smolm-autoreg-bpe-babylm-no_aann-all-det-removal-1e-3

This model is a fine-tuned version of models/smolm-autoreg-bpe-babylm-no_aann-all-det-removal-1e-3/config.json on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 3.1553
• Accuracy: 0.4326

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: 64
• eval_batch_size: 256
• seed: 42
• 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
3.5515	1.0	9181	3.6440	0.3711
3.2765	2.0	18362	3.3933	0.3949
3.1549	3.0	27543	3.3002	0.4061
3.0673	4.0	36724	3.2418	0.4127
2.988	5.0	45905	3.1911	0.4198
2.9271	6.0	55086	3.1648	0.4231
2.883	7.0	64267	3.1442	0.4257
2.842	8.0	73448	3.1381	0.4276
2.8068	9.0	82629	3.1224	0.4291
2.7789	10.0	91810	3.1231	0.4301
2.7503	11.0	100991	3.1208	0.4313
2.7201	12.0	110172	3.1191	0.4319
2.705	13.0	119353	3.1200	0.4324
2.6755	14.0	128534	3.1298	0.4318
2.651	15.0	137715	3.1362	0.4321
2.6293	16.0	146896	3.1347	0.4326
2.6117	17.0	156077	3.1423	0.4324
2.5863	18.0	165258	3.1416	0.4329
2.5659	19.0	174439	3.1501	0.4327
2.5465	20.0	183620	3.1553	0.4326

Framework versions

• Transformers 4.35.0
• Pytorch 2.1.0+cu121
• Datasets 2.12.0
• Tokenizers 0.14.1