TokenizerTest-32m-A / bpb_curve.json
Maggio33's picture
TokenizerTest-32m-A: final + every-10k weights, card, bpb curve
9e59882 verified
Raw History Blame Contribute Delete
2.22 kB
{
"definition": "bits/byte = sum of NLL over text tokens / ln 2 / UTF-8 bytes; non-overlapping 1024-token windows; EOS targets excluded",
"trainer_sha256": "0223c0830ed4ae323f6424038a2de7bfb6b425653f3591dd55844f95706f5d17",
"tokenizer_sha256": "60e23148ae8ee2f8b5e8e2a42d23cedbc550fdccb598e0262c15b8c6af60d9c5",
"sets": {
"W": {
"sha256": "234e7ec21bf2236bc74425591531f2c4772f8c49ff853d57f6fa16f33ee90ac7",
"docs": 463,
"text_bytes": 1430217,
"tokens_text": 341025
},
"val": {
"sha256": "748d13898e736666d96f1c9307222c52414a9b66b06bf212e7a07e765babc668",
"docs": 2723,
"text_bytes": 7232820,
"tokens_text": 1639802
}
},
"device": "cpu",
"ckpts": [
{
"step": 10000,
"ckpt_sha256": "985246af7a264adb3ccacb1968c6424c485153d0ed2d6afe1fe942a8f3854bf8",
"W": 1.2327165523642996,
"val": 1.1909538985721075
},
{
"step": 20000,
"ckpt_sha256": "089dbd5b141657416caab8d29c93139d0a993ecb443810bbe426e9e1e50039fe",
"W": 1.179908492896963,
"val": 1.1447952006986657
},
{
"step": 30000,
"ckpt_sha256": "930dbc5dd583cf6c489bea0f334805879c1e3cfb4eaa0161f5969362bea1f52a",
"W": 1.1499359005482026,
"val": 1.119994574275409
},
{
"step": 40000,
"ckpt_sha256": "fb54576a3504f496dd31a47043bfa4a13c38f375fbdafe1dbb15640bda9740a5",
"W": 1.1303579741310747,
"val": 1.1032037185642838
},
{
"step": 50000,
"ckpt_sha256": "a4b60f59dc85578435f37f53fdd5513d7907fa627b13f71330d387667c627750",
"W": 1.114065574242555,
"val": 1.0888208051528139
},
{
"step": 60000,
"ckpt_sha256": "2ee24f6e1d09faea31f17f472c72dce9fbc6f29b6f3906cb40eacc36bd8bc93c",
"W": 1.0999765100751717,
"val": 1.0772723645560955
},
{
"step": 70000,
"ckpt_sha256": "95150f77acc6ab16d0e877e0bf4cf91ee21b54225d5cc2e31989617164e20e59",
"W": 1.0885261369122678,
"val": 1.0674803839610683
},
{
"step": 80000,
"ckpt_sha256": "066adbf14e74e7cf6deb887c9b8f5536a9dac896b7fa0f40da16cae816f17cf7",
"W": 1.0817551480434846,
"val": 1.061001320817973
},
{
"step": 89000,
"ckpt_sha256": "19040910cb8bd11136364da88d01080f0fcb3f281383b3bfb27bf347f78cc92d",
"W": 1.078115063691179,
"val": 1.0582625876804834
}
]
}