db26969bbadbde51d229e2e3eef0b5f9

This model is a fine-tuned version of google/gemma-2b on the dair-ai/emotion [split] dataset. It achieves the following results on the evaluation set:

• Loss: 0.9766
• Data Size: 1.0
• Epoch Runtime: 95.6605
• Accuracy: 0.9315
• F1 Macro: 0.8954

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	19.6997	0	4.2534	0.0348	0.0160
No log	1	500	8.1670	0.0078	4.3495	0.3765	0.1546
No log	2	1000	5.3427	0.0156	6.6627	0.5408	0.2870
No log	3	1500	2.7666	0.0312	10.0068	0.7737	0.6525
No log	4	2000	2.1499	0.0625	12.6556	0.8417	0.7662
0.1755	5	2500	2.1083	0.125	17.8891	0.8498	0.8171
1.2868	6	3000	1.0770	0.25	28.7037	0.9062	0.8025
0.1437	7	3500	1.1153	0.5	50.5533	0.9088	0.8722
0.6542	8.0	4000	0.6933	1.0	96.2773	0.9239	0.8648
0.6561	9.0	4500	0.7342	1.0	95.6056	0.9239	0.8815
0.5744	10.0	5000	0.5925	1.0	93.6717	0.9355	0.8975
0.4555	11.0	5500	0.6979	1.0	97.3063	0.9309	0.8936
0.4072	12.0	6000	0.8161	1.0	97.3904	0.9309	0.8789
0.5086	13.0	6500	0.8292	1.0	93.9134	0.9234	0.8750
0.4498	14.0	7000	0.9766	1.0	95.6605	0.9315	0.8954

Framework versions

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