| { | |
| "H": 240, | |
| "W": 320, | |
| "R": 30, | |
| "C": 40, | |
| "levels": 4, | |
| "channels": [ | |
| 32, | |
| 64, | |
| 128, | |
| 256 | |
| ], | |
| "enc_channels": 1024, | |
| "enc_out": 256, | |
| "depth_resolution": 256.0, | |
| "weights": { | |
| "input_proj.w": { | |
| "offset": 0, | |
| "count": 96, | |
| "shape": [ | |
| 3, | |
| 32 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "input_proj.b": { | |
| "offset": 192, | |
| "count": 32, | |
| "shape": [ | |
| 32 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "prompt_proj.w": { | |
| "offset": 256, | |
| "count": 64, | |
| "shape": [ | |
| 2, | |
| 32 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "prompt_proj.b": { | |
| "offset": 384, | |
| "count": 32, | |
| "shape": [ | |
| 32 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "out_proj.w": { | |
| "offset": 448, | |
| "count": 32, | |
| "shape": [ | |
| 32, | |
| 1 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "out_proj.b": { | |
| "offset": 512, | |
| "count": 1, | |
| "shape": [ | |
| 1 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "down0.n1w": { | |
| "offset": 514, | |
| "count": 32, | |
| "shape": [ | |
| 32 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "down0.n1b": { | |
| "offset": 578, | |
| "count": 32, | |
| "shape": [ | |
| 32 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "down0.c1w8": { | |
| "offset": 642, | |
| "count": 27648, | |
| "shape": [ | |
| 27, | |
| 32, | |
| 32 | |
| ], | |
| "dtype": "int8" | |
| }, | |
| "down0.c1ws": { | |
| "offset": 28290, | |
| "count": 32, | |
| "shape": [ | |
| 32 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "down0.c1b": { | |
| "offset": 28418, | |
| "count": 32, | |
| "shape": [ | |
| 32 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "down0.c1clip": { | |
| "offset": 28482, | |
| "count": 1, | |
| "shape": [ | |
| 1 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "down0.c2w8": { | |
| "offset": 28486, | |
| "count": 27648, | |
| "shape": [ | |
| 27, | |
| 32, | |
| 32 | |
| ], | |
| "dtype": "int8" | |
| }, | |
| "down0.c2ws": { | |
| "offset": 56134, | |
| "count": 32, | |
| "shape": [ | |
| 32 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "down0.c2b": { | |
| "offset": 56262, | |
| "count": 32, | |
| "shape": [ | |
| 32 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "down0.c2clip": { | |
| "offset": 56326, | |
| "count": 1, | |
| "shape": [ | |
| 1 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "pool0.w8": { | |
| "offset": 56330, | |
| "count": 2048, | |
| "shape": [ | |
| 32, | |
| 64 | |
| ], | |
| "dtype": "int8" | |
| }, | |
| "pool0.ws": { | |
| "offset": 58378, | |
| "count": 64, | |
| "shape": [ | |
| 64 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "pool0.b": { | |
| "offset": 58634, | |
| "count": 64, | |
| "shape": [ | |
| 64 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "pool0.clip": { | |
| "offset": 58762, | |
| "count": 1, | |
| "shape": [ | |
| 1 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "down1.n1w": { | |
| "offset": 58766, | |
| "count": 64, | |
| "shape": [ | |
| 64 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "down1.n1b": { | |
| "offset": 58894, | |
| "count": 64, | |
| "shape": [ | |
| 64 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "down1.c1w8": { | |
| "offset": 59022, | |
| "count": 110592, | |
| "shape": [ | |
| 27, | |
| 64, | |
| 64 | |
| ], | |
| "dtype": "int8" | |
| }, | |
| "down1.c1ws": { | |
| "offset": 169614, | |
| "count": 64, | |
| "shape": [ | |
| 64 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "down1.c1b": { | |
| "offset": 169870, | |
| "count": 64, | |
| "shape": [ | |
| 64 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "down1.c1clip": { | |
| "offset": 169998, | |
| "count": 1, | |
| "shape": [ | |
| 1 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "down1.c2w8": { | |
| "offset": 170002, | |
| "count": 110592, | |
| "shape": [ | |
| 27, | |
| 64, | |
| 64 | |
| ], | |
| "dtype": "int8" | |
| }, | |
| "down1.c2ws": { | |
| "offset": 280594, | |
| "count": 64, | |
| "shape": [ | |
| 64 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "down1.c2b": { | |
| "offset": 280850, | |
| "count": 64, | |
| "shape": [ | |
| 64 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "down1.c2clip": { | |
| "offset": 280978, | |
| "count": 1, | |
| "shape": [ | |
