You can deploy this model and start earning money today!

nikitastheo/
babylm-zho-ell-interleaved

Name: babylm-zho-ell-interleaved
Availability: InStock

generated_from_trainer

text-generation-inference

endpoints_compatible

Views

No views yet

zho-ell-interleaved

This model is a fine-tuned version of gpt_base_config.json on the data/babybabellm/zho-ell/tokenized_interleaved/train_1.txt dataset. It achieves the following results on the evaluation set:

Loss: 4.5135
Accuracy: 0.2673

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: 0.0001
train_batch_size: 32
eval_batch_size: 32
seed: 42
optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-06 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 1326
training_steps: 13260

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
7.1024	0.4996	663	7.0863	0.1091
6.3502	0.9992	1326	6.3497	0.1313
5.9515	1.4989	1989	5.9362	0.1512
5.6972	1.9985	2652	5.6559	0.1661
5.4566	2.4981	3315	5.4470	0.1792
5.2731	2.9977	3978	5.2699	0.1912
5.04	3.4974	4641	5.1264	0.2038
4.9347	3.9970	5304	4.9856	0.2174
4.7772	4.4966	5967	4.8862	0.2287
4.6904	4.9962	6630	4.8007	0.2372
4.6965	5.4959	7293	4.7361	0.2444
4.5898	5.9955	7956	4.6836	0.2494
4.4765	6.4951	8619	4.6440	0.2537
4.3821	6.9947	9282	4.6052	0.2570
4.3645	7.4943	9945	4.5839	0.2598
4.416	7.9940	10608	4.5563	0.2623
4.3247	8.4936	11271	4.5413	0.2644
4.2423	8.9932	11934	4.5258	0.2658
4.2073	9.4928	12597	4.5187	0.2668
4.2745	9.9925	13260	4.5135	0.2673

Framework versions

Transformers 4.57.6
Pytorch 2.12.0+cu130
Datasets 4.8.5
Tokenizers 0.22.2

nikitastheo/babylm-zho-ell-interleaved

zho-ell-interleaved

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

nikitastheo/
babylm-zho-ell-interleaved