train_winogrande_101112_1760638069

This model is a fine-tuned version of meta-llama/Meta-Llama-3-8B-Instruct on the winogrande dataset. It achieves the following results on the evaluation set:

• Loss: 0.2501
• Num Input Tokens Seen: 38366624

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: 101112
• optimizer: Use 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: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Input Tokens Seen
0.2313	1.0	9090	0.2313	1917952
0.2347	2.0	18180	0.2315	3835840
0.2279	3.0	27270	0.2303	5753152
0.2312	4.0	36360	0.2299	7672000
0.226	5.0	45450	0.2286	9590080
0.215	6.0	54540	0.2256	11509088
0.2366	7.0	63630	0.2220	13427712
0.1858	8.0	72720	0.2162	15346672
0.2223	9.0	81810	0.2146	17265344
0.2639	10.0	90900	0.2104	19184224
0.2016	11.0	99990	0.2138	21102912
0.1812	12.0	109080	0.2102	23021312
0.1955	13.0	118170	0.2082	24938688
0.2231	14.0	127260	0.2090	26857088
0.1629	15.0	136350	0.2134	28775840
0.2237	16.0	145440	0.2109	30693088
0.1881	17.0	154530	0.2137	32612480
0.2269	18.0	163620	0.2165	34530176
0.1694	19.0	172710	0.2177	36447600
0.1498	20.0	181800	0.2176	38366624

Framework versions

• PEFT 0.17.1
• Transformers 4.51.3
• Pytorch 2.9.0+cu128
• Datasets 4.0.0
• Tokenizers 0.21.4