|
{ |
|
"_name_or_path": "kuleshov-group/bd3lm-owt-block_size4", |
|
"architectures": [ |
|
"BD3LM" |
|
], |
|
"attn_backend": "sdpa", |
|
"auto_map": { |
|
"AutoConfig": "configuration_bd3lm.BD3LMConfig", |
|
"AutoModelForMaskedLM": "modeling_bd3lm.BD3LM" |
|
}, |
|
"block_size": 4, |
|
"cond_dim": 128, |
|
"cross_attn": true, |
|
"dropout": 0.1, |
|
"hidden_dim": 768, |
|
"model_length": 1024, |
|
"model_type": "bd3lm", |
|
"n_blocks": 12, |
|
"n_heads": 12, |
|
"return_dict": false, |
|
"sampling_eps_max": 0.999, |
|
"sampling_eps_min": 0.001, |
|
"time_conditioning": false, |
|
"torch_dtype": "float32", |
|
"transformers_version": "4.49.0", |
|
"var_min": true, |
|
"vocab_size": 50258 |
|
} |
|
|