| 1 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "pool1.w8": { | |
| "offset": 280982, | |
| "count": 8192, | |
| "shape": [ | |
| 64, | |
| 128 | |
| ], | |
| "dtype": "int8" | |
| }, | |
| "pool1.ws": { | |
| "offset": 289174, | |
| "count": 128, | |
| "shape": [ | |
| 128 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "pool1.b": { | |
| "offset": 289686, | |
| "count": 128, | |
| "shape": [ | |
| 128 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "pool1.clip": { | |
| "offset": 289942, | |
| "count": 1, | |
| "shape": [ | |
| 1 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "down2.n1w": { | |
| "offset": 289946, | |
| "count": 128, | |
| "shape": [ | |
| 128 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "down2.n1b": { | |
| "offset": 290202, | |
| "count": 128, | |
| "shape": [ | |
| 128 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "down2.c1w8": { | |
| "offset": 290458, | |
| "count": 442368, | |
| "shape": [ | |
| 27, | |
| 128, | |
| 128 | |
| ], | |
| "dtype": "int8" | |
| }, | |
| "down2.c1ws": { | |
| "offset": 732826, | |
| "count": 128, | |
| "shape": [ | |
| 128 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "down2.c1b": { | |
| "offset": 733338, | |
| "count": 128, | |
| "shape": [ | |
| 128 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "down2.c1clip": { | |
| "offset": 733594, | |
| "count": 1, | |
| "shape": [ | |
| 1 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "down2.c2w8": { | |
| "offset": 733598, | |
| "count": 442368, | |
| "shape": [ | |
| 27, | |
| 128, | |
| 128 | |
| ], | |
| "dtype": "int8" | |
| }, | |
| "down2.c2ws": { | |
| "offset": 1175966, | |
| "count": 128, | |
| "shape": [ | |
| 128 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "down2.c2b": { | |
| "offset": 1176478, | |
| "count": 128, | |
| "shape": [ | |
| 128 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "down2.c2clip": { | |
| "offset": 1176734, | |
| "count": 1, | |
| "shape": [ | |
| 1 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "pool2.w8": { | |
| "offset": 1176738, | |
| "count": 32768, | |
| "shape": [ | |
| 128, | |
| 256 | |
| ], | |
| "dtype": "int8" | |
| }, | |
| "pool2.ws": { | |
| "offset": 1209506, | |
| "count": 256, | |
| "shape": [ | |
| 256 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "pool2.b": { | |
| "offset": 1210530, | |
| "count": 256, | |
| "shape": [ | |
| 256 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "pool2.clip": { | |
| "offset": 1211042, | |
| "count": 1, | |
| "shape": [ | |
| 1 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "down3.n1w": { | |
| "offset": 1211046, | |
| "count": 256, | |
| "shape": [ | |
| 256 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "down3.n1b": { | |
| "offset": 1211558, | |
| "count": 256, | |
| "shape": [ | |
| 256 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "down3.c1w8": { | |
| "offset": 1212070, | |
| "count": 1769472, | |
| "shape": [ | |
| 27, | |
| 256, | |
| 256 | |
| ], | |
| "dtype": "int8" | |
| }, | |
| "down3.c1ws": { | |
| "offset": 2981542, | |
| "count": 256, | |
| "shape": [ | |
| 256 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "down3.c1b": { | |
| "offset": 2982566, | |
| "count": 256, | |
| "shape": [ | |
| 256 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "down3.c1clip": { | |
| "offset": 2983078, | |
| "count": 1, | |
| "shape": [ | |
| 1 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "down3.c2w8": { | |
| "offset": 2983082, | |
| "count": 1769472, | |
| "shape": [ | |
| 27, | |
| 256, | |
| 256 | |
| ], | |
| "dtype": "int8" | |
| }, | |
| "down3.c2ws": { | |
| "offset": 4752554, | |
| "count": 256, | |
| "shape": [ | |
| 256 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "down3.c2b": { | |
| "offset": 4753578, | |
| "count": 256, | |
| "shape": [ | |
| 256 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "down3.c2clip": { | |
| "offset": 4754090, | |
| "count": 1, | |
| "shape": [ | |
| 1 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "encoder_fuse.w8": { | |
| "offset": 4754094, | |
| "count": 262144, | |
| "shape": [ | |
| 1024, | |
| 256 | |
| ], | |
| "dtype": "int8" | |
| }, | |
| "encoder_fuse.ws": { | |
| "offset": 5016238, | |
| "count": 256, | |
| "shape": [ | |
| 256 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "encoder_fuse.b": { | |
| "offset": 5017262, | |
| "count": 256, | |
| "shape": [ | |
