---
inference: false
license: other
---

<!-- header start -->
<div style="width: 100%;">
    <img src="https://i.imgur.com/EBdldam.jpg" alt="TheBlokeAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</div>
<div style="display: flex; justify-content: space-between; width: 100%;">
    <div style="display: flex; flex-direction: column; align-items: flex-start;">
        <p><a href="https://discord.gg/theblokeai">Chat & support: my new Discord server</a></p>
    </div>
    <div style="display: flex; flex-direction: column; align-items: flex-end;">
        <p><a href="https://www.patreon.com/TheBlokeAI">Want to contribute? TheBloke's Patreon page</a></p>
    </div>
</div>
<!-- header end -->

# Pankaj Mathur's Orca Mini v2 13B GGML

These files are GGML format model files for [Pankaj Mathur's Orca Mini v2 13B](https://huggingface.co/psmathur/orca_mini_v2_13b).

GGML files are for CPU + GPU inference using [llama.cpp](https://github.com/ggerganov/llama.cpp) and libraries and UIs which support this format, such as:
* [text-generation-webui](https://github.com/oobabooga/text-generation-webui)
* [KoboldCpp](https://github.com/LostRuins/koboldcpp)
* [LoLLMS Web UI](https://github.com/ParisNeo/lollms-webui)
* [llama-cpp-python](https://github.com/abetlen/llama-cpp-python)
* [ctransformers](https://github.com/marella/ctransformers)

## Repositories available

* [4-bit GPTQ models for GPU inference](https://huggingface.co/TheBloke/orca_mini_v2_13b-GPTQ)
* [2, 3, 4, 5, 6 and 8-bit GGML models for CPU+GPU inference](https://huggingface.co/TheBloke/orca_mini_v2_13b-GGML)
* [Unquantised fp16 model in pytorch format, for GPU inference and for further conversions](https://huggingface.co/psmathur/orca_mini_v2_13b)

## Prompt template: orca_mini

```
### System:
You are an AI assistant that follows instruction extremely well. Help as much as you can.

### User:
prompt

### Input:
input, if required

### Response:

```

<!-- compatibility_ggml start -->
## Compatibility

### Original llama.cpp quant methods: `q4_0, q4_1, q5_0, q5_1, q8_0`

These are guaranteed to be compatbile with any UIs, tools and libraries released since late May. They may be phased out soon, as they are largely superseded by the new k-quant methods.

### New k-quant methods: `q2_K, q3_K_S, q3_K_M, q3_K_L, q4_K_S, q4_K_M, q5_K_S, q6_K`

These new quantisation methods are compatible with llama.cpp as of June 6th, commit `2d43387`.

They are now also compatible with recent releases of text-generation-webui, KoboldCpp, llama-cpp-python, ctransformers, rustformers and most others. For compatibility with other tools and libraries, please check their documentation.

## Explanation of the new k-quant methods
<details>
  <summary>Click to see details</summary>

The new methods available are:
* GGML_TYPE_Q2_K - "type-1" 2-bit quantization in super-blocks containing 16 blocks, each block having 16 weight. Block scales and mins are quantized with 4 bits. This ends up effectively using 2.5625 bits per weight (bpw)
* GGML_TYPE_Q3_K - "type-0" 3-bit quantization in super-blocks containing 16 blocks, each block having 16 weights. Scales are quantized with 6 bits. This end up using 3.4375 bpw.
* GGML_TYPE_Q4_K - "type-1" 4-bit quantization in super-blocks containing 8 blocks, each block having 32 weights. Scales and mins are quantized with 6 bits. This ends up using 4.5 bpw.
* GGML_TYPE_Q5_K - "type-1" 5-bit quantization. Same super-block structure as GGML_TYPE_Q4_K resulting in 5.5 bpw
* GGML_TYPE_Q6_K - "type-0" 6-bit quantization. Super-blocks with 16 blocks, each block having 16 weights. Scales are quantized with 8 bits. This ends up using 6.5625 bpw
* GGML_TYPE_Q8_K - "type-0" 8-bit quantization. Only used for quantizing intermediate results. The difference to the existing Q8_0 is that the block size is 256. All 2-6 bit dot products are implemented for this quantization type.

