ShengdingHu/compacter_t5-base_cola
015
1{2 "backbone_checkpoint_name": "t5-base",3 "backbone_class": "T5ForConditionalGeneration",4 "backbone_hash": "6d029f2b782e1049f4ca25890fc9b113",5 "bottleneck_dim": null,6 "common_structure": true,7 "delta_type": "compacter",8 "factorized_phm": true,9 "factorized_phm_rule": false,10 "hypercomplex_division": 4,11 "hypercomplex_nonlinearity": "glorot-uniform",12 "kronecker_prod": null,13 "learn_phm": true,14 "modified_modules": [15 "attn",16 "ff"17 ],18 "non_linearity": "gelu_new",19 "opendelta_version": "0.0.1",20 "phm_c_init": "normal",21 "phm_init_range": 0.0001,22 "phm_rank": 1,23 "reduction_factor": 16,24 "sequential": null,25 "shared_W_phm": false,26 "shared_phm_rule": false,27 "transformers_version": "4.16.0.dev0",28 "unfrozen_modules": [29 "deltas",30 "layer_norm",31 "final_layer_norm"32 ],33 "use_bias_down_sampler": true,34 "use_bias_up_sampler": true35}36 