_d12": 0.0039,
"exact_d12": 0.0,
"acc_d13": 0.0029,
"exact_d13": 0.0,
"acc_d14": 0.0029,
"exact_d14": 0.0,
"acc_d15": 0.0059,
"exact_d15": 0.0,
"acc_d16": 0.001,
"exact_d16": 0.0,
"heldout_byte_acc": 0.0027669270833333335,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 8.999689320723215,
"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,
"clip_compiler": 0.125,
"clip_spine": 1046.705,
"step": 3000,
"loss": 2408.83837890625,
"train_bits": 2407.02,
"lam_mid": 0.0,
"cur_depth": 10,
"st": 1,
"s_per_step": 0.3151,
"wall_s": 945.4
},
{
"acc_d11": 0.0059,
"exact_d11": 0.0,
"acc_d12": 0.0039,
"exact_d12": 0.0,
"acc_d13": 0.002,
"exact_d13": 0.0,
"acc_d14": 0.0029,
"exact_d14": 0.0,
"acc_d15": 0.0059,
"exact_d15": 0.0,
"acc_d16": 0.001,
"exact_d16": 0.0,
"heldout_byte_acc": 0.0035807291666666665,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 8.825632909933725,
"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,
"clip_compiler": 0.318,
"clip_spine": 623.635,
"step": 3500,
"loss": 1957.0614013671875,
"train_bits": 1956.62,
"lam_mid": 0.0,
"cur_depth": 10,
"st": 1,
"s_per_step": 0.3149,
"wall_s": 1102.3
},
{
"acc_d11": 0.0049,
"exact_d11": 0.0,
"acc_d12": 0.0068,
"exact_d12": 0.0,
"acc_d13": 0.0088,
"exact_d13": 0.0,
"acc_d14": 0.0049,
"exact_d14": 0.0,
"acc_d15": 0.0039,
"exact_d15": 0.0,
"acc_d16": 0.002,
"exact_d16": 0.0,
"heldout_byte_acc": 0.005208333333333333,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 9.00603057940801,
"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,
"clip_compiler": 0.146,
"clip_spine": 300.717,
"step": 3999,
"loss": 2253.82421875,
"train_bits": 2253.55,
"lam_mid": 0.0,
"cur_depth": 10,
"st": 1,
"s_per_step": 0.3148,
"wall_s": 1258.8
}
],
"final": {
"acc_d11": 0.0049,
"exact_d11": 0.0,
"acc_d12": 0.0068,
"exact_d12": 0.0,
"acc_d13": 0.0088,
"exact_d13": 0.0,
"acc_d14": 0.0049,
"exact_d14": 0.0,
"acc_d15": 0.0039,
"exact_d15": 0.0,
"acc_d16": 0.002,
"exact_d16": 0.0,
"heldout_byte_acc": 0.005208333333333333,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 9.00603057940801,
"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,
"clip_compiler": 0.146,
"clip_spine": 300.717,
"step": 3999,
"loss": 2253.82421875,
"train_bits": 2253.55,
"lam_mid": 0.0,
"cur_depth": 10,
"st": 1,
"s_per_step": 0.3148,
"wall_s": 1258.8
}
}
],
"partial": false,
"result_file": "scratch/g1rep_random_frozen_s1.json",
"off_law": {},
"certification_geometry": true
}