| { | |
| "architectures": [ | |
| "VQAdapterModel" | |
| ], | |
| "clip_dim": 768, | |
| "codebook_lr": 10.0, | |
| "is_rq": false, | |
| "mlp_dim": 1028, | |
| "mlp_hidden_dim": 512, | |
| "mlp_layers": 1, | |
| "rq_quantize_dropout": true, | |
| "rq_quantize_dropout_cutoff_index": 1, | |
| "rq_quantize_dropout_multiple_of": 4, | |
| "torch_dtype": "float32", | |
| "transformers_version": "4.31.0.dev0", | |
| "vq_accept_image_fmap": false, | |
| "vq_affine_param": false, | |
| "vq_affine_param_batch_decay": 0.99, | |
| "vq_affine_param_codebook_decay": 0.9, | |
| "vq_channel_last": true, | |
| "vq_codebook_dim": 32, | |
| "vq_codebook_size": 64, | |
| "vq_commitment_use_cross_entropy_loss": false, | |
| "vq_commitment_weight": 0.05, | |
| "vq_decay": 0.85, | |
| "vq_ema_update": true, | |
| "vq_eps": 1e-05, | |
| "vq_heads": 32, | |
| "vq_kmeans_init": false, | |
| "vq_kmeans_iters": 20, | |
| "vq_learnable_codebook": false, | |
| "vq_orthogonal_reg_active_codes_only": false, | |
| "vq_orthogonal_reg_max_codes": null, | |
| "vq_orthogonal_reg_weight": 0.0, | |
| "vq_reinmax": false, | |
| "vq_sample_codebook_temp": 1.0, | |
| "vq_separate_codebook_per_head": true, | |
| "vq_stochastic_sample_codes": true, | |
| "vq_straight_through": false, | |
| "vq_sync_affine_param": false, | |
| "vq_sync_codebook": false, | |
| "vq_sync_kmeans": true, | |
| "vq_sync_update_v": 0.0, | |
| "vq_threshold_ema_dead_code": 2, | |
| "vq_use_cosine_sim": false | |
| } | |