| 256 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "encoder_fuse.clip": { | |
| "offset": 5017774, | |
| "count": 1, | |
| "shape": [ | |
| 1 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "fuse0.w8": { | |
| "offset": 5017778, | |
| "count": 131072, | |
| "shape": [ | |
| 512, | |
| 256 | |
| ], | |
| "dtype": "int8" | |
| }, | |
| "fuse0.ws": { | |
| "offset": 5148850, | |
| "count": 256, | |
| "shape": [ | |
| 256 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "fuse0.b": { | |
| "offset": 5149874, | |
| "count": 256, | |
| "shape": [ | |
| 256 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "fuse0.clip": { | |
| "offset": 5150386, | |
| "count": 1, | |
| "shape": [ | |
| 1 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "fuse2.w8": { | |
| "offset": 5150390, | |
| "count": 65536, | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "dtype": "int8" | |
| }, | |
| "fuse2.ws": { | |
| "offset": 5215926, | |
| "count": 256, | |
| "shape": [ | |
| 256 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "fuse2.b": { | |
| "offset": 5216950, | |
| "count": 256, | |
| "shape": [ | |
| 256 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "fuse2.clip": { | |
| "offset": 5217462, | |
| "count": 1, | |
| "shape": [ | |
| 1 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "bott.n1w": { | |
| "offset": 5217466, | |
| "count": 256, | |
| "shape": [ | |
| 256 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "bott.n1b": { | |
| "offset": 5217978, | |
| "count": 256, | |
| "shape": [ | |
| 256 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "bott.c1w8": { | |
| "offset": 5218490, | |
| "count": 1769472, | |
| "shape": [ | |
| 27, | |
| 256, | |
| 256 | |
| ], | |
| "dtype": "int8" | |
| }, | |
| "bott.c1ws": { | |
| "offset": 6987962, | |
| "count": 256, | |
| "shape": [ | |
| 256 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "bott.c1b": { | |
| "offset": 6988986, | |
| "count": 256, | |
| "shape": [ | |
| 256 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "bott.c1clip": { | |
| "offset": 6989498, | |
| "count": 1, | |
| "shape": [ | |
| 1 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "bott.c2w8": { | |
| "offset": 6989502, | |
| "count": 1769472, | |
| "shape": [ | |
| 27, | |
| 256, | |
| 256 | |
| ], | |
| "dtype": "int8" | |
| }, | |
| "bott.c2ws": { | |
| "offset": 8758974, | |
| "count": 256, | |
| "shape": [ | |
| 256 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "bott.c2b": { | |
| "offset": 8759998, | |
| "count": 256, | |
| "shape": [ | |
| 256 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "bott.c2clip": { | |
| "offset": 8760510, | |
| "count": 1, | |
| "shape": [ | |
| 1 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "up0.w8": { | |
| "offset": 8760514, | |
| "count": 32768, | |
| "shape": [ | |
| 256, | |
| 128 | |
| ], | |
| "dtype": "int8" | |
| }, | |
| "up0.ws": { | |
| "offset": 8793282, | |
| "count": 128, | |
| "shape": [ | |
| 128 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "up0.b": { | |
| "offset": 8793794, | |
| "count": 128, | |
| "shape": [ | |
| 128 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "up0.clip": { | |
| "offset": 8794050, | |
| "count": 1, | |
| "shape": [ | |
| 1 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "dec0.n1w": { | |
| "offset": 8794054, | |
| "count": 128, | |
| "shape": [ | |
| 128 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "dec0.n1b": { | |
| "offset": 8794310, | |
| "count": 128, | |
| "shape": [ | |
| 128 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "dec0.c1w8": { | |
| "offset": 8794566, | |
| "count": 442368, | |
| "shape": [ | |
| 27, | |
| 128, | |
| 128 | |
| ], | |
| "dtype": "int8" | |
| }, | |
| "dec0.c1ws": { | |
| "offset": 9236934, | |
| "count": 128, | |
| "shape": [ | |
| 128 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "dec0.c1b": { | |
| "offset": 9237446, | |
| "count": 128, | |
| "shape": [ | |
| 128 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "dec0.c1clip": { | |
| "offset": 9237702, | |
| "count": 1, | |
| "shape": [ | |
| 1 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "dec0.c2w8": { | |
| "offset": 9237706, | |
| "count": 442368, | |
| "shape": [ | |
| 27, | |
| 128, | |
| 128 | |
| ], | |
| "dtype": "int8" | |
| }, | |
| "dec0.c2ws": { | |
| "offset": 9680074, | |
| "count": 128, | |
| "shape": [ | |
