CooperW/gemma_2_27b_it_bnb_4bit_multi

This model has been trained for multi classification of the TON-Iot Network dataset.

Refer to adapter_config.json for LoRa Parameters.

Accuracy: 95.6%

Key Training Params:

• Steps: 100
• Batch Size: 64
• Optimizer: Adam_8bit
• lr_schedular: cosine_with_restarts
• lr_scheduler_kwargs = { "num_cycles": len(trainer.get_train_dataloader()) / 100 }

The datasets are an 80/20 partition of the TON-Iot Network dataset.

Train Dataset:

• train_network_multi.jsonl

Test Dataset:

• test_network_multi.jsonl