|
{ |
|
"architectures": [ |
|
"AnemoneForCausalLM" |
|
], |
|
"attention_dropout": 0.0, |
|
"attn_layer_offset": 5, |
|
"attn_layer_period": 6, |
|
"attn_num_experts": 16, |
|
"attn_router_aux_loss_coef": 0.05, |
|
"attn_top_k": 4, |
|
"bos_token_id": 1, |
|
"calc_logits_for_entire_prompt": true, |
|
"capacity": 128, |
|
"eos_token_id": 2, |
|
"expert_layer_offset": 1, |
|
"expert_layer_period": 2, |
|
"hidden_act": "silu", |
|
"hidden_size": 1024, |
|
"initializer_range": 0.02, |
|
"intermediate_size": 2512, |
|
"mamba_conv_bias": true, |
|
"mamba_d_conv": 4, |
|
"mamba_d_state": 16, |
|
"mamba_dt_rank": 256, |
|
"mamba_expand": 2, |
|
"mamba_inner_layernorms": true, |
|
"mamba_proj_bias": false, |
|
"mod_aux_loss_coef": 0.01, |
|
"mod_aux_routing": false, |
|
"mod_routing": true, |
|
"n_ctx": 262144, |
|
"num_attention_heads": 32, |
|
"num_experts": 8, |
|
"num_experts_per_tok": 2, |
|
"num_hidden_layers": 22, |
|
"num_key_value_heads": 8, |
|
"output_router_logits": true, |
|
"pad_token_id": 0, |
|
"rms_norm_eps": 1e-06, |
|
"router_aux_loss_coef": 0.001, |
|
"skip_blocks": 2, |
|
"sliding_window": null, |
|
"tie_word_embeddings": false, |
|
"torch_dtype": "bfloat16", |
|
"transformers_version": "4.39.3", |
|
"use_cache": true, |
|
"use_mamba_kernels": true, |
|
"vocab_size": 65536 |
|
} |
|
|