| { |
| "architectures": [ |
| "Qwen2MoeForCausalLM" |
| ], |
| "attention_dropout": 0.0, |
| "decoder_sparse_step": 1, |
| "hidden_act": "silu", |
| "hidden_dropout": 0.0, |
| "hidden_size": 256, |
| "initializer_range": 0.02, |
| "intermediate_size": 1024, |
| "max_position_embeddings": 2048, |
| "max_window_layers": 28, |
| "mlp_only_layers": [], |
| "model_type": "qwen2_moe", |
| "moe_intermediate_size": 64, |
| "norm_topk_prob": false, |
| "num_attention_heads": 2, |
| "num_experts": 8, |
| "num_experts_per_tok": 4, |
| "num_hidden_layers": 8, |
| "num_key_value_heads": 1, |
| "output_router_logits": false, |
| "qkv_bias": true, |
| "rms_norm_eps": 1e-06, |
| "rope_scaling": null, |
| "rope_theta": 10000.0, |
| "router_aux_loss_coef": 0.001, |
| "shared_expert_intermediate_size": 512, |
| "sliding_window": null, |
| "tie_word_embeddings": true, |
| "torch_dtype": "bfloat16", |
| "transformers_version": "4.54.0.dev0", |
| "use_cache": true, |
| "use_sliding_window": false, |
| "vocab_size": 128256 |
| } |
|
|
