FunctionGemma PocketAssist 270M
Fine-tuned FunctionGemma 270M for PocketAssist.
Accuracy: 35.0% on 440 examples
| Function | Accuracy |
|---|---|
| get_password | 0.0% |
| save_password | 80.0% |
| search | 30.0% |
Training
- Base:
unsloth/functiongemma-270m-it - Method: QLoRA 4-bit (Unsloth) + sequence packing
- LoRA rank: 16, alpha: 32
- LR: 0.0002, epochs: 5
- seq_len: 256, batch: 4×4
- Hardware: Colab T4 (16 GB VRAM)
Usage
from unsloth import FastLanguageModel
model, tokenizer = FastLanguageModel.from_pretrained(
"2796gauravc/functiongemma-pocketassist-270m",
load_in_4bit=True,
)
- Downloads last month
- 5
Inference Providers
NEW
This model isn't deployed by any Inference Provider.
🙋
Ask for provider support
Model tree for 2796gauravc/functiongemma-pocketassist-270m
Base model
google/functiongemma-270m-it