File size: 1,953 Bytes
47b7ad3 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 | {
"_comment": [
"This repository holds several EVOKE checkpoints. It is NOT a single loadable",
"diffusers pipeline root -- there is no top-level vae/ or transformer/, and",
"DiffusionPipeline.from_pretrained() on this repo id will not work.",
"",
"Every released model directory is the PARENT of a transformer/, and loads as",
" from_pretrained(<model_dir>, subfolder=\"transformer\")",
"with the shared components taken from evoke-base/.",
"",
"EVOKE also pins a development fork of diffusers that is not on PyPI. See",
"https://github.com/AlayaLab/Evoke for installation, weights layout and inference."
],
"_shipped_model": "evoke/stage3_post_distillation",
"_base_components": "evoke-base",
"_diffusers_fork_version": "0.37.0.dev0",
"_models": {
"evoke/stage1_camera_control": "multi-step camera-controllable model -- 50 steps, CFG 5.0",
"evoke/stage2_few_step_training": "few-step distillation -- 3-step pyramid",
"evoke/stage3_long_distillation": "30 s long-video distillation -- post-distill init",
"evoke/stage3_post_distillation": "the shipped model -- 3 steps, CFG-free",
"evoke/evoke_teacher": "dual-expert DMD teacher, {high,low}_noise -- training only"
},
"_components": {
"evoke/<model>/transformer": [
"evoke.modules.transformer_evoke",
"EvokeTransformer3DModel"
],
"evoke-base/vae": [
"diffusers",
"AutoencoderKLWan"
],
"evoke-base/text_encoder": [
"transformers",
"UMT5EncoderModel"
],
"evoke-base/tokenizer": [
"transformers",
"T5TokenizerFast"
],
"evoke-base/scheduler": [
"evoke.diffusers_version.scheduling_evoke_diffusers",
"EvokeScheduler"
]
},
"_external_weights": {
"ViGeo": "pkqbajng/ViGeo -- REQUIRED depth backend, downloaded separately",
"Depth-Anything-3": "depth-anything/da3-giant -- OPTIONAL, only for DEPTH_BACKEND=da3"
}
}
|