|
--- |
|
datasets: |
|
- its5Q/panorama |
|
language: |
|
- ru |
|
library_name: peft |
|
pipeline_tag: text-generation |
|
tags: |
|
- llama |
|
- llama-2 |
|
- news |
|
--- |
|
|
|
# Llama 2 7B, fine-tuned on Panorama media |
|
|
|
This repo contains the QLoRA adapter. |
|
|
|
Prompt: |
|
``` |
|
Write a hypothetical news story based on the given headline |
|
|
|
### Title: |
|
{prompt} |
|
|
|
Text: |
|
``` |
|
|
|
## 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: nf4 |
|
- bnb_4bit_use_double_quant: False |
|
- bnb_4bit_compute_dtype: float16 |
|
### Framework versions |
|
|
|
|
|
- PEFT 0.5.0.dev0 |
|
|
|
|
|
### Additional information |
|
|
|
Thanks [its5Q](https://huggingface.co/its5Q) for dataset and help |
