trident-ec / g1fs16_st_dense_shared_s0.json
farguney's picture
Upload g1fs16_st_dense_shared_s0.json with huggingface_hub
0be23a4 verified
Raw
History Blame Contribute Delete
8.98 kB
{
"screen": "G1-FS16",
"arm": "dense_shared",
"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": 8,
"max_mid_entropy": 1.0799012184143066,
"pass": false
},
"history": [
{
"acc_d11": 0.002,
"exact_d11": 0.0,
"acc_d12": 0.0039,
"exact_d12": 0.0,
"acc_d13": 0.0049,
"exact_d13": 0.0,
"acc_d14": 0.0029,
"exact_d14": 0.0,
"acc_d15": 0.0029,
"exact_d15": 0.0,
"acc_d16": 0.0088,
"exact_d16": 0.0,
"heldout_byte_acc": 0.004231770833333333,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 8.016099482774734,
"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": 4102.45068359375,
"lam_mid": 0.0,
"cur_depth": 1,
"s_per_step": 1.3926,
"wall_s": 1.4
},
{
"acc_d11": 0.0049,
"exact_d11": 0.0,
"acc_d12": 0.0049,
"exact_d12": 0.0,
"acc_d13": 0.0039,
"exact_d13": 0.0,
"acc_d14": 0.0088,
"exact_d14": 0.0,
"acc_d15": 0.0029,
"exact_d15": 0.0,
"acc_d16": 0.002,
"exact_d16": 0.0,
"heldout_byte_acc": 0.004557291666666667,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 12.30225932598114,
"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": 1000,
"loss": 1060.5235595703125,
"lam_mid": 0.0,
"cur_depth": 1,
"s_per_step": 0.0341,
"wall_s": 34.1
},
{
"acc_d11": 0.0029,
"exact_d11": 0.0,
"acc_d12": 0.002,
"exact_d12": 0.0,
"acc_d13": 0.0098,
"exact_d13": 0.0,
"acc_d14": 0.0098,
"exact_d14": 0.0,
"acc_d15": 0.0039,
"exact_d15": 0.0,
"acc_d16": 0.0039,
"exact_d16": 0.0,
"heldout_byte_acc": 0.00537109375,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 16.243216653664906,
"heldout_min_exact": 0.0,
"singleton_len1": 1.0,
"distinct_codes": 0.0,
"code_map": "[0, 0]",
"empty_word_rate": 0.0,
"word_purity": 0.16176470588235295,
"word_len_match": 0.16176470588235295,
"step": 2000,
"loss": 1192.646240234375,
"lam_mid": 0.0,
"cur_depth": 2,
"s_per_step": 0.0355,
"wall_s": 71.0
},
{
"acc_d11": 0.0059,
"exact_d11": 0.0,
"acc_d12": 0.002,
"exact_d12": 0.0,
"acc_d13": 0.0,
"exact_d13": 0.0,
"acc_d14": 0.0,
"exact_d14": 0.0,
"acc_d15": 0.0,
"exact_d15": 0.0,
"acc_d16": 0.0039,
"exact_d16": 0.0,
"heldout_byte_acc": 0.001953125,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 18.66830571492513,
"heldout_min_exact": 0.0,
"singleton_len1": 1.0,
"distinct_codes": 0.0,
"code_map": "[0, 0]",
"empty_word_rate": 0.0,
"word_purity": 0.1323529411764706,
"word_len_match": 0.1323529411764706,
"step": 3000,
"loss": 1058.250244140625,
"lam_mid": 0.0,
"cur_depth": 3,
"s_per_step": 0.0369,
"wall_s": 110.8
},
{
"acc_d11": 0.0049,
"exact_d11": 0.0,
"acc_d12": 0.0088,
"exact_d12": 0.0,
"acc_d13": 0.0029,
"exact_d13": 0.0,
"acc_d14": 0.0039,
"exact_d14": 0.0,
"acc_d15": 0.001,
"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": 20.684587319691975,
"heldout_min_exact": 0.0,
"singleton_len1": 1.0,
"distinct_codes": 0.0,
"code_map": "[0, 0]",
"empty_word_rate": 0.0,
"word_purity": 0.3382352941176471,
"word_len_match": 0.38235294117647056,
"step": 4000,
"loss": 1056.6663818359375,
"lam_mid": 0.0,
"cur_depth": 4,
"s_per_step": 0.0387,
"wall_s": 155.0
},
{
