| attention_logit_softcapping: 50.0 | |
| attention_scores_scalar: 256 | |
| bias: false | |
| block_size: 8192 | |
| final_logit_softcapping: 30.0 | |
| gelu_approximate: tanh | |
| head_size: 256 | |
| hf_config: | |
| name: gemma-2-2b | |
| org: google | |
| intermediate_size: 9216 | |
| lm_head_bias: false | |
| mlp_class_name: GemmaMLP | |
| n_embd: 2304 | |
| n_expert: 0 | |
| n_expert_per_token: 0 | |
| n_head: 8 | |
| n_layer: 26 | |
| n_query_groups: 4 | |
| name: Gemma-2-2b | |
| norm_class_name: RMSNorm | |
| norm_eps: 1.0e-05 | |
| padded_vocab_size: 256000 | |
| padding_multiple: 512 | |
| parallel_residual: false | |
| post_attention_norm: true | |
| post_mlp_norm: true | |
| rope_base: 10000 | |
| rope_condense_ratio: 1 | |
| rotary_percentage: 1.0 | |
| scale_embeddings: true | |
| shared_attention_norm: false | |
| sliding_window_layer_placing: 2 | |
| sliding_window_size: 4096 | |
| use_flash_attn: true | |
| vocab_size: 256000 | |