ali_hnet-ffn_s3 / final.json
suchirsalhan's picture
final: log, final.json, probes
ccb2a14 verified
Raw History Blame Contribute Delete
1.58 kB
{
"event": "final",
"run": "ali_hnet-ffn_s3",
"arch": "hnet_ffn",
"objective": "ratio",
"condition": "A",
"step": 6000,
"params": 22785033,
"seed": 3,
"per_lang": {
"en": {
"cpc": 0.2075504223409757,
"cps": 28.21875,
"bpb": 1.8950562349348476,
"bytes_per_chunk": 4.724209084601592
},
"tr": {
"cpc": 0.1969286858531735,
"cps": 27.8515625,
"bpb": 2.0103088068221227,
"bytes_per_chunk": 5.754576985555604
},
"fi": {
"cpc": 0.25112965817872523,
"cps": 36.90625,
"bpb": 2.1378381237650945,
"bytes_per_chunk": 4.441123901976231
}
},
"bpb_mean": 2.0144010551740217,
"gini_cpc": 0.0551151592286927,
"gini_cps": 0.0649245161471026,
"bpic": 4.76699437022099,
"mean_boundary_entropy": 0.10512667894363403,
"dechunk": "ema",
"router_noise": 0.0,
"transitions": [
{
"step": 0,
"phase": 1,
"objective": "global",
"router_trainable": true,
"encoder_trainable": true,
"frozen_router_params": 0
},
{
"step": 2400,
"phase": 2,
"objective": "global",
"router_trainable": true,
"encoder_trainable": true,
"frozen_router_params": 0
},
{
"step": 4200,
"phase": 3,
"objective": "global",
"router_trainable": true,
"encoder_trainable": true,
"frozen_router_params": 0
}
],
"probe_overhead_sec_total": 0.0,
"probe_overhead_seq_frac": 0.0,
"wall_sec": 488.30208444595337,
"ffn_params": 268041,
"matched_to_ali_params": 268096
}