"acc_d11": 0.0039,
"exact_d11": 0.0,
"acc_d12": 0.0049,
"exact_d12": 0.0,
"acc_d13": 0.0039,
"exact_d13": 0.0,
"acc_d14": 0.0029,
"exact_d14": 0.0,
"acc_d15": 0.0059,
"exact_d15": 0.0,
"acc_d16": 0.0039,
"exact_d16": 0.0,
"heldout_byte_acc": 0.004231770833333333,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 15.926769653956095,
"heldout_min_exact": 0.0,
"singleton_len1": 0.5,
"distinct_codes": 0.0,
"code_map": "[None, 0]",
"empty_word_rate": 0.0,
"word_purity": 0.0,
"word_len_match": 0.0,
"step": 5000,
"loss": 1064.9794921875,
"lam_mid": 0.0,
"cur_depth": 4,
"s_per_step": 0.0409,
"wall_s": 204.3
},
{
"acc_d11": 0.0029,
"exact_d11": 0.0,
"acc_d12": 0.0049,
"exact_d12": 0.0,
"acc_d13": 0.001,
"exact_d13": 0.0,
"acc_d14": 0.0039,
"exact_d14": 0.0,
"acc_d15": 0.002,
"exact_d15": 0.0,
"acc_d16": 0.0068,
"exact_d16": 0.0,
"heldout_byte_acc": 0.0035807291666666665,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 15.697057127952576,
"heldout_min_exact": 0.0,
"singleton_len1": 0.5,
"distinct_codes": 0.0,
"code_map": "[None, 0]",
"empty_word_rate": 0.0,
"word_purity": 0.0,
"word_len_match": 0.0,
"step": 6000,
"loss": 1079.1280517578125,
"lam_mid": 0.0,
"cur_depth": 5,
"s_per_step": 0.0426,
"wall_s": 255.9
},
{
"acc_d11": 0.0059,
"exact_d11": 0.0,
"acc_d12": 0.0039,
"exact_d12": 0.0,
"acc_d13": 0.001,
"exact_d13": 0.0,
"acc_d14": 0.0088,
"exact_d14": 0.0,
"acc_d15": 0.0029,
"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": 11.534734388192495,
"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": 7000,
"loss": 1068.650146484375,
"lam_mid": 0.0,
"cur_depth": 6,
"s_per_step": 0.0447,
"wall_s": 312.7
},
{
"acc_d11": 0.001,
"exact_d11": 0.0,
"acc_d12": 0.001,
"exact_d12": 0.0,
"acc_d13": 0.002,
"exact_d13": 0.0,
"acc_d14": 0.0068,
"exact_d14": 0.0,
"acc_d15": 0.002,
"exact_d15": 0.0,
"acc_d16": 0.0039,
"exact_d16": 0.0,
"heldout_byte_acc": 0.0027669270833333335,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 8.231780787309011,
"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": 8000,
"loss": 1202.6983642578125,
"lam_mid": 0.0,
"cur_depth": 7,
"s_per_step": 0.0468,
"wall_s": 374.0
},
{
"acc_d11": 0.002,
"exact_d11": 0.0,
"acc_d12": 0.001,
"exact_d12": 0.0,
"acc_d13": 0.0078,
"exact_d13": 0.0,
"acc_d14": 0.0029,
"exact_d14": 0.0,
"acc_d15": 0.0088,
"exact_d15": 0.0,
"acc_d16": 0.0049,
"exact_d16": 0.0,
"heldout_byte_acc": 0.004557291666666667,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 8.697717557350794,
"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": 9000,
"loss": 1068.110107421875,
"lam_mid": 0.0,
"cur_depth": 7,
"s_per_step": 0.0495,
"wall_s": 445.8
}
],
"final": {
"acc_d11": 0.002,
"exact_d11": 0.0,
"acc_d12": 0.001,
"exact_d12": 0.0,
"acc_d13": 0.0078,
"exact_d13": 0.0,
"acc_d14": 0.0029,
"exact_d14": 0.0,
"acc_d15": 0.0088,
"exact_d15": 0.0,
"acc_d16": 0.0049,
"exact_d16": 0.0,
"heldout_byte_acc": 0.004557291666666667,
"heldout_record_exact": 0.0,
"heldout_answer_bpb": 8.697717557350794,
"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": 9000,
"loss": 1068.110107421875,
"lam_mid": 0.0,
"cur_depth": 7,
"s_per_step": 0.0495,
"wall_s": 445.8
}
}
],
"partial": true
}