README.md

---
license: apache-2.0
datasets:
- pankajmathur/WizardLM_Orca
- teknium/trismegistus-project
- unalignment/toxic-dpo-v0.1
- Intel/orca_dpo_pairs
language:
- en
pipeline_tag: text-generation
---

## Mistral 7b Reverse Instruct

This model is sft (LoRA) fine tuned to reverse engineer the original prompt of a given LLM output/response.    
Use Case: The generation of synthetic instruct datasets for developing chatbots and domain specific fine tuning (e.g. "Summarization" & "Roleplay").


- base_model: mistralai/Mistral-7B-v0.1 (=checkpoint-v1)
- base_model: mistralai/Mistral-7B-v0.2 (>=checkpoint-v2)

For convinience the latest model export is provided under [/latest_model_export](https://huggingface.co/Philipp-Sc/mistral-7b-reverse-instruct/tree/main/latest_model_export) as well as gguf quantized versions under [/latest_ggml_models](https://huggingface.co/Philipp-Sc/mistral-7b-reverse-instruct/tree/main/latest_ggml_models)

## Response Format

"[INST]\n### System:\n{system}\n### Instruction:\n{instruction}\n[/INST]\n"

- Grammar File: [inst_format.gbnf](https://huggingface.co/Philipp-Sc/mistral-7b-reverse-instruct/blob/main/inst_format.gbnf)


## Prompt Template

"\n### System:\nYou craft instructions for generating the given output through reverse engineering.\n### Instruction:\nDecipher the steps used to produce the given output and articulate a refined set of instructions (System & Instruction).\n### OUTPUT:\n {output}"

(use the template without the " ")

## Training Dataset

About 21k items of the following datasets were used. (mostly coding-like tasks were removed)

- v1 & v2: [reverse-instruct_v1.json](https://huggingface.co/Philipp-Sc/mistral-7b-reverse-instruct/blob/main/reverse-instruct_v1.json)
- v3: [reverse-instruct_v2.json](https://huggingface.co/Philipp-Sc/mistral-7b-reverse-instruct/blob/main/reverse-instruct_v2.json)

The reverse instruct dataset has been compiled with entries from the following datasets:

- [alpaca_gpt4_data](https://raw.githubusercontent.com/Instruction-Tuning-with-GPT-4/GPT-4-LLM/main/data/alpaca_gpt4_data.json)
- [roleplay-instruct-v2.1](https://raw.githubusercontent.com/teknium1/GPTeacher/main/Roleplay%20Supplemental/roleplay-instruct-v2.1.json)
- [wizardlm_orca](https://huggingface.co/datasets/pankajmathur/WizardLM_Orca/resolve/main/wizardlm_orca.json)
- [toxic-dpo-v0.1](https://huggingface.co/datasets/unalignment/toxic-dpo-v0.1/resolve/main/toxic-dpo.parquet)
- [orca_dpo_pairs](https://huggingface.co/datasets/Intel/orca_dpo_pairs/resolve/main/orca_rlhf.jsonl)
- [occultexpert](https://huggingface.co/datasets/teknium/trismegistus-project/resolve/main/occultexpert.json)

## Training Procedure

```bash
!cd LLaMA-Factory && WANDB_DISABLED=True PYTORCH_CUDA_ALLOC_CONF=max_split_size_mb:256 accelerate launch \
    --multi_gpu \
    --mixed_precision fp16 \
    --num_processes 2 \
    --num_machines 1 \
    --rdzv_backend static \
    --same_network \
    --gpu_ids all \
    --machine_rank 0 \
    --main_training_function main \
    --  src/train_bash.py  \
    --stage sft \
    --model_name_or_path mistralai/Mistral-7B-Instruct-v0.2 \
    --adapter_name_or_path path_to_checkpoint \
    --flash_attn \
    --neftune_noise_alpha 5 \
    --do_train \
    --dataset default \
    --template vanilla \
    --finetuning_type lora \
    --lora_target q_proj,v_proj \
    --output_dir path_to_sft_checkpoint \
    --overwrite_cache \
    --per_device_train_batch_size 1 \
    --gradient_accumulation_steps 1 \
    --lr_scheduler_type cosine \
    --logging_steps 10 \
    --save_steps 10 \
    --save_total_limit 3 \
    --learning_rate 5e-5 \
    --num_train_epochs 9.0 \
    --plot_loss \
    --fp16 \
    --overwrite_output_dir \
    --cutoff_len 4096 \
    --quantization_bit 4
```

## Training Time

- v1: ~12h on Kaggle's P100 GPU
- v2: >30h on Kaggle's T4 x2
- v3: coming soon

### Framework versions

- LLaMA-Factory