InkubaLM-0.5B-multi-instruct

This model is a fine-tuned version of lelapa/InkubaLM-0.4B on the None dataset. It achieves the following results on the evaluation set:

• Loss: 2.6574

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: 0.0002
• train_batch_size: 1
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 16
• total_train_batch_size: 16
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 100
• num_epochs: 4
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
3.6403	0.1572	200	3.6194
3.3373	0.3145	400	3.4392
3.1929	0.4717	600	3.2638
3.1726	0.6290	800	3.1232
3.0448	0.7862	1000	3.0049
2.8797	0.9435	1200	2.9117
2.3500	1.1006	1400	2.8947
2.2215	1.2579	1600	2.8335
2.2400	1.4151	1800	2.7583
2.1957	1.5724	2000	2.6760
2.1691	1.7296	2200	2.5903
2.1501	1.8869	2400	2.5254
1.2087	2.0440	2600	2.6767
1.1884	2.2013	2800	2.6625
1.2160	2.3585	3000	2.6542
1.2006	2.5158	3200	2.6120
1.2054	2.6730	3400	2.5835
1.1190	2.8303	3600	2.5548
1.0761	2.9875	3800	2.5319
0.4459	3.1447	4000	2.6569
0.4153	3.3019	4200	2.6612
0.4336	3.4592	4400	2.6602
0.3943	3.6164	4600	2.6574
0.4209	3.7737	4800	2.6583
0.4042	3.9309	5000	2.6574

Framework versions

• Transformers 5.0.0
• Pytorch 2.10.0+cu128
• Datasets 4.0.0
• Tokenizers 0.22.2