SmolLM2-1.7B-TemporalQuestions

This model is a fine-tuned version of HuggingFaceTB/SmolLM2-1.7B on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.0287
• F1: 0.9837

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: 4
• eval_batch_size: 4
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 128
• total_train_batch_size: 2048
• total_eval_batch_size: 16
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine_with_restarts
• lr_scheduler_warmup_ratio: 0.05
• num_epochs: 30

Training results

Training Loss	Epoch	Step	Validation Loss	F1
0.2142	1.0	112	0.1175	0.9066
0.1218	2.0	224	0.0498	0.9606
0.0219	3.0	336	0.0363	0.9755
0.1146	4.0	448	0.0350	0.9749
0.0062	5.0	560	0.0297	0.9775
0.079	6.0	672	0.0361	0.9767
0.0261	7.0	784	0.0297	0.9818
0.0225	8.0	896	0.0290	0.9804
0.0004	9.0	1008	0.0308	0.9816
0.0105	10.0	1120	0.0333	0.9829
0.0014	11.0	1232	0.0287	0.9837
0.0035	12.0	1344	0.0301	0.9848
0.0035	13.0	1456	0.0290	0.9835
0.0785	14.0	1568	0.0291	0.9846
0.0014	15.0	1680	0.0303	0.9840
0.0003	16.0	1792	0.0356	0.9837
0.0002	17.0	1904	0.0347	0.9862
0.0003	18.0	2016	0.0338	0.9865
0.0002	19.0	2128	0.0335	0.9872
0.0001	20.0	2240	0.0352	0.9878
0.0001	21.0	2352	0.0362	0.9878
0.0001	22.0	2464	0.0378	0.9875
0.0001	23.0	2576	0.0386	0.9874
0.0001	24.0	2688	0.0391	0.9873
0.0001	25.0	2800	0.0394	0.9874
0.0001	26.0	2912	0.0396	0.9874
0.0001	27.0	3024	0.0397	0.9874
0.0001	28.0	3136	0.0397	0.9874
0.0001	29.0	3248	0.0398	0.9874
0.0036	29.7385	3330	0.0398	0.9874

Framework versions

• Transformers 4.47.1
• Pytorch 2.5.1+cu124
• Datasets 3.0.1
• Tokenizers 0.21.0