Cabra: A portuguese finetuned instruction Open-LLaMA

LoRA adapter created with the procedures detailed at the GitHub repository: https://github.com/gustrd/cabra .

This training was done at 2 epochs using one A4000 at Paperspace.

The GGML version was created with llama.cpp "convert-lora-to-ggml.py".

This LoRA adapter was created following the procedure

library_name: peft

Training procedure

The following bitsandbytes quantization config was used during training:

• load_in_8bit: False
• load_in_4bit: True
• 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 0.5.0.dev0