arch: hnet bp_per_token: 32 max_len: 12800 max_length: 12800 min_routing_tokens: 8 r_hi: 0.3 r_low: 0.0 r_warm_up_end: 750 r_warm_up_start: 200 reference_loss: null region_info: promoter1_cds1_utr2_exon2_intron8_nig8_dig16 strictness_exp: 1.0 strictness_max: 1.0 tokenizer: fast use_routing_floor: true model_cfg: arch_layout: - m2 - - m2 - - m15 - m2 - m2 attn_cfg: num_heads: - 8 - 8 - 12 rotary_emb_dim: - 16 - 16 - 24 window_size: - 511 - 511 - -1 d_intermediate: - 0 - 0 - 2048 d_model: - 512 - 512 - 768 min_routing_tokens: 8 n_gpt: 1.0 r_hi: 0.3 r_low: 0.0 r_warm_up_end: 750 r_warm_up_start: 200 ssm_cfg: chunk_size: 256 d_conv: 4 d_state: 64 expand: 2 head_dim: 64 tie_embeddings: true vocab_size: 12 hf_user: andyjzhao hf_repo: andyjzhao/GeneZip-70M-Transcript-balanced model_name: GeneZip-70M-Transcript-balanced upload_to_hf: false routing_floor: k_min_list: - 8