a45bb74f6685a7934ae6a58d57b22954

This model is a fine-tuned version of studio-ousia/mluke-large on the contemmcm/trec dataset. It achieves the following results on the evaluation set:

• Loss: 0.2309
• Data Size: 1.0
• Epoch Runtime: 33.4905
• Accuracy: 0.9646
• F1 Macro: 0.9596

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: 8
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• total_train_batch_size: 32
• total_eval_batch_size: 32
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: constant
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Data Size	Epoch Runtime	Accuracy	F1 Macro
No log	0	0	1.7715	0	1.5374	0.2437	0.1334
No log	1	170	1.7500	0.0078	2.3190	0.2062	0.1120
No log	2	340	1.6979	0.0156	3.4232	0.2875	0.1029
No log	3	510	1.7055	0.0312	4.3840	0.1875	0.0735
No log	4	680	1.5677	0.0625	5.9086	0.3729	0.3270
0.096	5	850	1.3562	0.125	8.0630	0.8292	0.6949
0.096	6	1020	0.4455	0.25	12.3315	0.9104	0.7669
0.4128	7	1190	0.2466	0.5	19.7750	0.9333	0.7906
0.2329	8.0	1360	0.1743	1.0	36.2543	0.9604	0.9353
0.2014	9.0	1530	0.2021	1.0	33.0901	0.9625	0.9526
0.1881	10.0	1700	0.2480	1.0	33.4578	0.9563	0.9538
0.071	11.0	1870	0.1227	1.0	34.3113	0.9729	0.9666
0.1018	12.0	2040	0.2922	1.0	34.1444	0.9604	0.9649
0.0853	13.0	2210	0.1996	1.0	35.1034	0.9625	0.9649
0.0709	14.0	2380	0.1810	1.0	33.0739	0.9667	0.9607
0.0924	15.0	2550	0.2309	1.0	33.4905	0.9646	0.9596

Framework versions

• Transformers 4.57.0
• Pytorch 2.8.0+cu128
• Datasets 4.3.0
• Tokenizers 0.22.1