| { | |
| "attn_depth": 12, | |
| "attn_dim_head": 64, | |
| "attn_heads": 12, | |
| "attn_kwargs": { | |
| "attn_num_mem_kv": 4, | |
| "ff_glu": true | |
| }, | |
| "coarse_pre_gateloop_depth": 2, | |
| "condition_on_text": true, | |
| "cross_attn_num_mem_kv": 4, | |
| "dim": 768, | |
| "dropout": 0.0, | |
| "fine_attn_depth": 2, | |
| "fine_attn_dim_head": 32, | |
| "fine_attn_heads": 8, | |
| "fine_cross_attend_text": true, | |
| "fine_pre_gateloop_depth": 2, | |
| "flash_attn": true, | |
| "gateloop_use_heinsen": false, | |
| "max_seq_len": 1500, | |
| "num_sos_tokens": 1, | |
| "pad_id": -1, | |
| "quads": false, | |
| "text_cond_with_film": false, | |
| "text_condition_cond_drop_prob": 0.0, | |
| "text_condition_model_types": "bge" | |
| } |