27afceb9392c23d1674a5a2901ca7d34

This model is a fine-tuned version of albert/albert-xxlarge-v2 on the dair-ai/emotion [split] dataset. It achieves the following results on the evaluation set:

• Loss: 0.3531
• Data Size: 1.0
• Epoch Runtime: 41.1003
• Accuracy: 0.9088
• F1 Macro: 0.8659

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	2.2406	0	1.9154	0.0938	0.0641
No log	1	500	1.7164	0.0078	2.2981	0.2359	0.1245
No log	2	1000	1.7238	0.0156	2.6170	0.2747	0.1135
No log	3	1500	1.5864	0.0312	3.4158	0.3508	0.0944
No log	4	2000	1.5906	0.0625	4.7034	0.3054	0.1008
0.0883	5	2500	1.4474	0.125	7.0518	0.5050	0.2052
1.3713	6	3000	1.3480	0.25	12.0165	0.5469	0.2198
0.1954	7	3500	1.0987	0.5	21.9685	0.6144	0.2500
0.4637	8.0	4000	0.4034	1.0	41.8969	0.8831	0.7363
0.4674	9.0	4500	0.4547	1.0	41.4539	0.8866	0.7876
0.3519	10.0	5000	0.3257	1.0	41.3873	0.9057	0.8543
0.2821	11.0	5500	0.2998	1.0	41.3973	0.9113	0.8625
0.2597	12.0	6000	0.2644	1.0	41.4418	0.9148	0.8722
0.2649	13.0	6500	0.3314	1.0	41.2494	0.9158	0.8698
0.255	14.0	7000	0.2704	1.0	41.4583	0.9113	0.8641
0.1864	15.0	7500	0.3561	1.0	41.3138	0.9178	0.8721
0.2955	16.0	8000	0.3531	1.0	41.1003	0.9088	0.8659

Framework versions

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