EmoN1

Fine-tuned from google/gemma-3-27b-it using QLoRA.

Training Details

• Base Model: google/gemma-3-27b-it
• Method: QLoRA (4-bit quantization + LoRA)
• LoRA Rank: 32
• LoRA Alpha: 64
• Sequence Length: 8192
• Epochs: 3
• Learning Rate: 2e-4

Training Results

Training Loss	Epoch	Step	Validation Loss
0.9058	1.0	63	0.8959
0.8279	2.0	126	0.8607

Framework Versions

• PEFT 0.17.1
• Transformers 4.55.4
• Pytorch 2.7.1+cu126