|
{ |
|
"architectures": [ |
|
"GPTNeoXCodebookModel" |
|
], |
|
"codebook_at": [ |
|
"preproj_attention", |
|
"mlp" |
|
], |
|
"codebook_kwargs": {}, |
|
"codebook_type": [ |
|
"group", |
|
"vanilla" |
|
], |
|
"k_codebook": [ |
|
1, |
|
1 |
|
], |
|
"kmeans_init": false, |
|
"kmeans_init_examples": 1000, |
|
"kmeans_kwargs": null, |
|
"kmeans_path": null, |
|
"layers_to_snap": [ |
|
0, |
|
1, |
|
2, |
|
3 |
|
], |
|
"loss": "aeloss", |
|
"model_type": "codebook", |
|
"num_codebooks": [ |
|
4, |
|
1 |
|
], |
|
"num_codes": 10000, |
|
"replace_codes": false, |
|
"similarity_metric": "inner_product", |
|
"torch_dtype": "float32", |
|
"transformers_version": "4.28.1" |
|
} |
|
|