|
{ |
|
"backbone_checkpoint_name": "t5-large", |
|
"backbone_class": "T5ForConditionalGeneration", |
|
"backbone_hash": "6297bd1acc36524547c8a76cc03fef5c", |
|
"bottleneck_dim": null, |
|
"common_structure": null, |
|
"delta_type": "compacter", |
|
"factorized_phm": true, |
|
"factorized_phm_rule": false, |
|
"hypercomplex_division": 4, |
|
"hypercomplex_nonlinearity": "glorot-uniform", |
|
"kronecker_prod": null, |
|
"learn_phm": true, |
|
"modified_modules": [ |
|
"SelfAttention", |
|
"DenseReluDense" |
|
], |
|
"non_linearity": "gelu_new", |
|
"opendelta_version": "0.0.1", |
|
"phm_c_init": "normal", |
|
"phm_init_range": 0.0001, |
|
"phm_rank": 1, |
|
"reduction_factor": 16, |
|
"sequential": null, |
|
"shared_W_phm": false, |
|
"shared_phm_rule": false, |
|
"transformers_version": "4.17.0", |
|
"use_bias_down_sampler": true, |
|
"use_bias_up_sampler": true |
|
} |
|
|