o1 / benchmarks.json
davidfoss's picture
Upload folder using huggingface_hub
f880f98 verified
Raw History Blame Contribute Delete
6.86 kB
{
"meta": {
"checkpoint": "o1.pt",
"checkpoint_bytes": 13487357,
"parameter_count": 3318465,
"carrier_bytes": 32512,
"active_crystals": 3768,
"vocab_size": 4096,
"invariant_16d_norm": 124.371132,
"device": "mps"
},
"symbolic_and_identity": {
"organ_hits": 12,
"organ_total": 16,
"organ_details": [
{
"prompt": "seven larger three is",
"expected": "seven",
"answer": "seven",
"ok": true
},
{
"prompt": "9 smaller 4 is",
"expected": "4",
"answer": "4",
"ok": true
},
{
"prompt": "median of 2, 9 and 5 is",
"expected": "5",
"answer": "3",
"ok": false
},
{
"prompt": "closer to 10 between 3 and 8 is",
"expected": "8",
"answer": "8",
"ok": true
},
{
"prompt": "count of 4, 4, 7, 4 is",
"expected": "3",
"answer": "4",
"ok": false
},
{
"prompt": "distance between 15 and 9 is",
"expected": "6",
"answer": "9",
"ok": false
},
{
"prompt": "8 add 5 =",
"expected": "13",
"answer": "13",
"ok": true
},
{
"prompt": "14 subtract 6 =",
"expected": "8",
"answer": "8",
"ok": true
},
{
"prompt": "7 times 6 =",
"expected": "42",
"answer": "42",
"ok": true
},
{
"prompt": "20 divide 4 =",
"expected": "5",
"answer": "10",
"ok": false
},
{
"prompt": "17 mod 5 =",
"expected": "2",
"answer": "2",
"ok": true
},
{
"prompt": "parity of 14 is",
"expected": "even",
"answer": "even",
"ok": true
},
{
"prompt": "3 glorp 4 =",
"expected": "10",
"answer": "10",
"ok": true
},
{
"prompt": "6 hypot 8 =",
"expected": "10",
"answer": "10",
"ok": true
},
{
"prompt": "3 phasor 4 =",
"expected": "5",
"answer": "5",
"ok": true
},
{
"prompt": "3 reladd 4 =",
"expected": "0.538462",
"answer": "0.538462",
"ok": true
}
],
"identity_probes": {
"Who created o1?": "o1 was created by David Tom Foss at Wirkung",
"Author:": "David Tom Foss (Wirkung Inc.,",
"Contact email:": "david@wirku.ng (David Tom Fo",
"X profile:": "https://x.com/fossdt (Da"
},
"weltbuch_verify": {
"valid": true,
"checked_count": 48,
"total_triples": 48,
"tip_fnv64": "0x7da8573a337604fe",
"tip_sha256": "007e8dba7436ff65c05d05afadc594b80ecba55f224c71731811f72a4d33773a"
},
"weltbuch_triples": 48
},
"wikitext2": {
"utf8_bytes": 1287656,
"crystal_tokens": 546887,
"bytes_per_token": 2.3545,
"static_fresh_bpb": 1.962,
"static_continuous_bpb": 2.3223,
"static_continuous_ppl_tok": 44.26,
"online_gssm_bpb": 2.5635,
"online_gssm_ppl_tok": 65.61,
"online_early_bpb": 2.4985,
"online_late_bpb": 2.3846,
"online_updates": 134,
"online_tok_per_sec": 3437.3,
"carrier_bytes_after_online": 32512
},
"arc_easy": {
"n": 2376,
"acc": 0.2588,
"acc_norm": 0.2643,
"correct_raw": 615,
