| { |
| "activation": "swiglu", |
| "alternate_sparsity": true, |
| "architectures": [ |
| "BalmMoEForMaskedLM" |
| ], |
| "attention_classifier": false, |
| "attention_dropout": 0.1, |
| "classifier_activation": "tanh", |
| "classifier_attention_heads": 20, |
| "classifier_freeze_base": true, |
| "dropout": 0.1, |
| "expert_activation": "swiglu", |
| "expert_bias": true, |
| "expert_capacity": 1.0, |
| "expert_capacity_type": "multiplier", |
| "expert_dropout": 0.1, |
| "expert_intermediate_size": [ |
| 2560, |
| 2560, |
| 2560, |
| 2560, |
| 2560, |
| 2560, |
| 2560, |
| 2560 |
| ], |
| "ffn_bias": true, |
| "hidden_dropout": 0.1, |
| "hidden_size": 640, |
| "homogeneous_experts": false, |
| "initializer_range": 0.02, |
| "intermediate_size": 2560, |
| "layer_norm_eps": 1e-05, |
| "mask_token_id": 31, |
| "max_position_embeddings": 256, |
| "mlm_activation": "gelu", |
| "model_type": "balm_moe", |
| "num_attention_heads": 20, |
| "num_experts": 8, |
| "num_experts_per_tok": 2, |
| "num_hidden_layers": 30, |
| "num_initial_dense_layers": 1, |
| "num_shared_experts": 0, |
| "output_classifier_attentions": false, |
| "output_expert_indexes": false, |
| "output_router_logits": false, |
| "pad_token_id": 1, |
| "position_embedding_type": "rotary", |
| "router_aux_loss_coef": 0.01, |
| "router_bias": false, |
| "router_dtype": "float32", |
| "router_dynamic_loss_coef": 0.0001, |
| "router_jitter": 0.0, |
| "router_mask_aux_loss": true, |
| "router_mask_pad_logits": false, |
| "router_mask_pad_probs": true, |
| "router_penalty_loss_coef": 0.1, |
| "router_type": "top-k", |
| "router_use_penalty_loss": false, |
| "router_z_loss_coef": 0.001, |
| "shared_expert_intermediate_size": 2560, |
| "top_p_threshold": 0.7, |
| "torch_dtype": "float32", |
| "transformers_version": "4.51.0", |
| "use_cache": true, |
| "vocab_size": 32 |
| } |
|
|
