| { | |
| "activation_dropout": 0.0, | |
| "activation_function": "relu", | |
| "architectures": [ | |
| "NllbMoeForConditionalGeneration" | |
| ], | |
| "attention_dropout": 0.1, | |
| "batch_prioritized_routing": false, | |
| "bos_token_id": 0, | |
| "d_model": 16, | |
| "decoder_attention_heads": 4, | |
| "decoder_ffn_dim": 4, | |
| "decoder_layerdrop": 0.0, | |
| "decoder_layers": 4, | |
| "decoder_sparse_step": 1, | |
| "decoder_start_token_id": 2, | |
| "dropout": 0.1, | |
| "encoder_attention_heads": 4, | |
| "encoder_ffn_dim": 4, | |
| "encoder_layerdrop": 0.0, | |
| "encoder_layers": 4, | |
| "encoder_sparse_step": 2, | |
| "eos_token_id": 2, | |
| "expert_capacity": 100, | |
| "init_std": 0.02, | |
| "is_encoder_decoder": true, | |
| "max_position_embeddings": 20, | |
| "model_type": "nllb_moe", | |
| "moe_eval_capacity_token_fraction": 1.0, | |
| "moe_token_dropout": 0.2, | |
| "normalize_router_prob_before_dropping": false, | |
| "num_experts": 4, | |
| "num_hidden_layers": 4, | |
| "output_router_logits": false, | |
| "pad_token_id": 1, | |
| "router_aux_loss_coef": 0.001, | |
| "router_bias": false, | |
| "router_dtype": "float32", | |
| "router_ignore_padding_tokens": false, | |
| "router_jitter_noise": 0.0, | |
| "router_z_loss_coef": 0.001, | |
| "scale_embedding": true, | |
| "second_expert_policy": "all", | |
| "torch_dtype": "float32", | |
| "transformers_version": "4.28.0.dev0", | |
| "use_cache": true, | |
| "vocab_size": 1199 | |
| } | |