|
--- |
|
license: cc-by-sa-4.0 |
|
datasets: |
|
- izumi-lab/llm-japanese-dataset |
|
language: |
|
- ja |
|
- en |
|
--- |
|
|
|
## This model is a fine-tuned Llama2-7b-chat-hf model with Japanese dataset with LoRA. |
|
# This model is finetuned by the joint efforts of Sparticle Inc. and A. I. Hakusan Inc. |
|
|
|
The training set of this model contains: |
|
|
|
5% of randomly chosen data from llm-japanese-dataset by izumi-lab. |
|
|
|
Japanese-alpaca-lora dataset, retrieved from https://github.com/masa3141/japanese-alpaca-lora/tree/main |
|
|
|
For inference, please follow the instructions in https://github.com/tloen/alpaca-lora/ . |
|
|
|
## Training procedure |
|
|
|
The following `bitsandbytes` quantization config was used during training: |
|
- load_in_8bit: True |
|
- llm_int8_threshold: 6.0 |
|
- llm_int8_skip_modules: None |
|
- llm_int8_enable_fp32_cpu_offload: False |
|
|
|
### Framework versions |
|
|
|
- PEFT 0.5.0.dev0 |
|
|
|
You must agree with Meta's agreements when using this LoRA adapter with Llama-2. |