metadata
datasets:
- yahma/alpaca_cleaned
- lksy/ru_instruct_gpt4
language:
- ru
pipeline_tag: text2text-generation
inference: false
Based on LLaMA 30B.
Trained on 4 LoRA modules.
Parameters:
{
"base_model_name_or_path": "./llama-30b-hf",
"bias": "none",
"enable_lora": null,
"fan_in_fan_out": false,
"inference_mode": true,
"lora_alpha": 16,
"lora_dropout": 0.05,
"merge_weights": false,
"modules_to_save": null,
"peft_type": "LORA",
"r": 16,
"target_modules": [
"q_proj",
"v_proj",
"k_proj",
"o_proj"
],
"task_type": "CAUSAL_LM"
}
Cutoff length set to 512
Prompt template:
{
"description": "A shorter template to experiment with.",
"prompt_input": "### Задание:\n{instruction}\n\n### Вход:\n{input}\n\n### Ответ:\n",
"prompt_no_input": "### Задание:\n{instruction}\n\n### Ответ:\n",
"response_split": "### Ответ:"
}
Epochs: 3
Loss: 0.774 (Might be overfit a bit, try to use a checkpoint)