Load with the stock Diffusers and Transformers classes: one-hot timestep code, 51-layer text encoder

4a14bfd verified 2 days ago

1 kB

{
  "_class_name": "MiniMaxH3ImageTransformer2DModel",
  "_diffusers_version": "0.42.0.dev0",
  "num_attention_heads": 56,
  "attention_head_dim": 128,
  "hidden_size": 5376,
  "num_layers": 50,
  "num_refiner_layers": 2,
  "ffn_dim": 14336,
  "in_channels": 24,
  "patch_size": [
    1,
    2,
    2
  ],
  "text_dim": 5120,
  "freq_dim": 256,
  "time_embed_hidden_dim": 16,
  "time_embed_dim": 8,
  "rope_freq_dim": 16,
  "rope_theta": 10000.0,
  "norm_eps": 1e-05,
  "qk_norm_eps": 1e-05,
  "final_norm_eps": 1e-05,
  "num_pdd_heads": 6,
  "_timestep_grid": [
    0.0,
    0.03247630596160889,
    0.11245149374008179,
    0.25,
    0.4672260880470276,
    0.7679911255836487,
    0.9990000128746033
  ],
  "_notice": "Modified by WaveCut from MiniMaxAI/MiniMax-H3-Image@492fd8c4eca666f1085f57e30738c1e0d86b3439: time_embedder replaced by a one-hot code of the 7 sampling timesteps; adaln_proj.linear and norm_out.linear reduced to 8 input columns holding the original outputs for those timesteps."
}

	{
	"_class_name": "MiniMaxH3ImageTransformer2DModel",
	"_diffusers_version": "0.42.0.dev0",
	"num_attention_heads": 56,
	"attention_head_dim": 128,
	"hidden_size": 5376,
	"num_layers": 50,
	"num_refiner_layers": 2,
	"ffn_dim": 14336,
	"in_channels": 24,
	"patch_size": [
	1,
	2,
	2
	],
	"text_dim": 5120,
	"freq_dim": 256,
	"time_embed_hidden_dim": 16,
	"time_embed_dim": 8,
	"rope_freq_dim": 16,
	"rope_theta": 10000.0,
	"norm_eps": 1e-05,
	"qk_norm_eps": 1e-05,
	"final_norm_eps": 1e-05,
	"num_pdd_heads": 6,
	"_timestep_grid": [
	0.0,
	0.03247630596160889,
	0.11245149374008179,
	0.25,
	0.4672260880470276,
	0.7679911255836487,
	0.9990000128746033
	],
	"_notice": "Modified by WaveCut from MiniMaxAI/MiniMax-H3-Image@492fd8c4eca666f1085f57e30738c1e0d86b3439: time_embedder replaced by a one-hot code of the 7 sampling timesteps; adaln_proj.linear and norm_out.linear reduced to 8 input columns holding the original outputs for those timesteps."
	}