hrm-rec-strict / config.json
PatrickHaller's picture
final chck_1000M
936fb6a verified
Raw
History Blame Contribute Delete
871 Bytes
{
"H_cycles": 2,
"H_layers": null,
"L_cycles": 5,
"L_layers": null,
"architectures": [
"HRMTextForCausalLM"
],
"auto_map": {
"AutoConfig": "modeling_hrm.HRMTextConfig",
"AutoModelForCausalLM": "modeling_hrm.HRMTextForCausalLM",
"AutoModel": "modeling_hrm.HRMTextModel"
},
"bos_token_id": 1,
"bp_max_steps": 5,
"bp_min_steps": 2,
"bp_warmup_steps": 20000,
"dtype": "float32",
"eos_token_id": 2,
"expansion": 4.0,
"gdn_head_dim": 128,
"half_layers": true,
"hidden_size": 768,
"init_type": "lecun_normal",
"injection": "gru_l",
"max_seq_len": 1024,
"mixer_householder": 2,
"mixing": "gdn",
"model_type": "hrm_text",
"n_layers": 12,
"norm_eps": 1e-06,
"num_heads": 12,
"pad_token_id": 3,
"rope_theta": 10000.0,
"tie_word_embeddings": false,
"transformers_version": "5.9.0",
"vocab_size": 16384
}