"correct_norm": 628,
"eval_sec": 13.65,
"tok_per_sec": 52693.9
},
"arc_challenge": {
"n": 1172,
"acc": 0.1826,
"acc_norm": 0.2381,
"correct_raw": 214,
"correct_norm": 279,
"eval_sec": 7.94,
"tok_per_sec": 50226.6
},
"hellaswag": {
"n": 10042,
"acc": 0.2616,
"acc_norm": 0.2577,
"correct_raw": 2627,
"correct_norm": 2588,
"eval_sec": 144.65,
"tok_per_sec": 46978.6
},
"mmlu": {
"n": 14042,
"acc": 0.2347,
"acc_norm": 0.252,
"correct_raw": 3296,
"correct_norm": 3538,
"eval_sec": 164.29,
"tok_per_sec": 50069.3
},
"mmlu_dev_online": {
"n": 14042,
"acc": 0.235,
"acc_norm": 0.2508,
"correct_raw": 3300,
"correct_norm": 3522,
"eval_sec": 164.26,
"tok_per_sec": 50079.9,
"dev_adapt_steps": 3,
"dev_adapt_items": 285,
"dev_adapt_sec": 18.15
},
"gsm8k": {
"n": 1319,
"exact_match": 0.0152,
"correct": 20,
"gen_tok_per_sec": 2878.3
},
"humaneval": {
"n": 164,
"compiled": 80,
"pass_at_1": 0.0,
"passed": 0,
"gen_tok_per_sec": 759.3
},
"mbpp": {
"n": 500,
"compiled": 241,
"pass_at_1": 0.0,
"passed": 0,
"gen_tok_per_sec": 2602.0
},
"niah": {
"lengths": [
{
"T": 128,
"trials": 5,
"mc10_acc": 1.0,
"greedy_em": 0.6,
"carrier_bytes": 32512,
"stream_tok_per_sec": 1484.0
},
{
"T": 512,
"trials": 5,
"mc10_acc": 1.0,
"greedy_em": 0.4,
"carrier_bytes": 32512,
"stream_tok_per_sec": 3309.6
},
{
"T": 1024,
"trials": 5,
"mc10_acc": 1.0,
"greedy_em": 0.4,
"carrier_bytes": 32512,
"stream_tok_per_sec": 3936.5
},
{
"T": 2048,
"trials": 5,
"mc10_acc": 1.0,
"greedy_em": 0.2,
"carrier_bytes": 32512,
"stream_tok_per_sec": 4447.1
},
{
"T": 4096,
"trials": 5,
"mc10_acc": 0.2,
"greedy_em": 0.0,
"carrier_bytes": 32512,
"stream_tok_per_sec": 4720.4
},
{
"T": 8192,
"trials": 5,
"mc10_acc": 0.8,
"greedy_em": 0.0,
"carrier_bytes": 32512,
"stream_tok_per_sec": 4878.0
},
{
"T": 16384,
"trials": 5,
"mc10_acc": 0.2,
"greedy_em": 0.0,
"carrier_bytes": 32512,
"stream_tok_per_sec": 4946.9
},
{
"T": 32768,
"trials": 5,
"mc10_acc": 0.0,
"greedy_em": 0.0,
"carrier_bytes": 32512,
"stream_tok_per_sec": 5005.4
}
],
"constant_carrier_bytes_all_T": true,
"rss_delta_mb": 0.0
},
"ane_hardware": {
"ane_hardware_available": true,
"backend": "direct_private_ane_pingpong",
"compile_ms": 756.38,
"load_ms": 20.89,
"tokens_512_ane_ms": 152.578,
"tokens_per_sec_ane": 3355.7,
"us_per_token_ane": 298.0,
"recompile_count": 0,
"pytorch_parity_mean_cosine": 0.997528,
"pytorch_parity_min_cosine": 0.977215,
"path_a_us_per_step": 131.12,
"path_a_loss_drop": [
0.05118,
0.04095
],
"path_a_ane_energy_uj": 4000.0,
"path_b_ane_us_per_tok": 3.33,
"path_b_crystal_flash_us": 8.25,
"hcgm_total_sram_bytes": 2806914,
"hcgm_total_sram_mb": 2.6769,
"hcgm_mean_cosine": 0.997844
}
}