| 128 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "dec0.c2b": { | |
| "offset": 9680586, | |
| "count": 128, | |
| "shape": [ | |
| 128 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "dec0.c2clip": { | |
| "offset": 9680842, | |
| "count": 1, | |
| "shape": [ | |
| 1 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "up1.w8": { | |
| "offset": 9680846, | |
| "count": 8192, | |
| "shape": [ | |
| 128, | |
| 64 | |
| ], | |
| "dtype": "int8" | |
| }, | |
| "up1.ws": { | |
| "offset": 9689038, | |
| "count": 64, | |
| "shape": [ | |
| 64 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "up1.b": { | |
| "offset": 9689294, | |
| "count": 64, | |
| "shape": [ | |
| 64 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "up1.clip": { | |
| "offset": 9689422, | |
| "count": 1, | |
| "shape": [ | |
| 1 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "dec1.n1w": { | |
| "offset": 9689426, | |
| "count": 64, | |
| "shape": [ | |
| 64 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "dec1.n1b": { | |
| "offset": 9689554, | |
| "count": 64, | |
| "shape": [ | |
| 64 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "dec1.c1w8": { | |
| "offset": 9689682, | |
| "count": 110592, | |
| "shape": [ | |
| 27, | |
| 64, | |
| 64 | |
| ], | |
| "dtype": "int8" | |
| }, | |
| "dec1.c1ws": { | |
| "offset": 9800274, | |
| "count": 64, | |
| "shape": [ | |
| 64 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "dec1.c1b": { | |
| "offset": 9800530, | |
| "count": 64, | |
| "shape": [ | |
| 64 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "dec1.c1clip": { | |
| "offset": 9800658, | |
| "count": 1, | |
| "shape": [ | |
| 1 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "dec1.c2w8": { | |
| "offset": 9800662, | |
| "count": 110592, | |
| "shape": [ | |
| 27, | |
| 64, | |
| 64 | |
| ], | |
| "dtype": "int8" | |
| }, | |
| "dec1.c2ws": { | |
| "offset": 9911254, | |
| "count": 64, | |
| "shape": [ | |
| 64 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "dec1.c2b": { | |
| "offset": 9911510, | |
| "count": 64, | |
| "shape": [ | |
| 64 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "dec1.c2clip": { | |
| "offset": 9911638, | |
| "count": 1, | |
| "shape": [ | |
| 1 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "up2.w8": { | |
| "offset": 9911642, | |
| "count": 2048, | |
| "shape": [ | |
| 64, | |
| 32 | |
| ], | |
| "dtype": "int8" | |
| }, | |
| "up2.ws": { | |
| "offset": 9913690, | |
| "count": 32, | |
| "shape": [ | |
| 32 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "up2.b": { | |
| "offset": 9913818, | |
| "count": 32, | |
| "shape": [ | |
| 32 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "up2.clip": { | |
| "offset": 9913882, | |
| "count": 1, | |
| "shape": [ | |
| 1 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "dec2.n1w": { | |
| "offset": 9913886, | |
| "count": 32, | |
| "shape": [ | |
| 32 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "dec2.n1b": { | |
| "offset": 9913950, | |
| "count": 32, | |
| "shape": [ | |
| 32 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "dec2.c1w8": { | |
| "offset": 9914014, | |
| "count": 27648, | |
| "shape": [ | |
| 27, | |
| 32, | |
| 32 | |
| ], | |
| "dtype": "int8" | |
| }, | |
| "dec2.c1ws": { | |
| "offset": 9941662, | |
| "count": 32, | |
| "shape": [ | |
| 32 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "dec2.c1b": { | |
| "offset": 9941790, | |
| "count": 32, | |
| "shape": [ | |
| 32 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "dec2.c1clip": { | |
| "offset": 9941854, | |
| "count": 1, | |
| "shape": [ | |
| 1 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "dec2.c2w8": { | |
| "offset": 9941858, | |
| "count": 27648, | |
| "shape": [ | |
| 27, | |
| 32, | |
| 32 | |
| ], | |
| "dtype": "int8" | |
| }, | |
| "dec2.c2ws": { | |
| "offset": 9969506, | |
| "count": 32, | |
| "shape": [ | |
| 32 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "dec2.c2b": { | |
| "offset": 9969634, | |
| "count": 32, | |
| "shape": [ | |
| 32 | |
| ], | |
| "dtype": "float16" | |
| }, | |
| "dec2.c2clip": { | |
| "offset": 9969698, | |
| "count": 1, | |
| "shape": [ | |
| 1 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| "uv_bias": { | |
| "offset": 9969702, | |
| "count": 307200, | |
| "shape": [ | |
| 256, | |
| 30, | |
| 40 | |
| ], | |
| "dtype": "float32" | |
| } | |
| } | |
| } |