trident-ec / scratch /g1rep_dense_shared_s1.json
farguney's picture
Upload scratch/g1rep_dense_shared_s1.json with huggingface_hub
57d474d verified
Raw
History Blame Contribute Delete
9.56 kB
{
"screen": "G1-FS16",
"arm": "dense_shared",
"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": 1.042675256729126,
"pass": false
},
"history": [
{
"acc_d11": 0.002,
"exact_d11": 0.0,
"acc_d12": 0.0059,
"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": 0.0029,
"exact_d16": 0.0,
"heldout_byte_acc": 0.004231770833333333,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 8.055380572875341,
"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,
"clip_compiler": 1.844,
"clip_actions": 16.123,
"clip_spine": 4387.313,
"step": 0,
"loss": 4229.3623046875,
"train_bits": 4175.81,
"lam_mid": 0.0,
"cur_depth": 10,
"st": 1,
"s_per_step": 2.0601,
"wall_s": 2.1
},
{
"acc_d11": 0.0029,
"exact_d11": 0.0,
"acc_d12": 0.0098,
"exact_d12": 0.0,
"acc_d13": 0.001,
"exact_d13": 0.0,
"acc_d14": 0.002,
"exact_d14": 0.0,
"acc_d15": 0.0059,
"exact_d15": 0.0,
"acc_d16": 0.001,
"exact_d16": 0.0,
"heldout_byte_acc": 0.0037434895833333335,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 8.353239913781485,
"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": 4.024,
"clip_actions": 0.0,
"clip_spine": 789.729,
"step": 500,
"loss": 2219.36767578125,
"train_bits": 2217.72,
"lam_mid": 0.0,
"cur_depth": 10,
"st": 1,
"s_per_step": 0.3361,
"wall_s": 168.0
},
{
"acc_d11": 0.001,
"exact_d11": 0.0,
"acc_d12": 0.0049,
"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": 0.002,
"exact_d16": 0.0,
"heldout_byte_acc": 0.0030924479166666665,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 8.458874146143595,
"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": 1.664,
"clip_actions": 0.0,
"clip_spine": 386.074,
"step": 1000,
"loss": 2302.363037109375,
"train_bits": 2301.92,
"lam_mid": 0.0,
"cur_depth": 10,
"st": 1,
"s_per_step": 0.3348,
"wall_s": 334.8
},
{
"acc_d11": 0.0059,
"exact_d11": 0.0,
"acc_d12": 0.002,
"exact_d12": 0.0,
"acc_d13": 0.0049,
"exact_d13": 0.0,
"acc_d14": 0.0049,
"exact_d14": 0.0,
"acc_d15": 0.0029,
"exact_d15": 0.0,
"acc_d16": 0.0029,
"exact_d16": 0.0,
"heldout_byte_acc": 0.00390625,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 8.460377971331278,
"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": 1.96,
"clip_actions": 0.0,
"clip_spine": 602.461,
"step": 1500,
"loss": 2065.40576171875,
"train_bits": 2064.65,
"lam_mid": 0.0,
"cur_depth": 10,
"st": 1,
"s_per_step": 0.3344,
"wall_s": 501.6
},
{
"acc_d11": 0.0039,
"exact_d11": 0.0,
"acc_d12": 0.0059,
"exact_d12": 0.0,
"acc_d13": 0.0059,
"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": 0.004720052083333333,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 8.546362419923147,
"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": 2.622,
"clip_actions": 0.0,
"clip_spine": 738.279,
"step": 2000,
"loss": 2328.634033203125,
"train_bits": 2326.64,
"lam_mid": 0.0,
"cur_depth": 10,
"st": 1,
"s_per_step": 0.3341,
"wall_s": 668.1
},
{
"acc_d11": 0.0039,
"exact_d11": 0.0,
"acc_d12": 0.0039,
"exact_d12": 0.0,
"acc_d13": 0.0029,
"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": 0.0035807291666666665,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 8.451546092828115,
"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": 2.078,
"clip_actions": 21.553,
"clip_spine": 554.736,
"step": 2500,
"loss": 1730.2301025390625,
"train_bits": 1717.51,
"lam_mid": 0.0,
"cur_depth": 10,
"st": 1,
"s_per_step": 0.334,
"wall_s": 835.0
},
{
"acc_d11": 0.0039,
"exact_d11": 0.0,
"acc_d12": 0.0068,
"exact_d12": 0.0,
"acc_d13": 0.0078,
"exact_d13": 0.0,
"acc_d14": 0.0068,
"exact_d14": 0.0,
"acc_d15": 0.0029,
"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": 8.444048583507538,
"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": 1.536,
"clip_actions": 0.0,
"clip_spine": 2353.271,
"step": 3000,
"loss": 2311.021728515625,
"train_bits": 2310.29,
"lam_mid": 0.0,
"cur_depth": 10,
"st": 1,
"s_per_step": 0.334,
"wall_s": 1001.9
},
{
"acc_d11": 0.0059,
"exact_d11": 0.0,
"acc_d12": 0.0049,
"exact_d12": 0.0,
"acc_d13": 0.0059,
"exact_d13": 0.0,
"acc_d14": 0.0049,
"exact_d14": 0.0,
"acc_d15": 0.002,
"exact_d15": 0.0,
"acc_d16": 0.002,
"exact_d16": 0.0,
"heldout_byte_acc": 0.004231770833333333,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 8.562574605147043,
"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": 5.315,
"clip_actions": 29.348,
"clip_spine": 364.261,
"step": 3500,
"loss": 1571.31982421875,
"train_bits": 1558.2,
"lam_mid": 0.0,
"cur_depth": 10,
"st": 1,
"s_per_step": 0.3338,
"wall_s": 1168.5
},
{
"acc_d11": 0.0059,
"exact_d11": 0.0,
"acc_d12": 0.0049,
"exact_d12": 0.0,
"acc_d13": 0.0059,
"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": 0.0048828125,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 8.605278233687082,
"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": 1.534,
"clip_actions": 0.0,
"clip_spine": 249.589,
"step": 3999,
"loss": 2170.36083984375,
"train_bits": 2169.55,
"lam_mid": 0.0,
"cur_depth": 10,
"st": 1,
"s_per_step": 0.3338,
"wall_s": 1334.8
}
],
"final": {
"acc_d11": 0.0059,
"exact_d11": 0.0,
"acc_d12": 0.0049,
"exact_d12": 0.0,
"acc_d13": 0.0059,
"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": 0.0048828125,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 8.605278233687082,
"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": 1.534,
"clip_actions": 0.0,
"clip_spine": 249.589,
"step": 3999,
"loss": 2170.36083984375,
"train_bits": 2169.55,
"lam_mid": 0.0,
"cur_depth": 10,
"st": 1,
"s_per_step": 0.3338,
"wall_s": 1334.8
}
}
],
"partial": false,
"result_file": "scratch/g1rep_dense_shared_s1.json",
"off_law": {},
"certification_geometry": true
}