ra2 / transformer /config.json
recoilme's picture
Ra sdxs-VAE (step 61000)
e995b65 verified
Raw History Blame Contribute Delete
1.73 kB
{
"_class_name": "UDiT",
"format": "udm/v1",
"storage_dtype": "float16",
"latent_channels": 32,
"text_dim": 768,
"body": {
"kind": "пирамидальный U: ширина растёт к низу",
"widths": [
192,
512,
1024,
1536
],
"heads": [
6,
8,
8,
8
],
"blocks": {
"e0": 6,
"e0x": 3,
"d0x": 3,
"e1": 5,
"e2": 5,
"e3": 4,
"d2": 4,
"d1": 4,
"d0": 4
},
"levels": [
{
"level": 0,
"tokens": "÷1",
"width": 192,
"blocks": "conv 6 enc / 4 dec + 2 self-attn на сторону"
},
{
"level": 1,
"tokens": "÷2",
"width": 512,
"blocks": "attn 5 enc / 4 dec"
},
{
"level": 2,
"tokens": "÷4",
"width": 1024,
"blocks": "attn 5 enc / 4 dec"
},
{
"level": 3,
"tokens": "÷8",
"width": 1536,
"blocks": "attn 4 (боттлнек)"
}
],
"skips": "аддитивные (SD-стиль): e0/e1/e2 прибавляются в декодере",
"down": "mean-pool 2x + Linear",
"up": "Linear + nearest 2x",
"positions": {
"rope": {
"kind": "2d",
"base": 100,
"phase": "в латентных пикселях на каждом уровне"
},
"absolute": {
"kind": "2d sincos",
"dim": 64,
"into": "zero-init Linear(64 -> width) к выходу stem, в паре с RoPE"
}
},
"text": "кросс-внимание на всех уровнях",
"parametrization": "x0 латента",
"params": 240.04
}
}