|
{ |
|
"is_super_res": false, |
|
"tokenizer_config": { |
|
"max_seq_len": 384, |
|
"char_level": true, |
|
"legacy_padding_behavior": true |
|
}, |
|
"image_size": 64, |
|
"num_channels": 256, |
|
"num_res_blocks": 2, |
|
"num_heads": 16, |
|
"num_heads_upsample": -1, |
|
"attention_resolutions": "4,8,16", |
|
"dropout": 0.0, |
|
"learn_sigma": true, |
|
"sigma_small": false, |
|
"class_cond": false, |
|
"diffusion_steps": 1000, |
|
"noise_schedule": "cosine", |
|
"timestep_respacing": "", |
|
"use_kl": false, |
|
"predict_xstart": false, |
|
"rescale_timesteps": false, |
|
"rescale_learned_sigmas": false, |
|
"resblock_updown": true, |
|
"use_checkpoint": false, |
|
"use_scale_shift_norm": true, |
|
"channels_per_head": 64, |
|
"channels_per_head_upsample": -1, |
|
"channel_mult": "1,2,2,4,4", |
|
"use_checkpoint_down": false, |
|
"use_checkpoint_middle": false, |
|
"use_checkpoint_up": false, |
|
"txt": true, |
|
"txt_dim": 512, |
|
"txt_depth": 4, |
|
"max_seq_len": 384, |
|
"txt_resolutions": "8,16,32", |
|
"cross_attn_channels_per_head": 128, |
|
"cross_attn_init_gain": 1.0, |
|
"cross_attn_gain_scale": 1.0, |
|
"text_lr_mult": null, |
|
"txt_output_layers_only": true, |
|
"monochrome": false, |
|
"monochrome_adapter": false, |
|
"txt_attn_before_attn": false, |
|
"txt_avoid_groupnorm": false, |
|
"cross_attn_orth_init": true, |
|
"cross_attn_q_t_emb": true, |
|
"txt_rezero": false, |
|
"cross_attn_rezero": false, |
|
"cross_attn_rezero_keeps_prenorm": true, |
|
"cross_attn_use_layerscale": false, |
|
"verbose": false, |
|
"txt_t5": true, |
|
"txt_rotary": false, |
|
"rgb_adapter": false, |
|
"weave_attn": true, |
|
"weave_use_ff": true, |
|
"weave_ff_rezero": false, |
|
"weave_ff_force_prenorm": false, |
|
"weave_ff_mult": 2, |
|
"weave_ff_glu": false, |
|
"weave_qkv_dim_always_text": true, |
|
"channels_last_mem": false, |
|
"txt_ff_glu": true, |
|
"txt_ff_mult": 3, |
|
"weave_v2": false, |
|
"use_checkpoint_lowcost": false, |
|
"weave_use_ff_gain": true, |
|
"return_diffusion_factory": false, |
|
"use_balanced_loss": false, |
|
"use_v_loss": false, |
|
"use_snr_plus_one_loss": false, |
|
"bread_adapter_at_ds": -1, |
|
"bread_adapter_nearest_in": false, |
|
"bread_adapter_zero_conv_in": false, |
|
"bread_adapter_only": false, |
|
"expand_timestep_base_dim": -1, |
|
"silu_impl": "fused", |
|
"using_capt": true, |
|
"xattn_capt": true, |
|
"weave_capt": false, |
|
"glide_style_capt_attn": true, |
|
"glide_style_capt_emb": false, |
|
"glide_style_capt_emb_init_scale": 0.1, |
|
"glide_style_capt_emb_nonlin": false, |
|
"clipname": "ViT-L/14@336px", |
|
"clip_use_penultimate_layer": true, |
|
"use_checkpoint_below_res": -1, |
|
"vb_loss_ratio": 1000.0, |
|
"no_attn": false, |
|
"no_attn_substitute_resblock": false, |
|
"freeze_capt_encoder": true, |
|
"clipmod": null, |
|
"post_txt_image_attn": "none", |
|
"txt_groupnorm_1group": true |
|
} |