|
{ |
|
"architectures": [ |
|
"GPTNeoCodebookModel" |
|
], |
|
"codebook_at": [ |
|
"preproj_attention" |
|
], |
|
"codebook_kwargs": {}, |
|
"codebook_type": [ |
|
"compositional" |
|
], |
|
"k_codebook": [ |
|
8 |
|
], |
|
"kmeans_init": false, |
|
"kmeans_init_examples": 1000, |
|
"kmeans_kwargs": { |
|
"batch_size": 24576, |
|
"n_init": "auto" |
|
}, |
|
"kmeans_path": "/.cache/cb_volume/huggingface/kmeans_embeddings.pt", |
|
"layers_to_snap": [ |
|
0 |
|
], |
|
"loss": "aeloss", |
|
"model_type": "codebook", |
|
"num_codebooks": [ |
|
16 |
|
], |
|
"num_codes": 25000, |
|
"replace_codes": false, |
|
"similarity_metric": "inner_product", |
|
"torch_dtype": "float32", |
|
"transformers_version": "4.29.2" |
|
} |
|
|