train_mnli_42_1775732963

This model is a fine-tuned version of meta-llama/Llama-3.2-1B-Instruct on the mnli dataset. It achieves the following results on the evaluation set:

• Loss: 0.1219
• Num Input Tokens Seen: 191491960

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: 5e-06
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• 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
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Input Tokens Seen
0.1769	0.2500	11045	0.1287	9565376
0.0872	0.5000	22090	0.1304	19168640
0.1314	0.7500	33135	0.1342	28714304
0.079	1.0000	44180	0.1219	38289824
0.0622	1.2500	55225	0.1576	47877216
0.0476	1.5000	66270	0.1587	57416032
0.0615	1.7500	77315	0.1618	66982176
0.0343	2.0000	88360	0.1496	76602496
0.0005	2.2501	99405	0.2128	86154496
0.0401	2.5001	110450	0.2214	95709312
0.0901	2.7501	121495	0.2072	105304960
0.0894	3.0001	132540	0.2248	114898176
0.0	3.2501	143585	0.2610	124468928
0.0002	3.5001	154630	0.2781	134028992
0.0001	3.7501	165675	0.2688	143607232
0.0001	4.0001	176720	0.2775	153206432
0.0	4.2501	187765	0.3348	162770528
0.0	4.5001	198810	0.3502	172345120
0.0	4.7501	209855	0.3450	181948192

Framework versions

• Transformers 4.51.3
• Pytorch 2.10.0+cu128
• Datasets 4.0.0
• Tokenizers 0.21.4