---
library_name: peft
inference: true
---
## Training procedure

This mode has been trained on almost 20 000 using this [dataset](https://huggingface.co/datasets/Snit/french-conversation) on 2 epochs. 
Graph GPU used for this training: Nvidia A100.

Timing: less than 24 hours.
We're grateful to Nvidia Inception Program for their support!

## Contact
contact@faradaylab.fr
Join us on Discord: https://discord.gg/qc7Ecesqwc

The following `bitsandbytes` quantization config was used during training:
- quant_method: bitsandbytes
- load_in_8bit: True
- load_in_4bit: False
- llm_int8_threshold: 6.0
- llm_int8_skip_modules: None
- llm_int8_enable_fp32_cpu_offload: False
- llm_int8_has_fp16_weight: False
- bnb_4bit_quant_type: fp4
- bnb_4bit_use_double_quant: False
- bnb_4bit_compute_dtype: float32
### Framework versions


- PEFT 0.6.0.dev0