This model was trained from scratch on the kanishka/counterfactual-babylm-pipps_and_keys_to_it_all_removal dataset. It achieves the following results on the evaluation set:

Loss: 3.4099
Accuracy: 0.4102

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 0.001
train_batch_size: 32
eval_batch_size: 64
seed: 1024
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 32000
num_epochs: 20.0
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
3.6061	1.0	18594	3.7643	0.3602
3.3838	2.0	37188	3.5759	0.3804
3.259	3.0	55782	3.4653	0.3916
3.1772	4.0	74376	3.3950	0.3993
3.1278	5.0	92970	3.4000	0.4015
3.0787	6.0	111564	3.3889	0.4041
3.049	7.0	130158	3.3549	0.4067
3.0111	8.0	148752	3.3748	0.4067
2.9822	9.0	167346	3.3543	0.4089
2.9636	10.0	185940	3.3444	0.4092
2.938	11.0	204534	3.3562	0.4088
2.9153	12.0	223128	3.3650	0.4095
2.8913	13.0	241722	3.3697	0.4100
2.8725	14.0	260316	3.3670	0.4097
2.8528	15.0	278910	3.3746	0.4102
2.8293	16.0	297504	3.3919	0.4095
2.8131	17.0	316098	3.3852	0.4103
2.792	18.0	334692	3.3966	0.4101
2.7761	19.0	353286	3.3991	0.4106
2.7551	20.0	371880	3.4099	0.4102

Framework versions

Transformers 4.37.2
Pytorch 2.1.0+cu121
Datasets 2.16.1
Tokenizers 0.15.1

kanishka/smolm-autoreg-bpe-counterfactual-babylm-pipps_and_keys_to_it_all_removal-seed_1024-1e-3

smolm-autoreg-bpe-counterfactual-babylm-pipps_and_keys_to_it_all_removal-seed_1024-1e-3

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

kanishka/
smolm-autoreg-bpe-counterfactual-babylm-pipps_and_keys_to_it_all_removal-seed_1024-1e-3