train_wic_1753094168

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

• Loss: 0.2399
• Num Input Tokens Seen: 4213808

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-05
• train_batch_size: 4
• eval_batch_size: 4
• seed: 123
• 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: 10.0

Training results

Training Loss	Epoch	Step	Validation Loss	Input Tokens Seen
0.4151	0.5	611	0.3295	210240
0.2389	1.0	1222	0.2997	421528
0.2716	1.5	1833	0.2854	632632
0.1819	2.0	2444	0.2773	843368
0.3153	2.5	3055	0.2658	1054024
0.252	3.0	3666	0.2634	1264408
0.2181	3.5	4277	0.2576	1475000
0.2114	4.0	4888	0.2570	1685768
0.1929	4.5	5499	0.2561	1895752
0.3225	5.0	6110	0.2448	2106968
0.3734	5.5	6721	0.2439	2318136
0.2293	6.0	7332	0.2432	2528648
0.2122	6.5	7943	0.2424	2739720
0.3024	7.0	8554	0.2423	2949592
0.3562	7.5	9165	0.2420	3160056
0.2596	8.0	9776	0.2408	3371056
0.2664	8.5	10387	0.2399	3581616
0.2097	9.0	10998	0.2412	3792672
0.283	9.5	11609	0.2402	4003136
0.1575	10.0	12220	0.2404	4213808

Framework versions

• PEFT 0.15.2
• Transformers 4.51.3
• Pytorch 2.7.1+cu126
• Datasets 3.6.0
• Tokenizers 0.21.1