collapse_gemma-2-2b_hs2_accumulatesubsample_iter9_sftsd0

This model is a fine-tuned version of google/gemma-2-2b on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.1956
• Num Input Tokens Seen: 5023456

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: 8e-06
• train_batch_size: 8
• eval_batch_size: 16
• seed: 0
• gradient_accumulation_steps: 16
• total_train_batch_size: 128
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: constant_with_warmup
• lr_scheduler_warmup_ratio: 0.05
• num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss	Input Tokens Seen
No log	0	0	1.3909	0
1.3304	0.0541	5	1.2738	270256
1.1448	0.1082	10	1.2117	540696
1.0747	0.1623	15	1.2004	821656
1.0687	0.2164	20	1.2026	1098712
0.7675	0.2705	25	1.2266	1366240
0.7907	0.3245	30	1.2142	1643704
0.6795	0.3786	35	1.2293	1908672
0.6652	0.4327	40	1.2036	2181616
0.7058	0.4868	45	1.2256	2456960
0.6919	0.5409	50	1.2012	2735528
0.6422	0.5950	55	1.2120	3015304
0.6387	0.6491	60	1.2085	3286712
0.4768	0.7032	65	1.2063	3557760
0.5572	0.7573	70	1.1910	3824824
0.5535	0.8114	75	1.2021	4097760
0.4666	0.8654	80	1.1937	4378176
0.4766	0.9195	85	1.1977	4649040
0.5041	0.9736	90	1.1999	4914024

Framework versions

• Transformers 4.44.0
• Pytorch 2.4.0+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1