gemma2b-summarize-gemini1_5flash-256k

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

• Loss: 2.5681

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: 8
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• gradient_accumulation_steps: 2
• total_train_batch_size: 128
• total_eval_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss
1.0246	0.9976	207	2.4550
0.9556	2.0	415	2.4530
0.9114	2.9976	622	2.4641
0.8927	4.0	830	2.4882
0.8752	4.9976	1037	2.5081
0.8602	6.0	1245	2.5277
0.8464	6.9976	1452	2.5513
0.8353	8.0	1660	2.5615
0.8267	8.9976	1867	2.5674
0.8289	9.9976	2070	2.5681

Framework versions

• PEFT 0.11.1
• Transformers 4.41.2
• Pytorch 2.3.1+cu121
• Datasets 2.19.2
• Tokenizers 0.19.1