f4b3f7e96e6d78edcc712781656ef0aa

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

• Loss: 0.4000
• Data Size: 1.0
• Epoch Runtime: 20.5374
• Accuracy: 0.9204
• F1 Macro: 0.8778

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	1.7160	0	1.4263	0.1648	0.0832
No log	1	500	1.5913	0.0078	1.8653	0.2913	0.0805
No log	2	1000	1.5800	0.0156	1.7604	0.3493	0.0868
No log	3	1500	1.5109	0.0312	2.4209	0.3503	0.0886
No log	4	2000	1.1608	0.0625	2.7625	0.5817	0.2402
0.0707	5	2500	0.8713	0.125	3.9718	0.7026	0.4609
0.5561	6	3000	0.4284	0.25	6.3407	0.8574	0.7385
0.048	7	3500	0.3060	0.5	11.1289	0.9032	0.8526
0.2222	8.0	4000	0.2181	1.0	21.6270	0.9214	0.8737
0.1568	9.0	4500	0.2477	1.0	20.2584	0.9153	0.8762
0.1496	10.0	5000	0.2057	1.0	20.2410	0.9309	0.8961
0.131	11.0	5500	0.1953	1.0	20.7297	0.9294	0.8888
0.1234	12.0	6000	0.1912	1.0	20.8098	0.9269	0.8856
0.1211	13.0	6500	0.1740	1.0	20.9895	0.9279	0.8884
0.1135	14.0	7000	0.2479	1.0	20.7696	0.9284	0.8842
0.0859	15.0	7500	0.2452	1.0	20.7262	0.9244	0.8804
0.0679	16.0	8000	0.3947	1.0	20.5836	0.9259	0.8892
0.0809	17.0	8500	0.4000	1.0	20.5374	0.9204	0.8778

Framework versions

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