Refer to the Provided Files table below to see what files use which methods, and how.
</details>
<!-- compatibility_ggml end -->

## Provided files
| Name | Quant method | Bits | Size | Max RAM required | Use case |
| ---- | ---- | ---- | ---- | ---- | ----- |
| orca_mini_v2_13b.ggmlv3.q2_K.bin | q2_K | 2 | 5.51 GB| 8.01 GB | New k-quant method. Uses GGML_TYPE_Q4_K for the attention.vw and feed_forward.w2 tensors, GGML_TYPE_Q2_K for the other tensors. |
| orca_mini_v2_13b.ggmlv3.q3_K_S.bin | q3_K_S | 3 | 5.66 GB| 8.16 GB | New k-quant method. Uses GGML_TYPE_Q3_K for all tensors |
| orca_mini_v2_13b.ggmlv3.q3_K_M.bin | q3_K_M | 3 | 6.31 GB| 8.81 GB | New k-quant method. Uses GGML_TYPE_Q4_K for the attention.wv, attention.wo, and feed_forward.w2 tensors, else GGML_TYPE_Q3_K |
| orca_mini_v2_13b.ggmlv3.q3_K_L.bin | q3_K_L | 3 | 6.93 GB| 9.43 GB | New k-quant method. Uses GGML_TYPE_Q5_K for the attention.wv, attention.wo, and feed_forward.w2 tensors, else GGML_TYPE_Q3_K |
| orca_mini_v2_13b.ggmlv3.q4_K_S.bin | q4_K_S | 4 | 7.37 GB| 9.87 GB | New k-quant method. Uses GGML_TYPE_Q4_K for all tensors |
| orca_mini_v2_13b.ggmlv3.q4_K_M.bin | q4_K_M | 4 | 7.87 GB| 10.37 GB | New k-quant method. Uses GGML_TYPE_Q6_K for half of the attention.wv and feed_forward.w2 tensors, else GGML_TYPE_Q4_K |
| orca_mini_v2_13b.ggmlv3.q4_0.bin | q4_0 | 4 | 7.32 GB| 9.82 GB | Original quant method, 4-bit. |
| orca_mini_v2_13b.ggmlv3.q4_1.bin | q4_1 | 4 | 8.14 GB| 10.64 GB | Original quant method, 4-bit. Higher accuracy than q4_0 but not as high as q5_0. However has quicker inference than q5 models. |
| orca_mini_v2_13b.ggmlv3.q5_K_S.bin | q5_K_S | 5 | 8.97 GB| 11.47 GB | New k-quant method. Uses GGML_TYPE_Q5_K for all tensors |
| orca_mini_v2_13b.ggmlv3.q5_K_M.bin | q5_K_M | 5 | 9.23 GB| 11.73 GB | New k-quant method. Uses GGML_TYPE_Q6_K for half of the attention.wv and feed_forward.w2 tensors, else GGML_TYPE_Q5_K |
| orca_mini_v2_13b.ggmlv3.q5_0.bin | q5_0 | 5 | 8.95 GB| 11.45 GB | Original quant method, 5-bit. Higher accuracy, higher resource usage and slower inference. |
| orca_mini_v2_13b.ggmlv3.q5_1.bin | q5_1 | 5 | 9.76 GB| 12.26 GB | Original quant method, 5-bit. Even higher accuracy, resource usage and slower inference. |
| orca_mini_v2_13b.ggmlv3.q6_K.bin | q6_K | 6 | 10.68 GB| 13.18 GB | New k-quant method. Uses GGML_TYPE_Q8_K - 6-bit quantization - for all tensors |
| orca_mini_v2_13b.ggmlv3.q8_0.bin | q8_0 | 8 | 13.83 GB| 16.33 GB | Original quant method, 8-bit. Almost indistinguishable from float16. High resource use and slow. Not recommended for most users. |


**Note**: the above RAM figures assume no GPU offloading. If layers are offloaded to the GPU, this will reduce RAM usage and use VRAM instead.

## How to run in `llama.cpp`

I use the following command line; adjust for your tastes and needs:

```
./main -t 10 -ngl 32 -m ggml.ggmlv3.q5_0.bin --color -c 2048 --temp 0.7 --repeat_penalty 1.1 -n -1 -p "### Instruction: Write a story about llamas\n### Response:"
```
Change `-t 10` to the number of physical CPU cores you have. For example if your system has 8 cores/16 threads, use `-t 8`.

Change `-ngl 32` to the number of layers to offload to GPU. Remove it if you don't have GPU acceleration.

If you want to have a chat-style conversation, replace the `-p <PROMPT>` argument with `-i -ins`

## How to run in `text-generation-webui`

Further instructions here: [text-generation-webui/docs/llama.cpp-models.md](https://github.com/oobabooga/text-generation-webui/blob/main/docs/llama.cpp-models.md).

<!-- footer start -->
## Discord

For further support, and discussions on these models and AI in general, join us at:

[TheBloke AI's Discord server](https://discord.gg/theblokeai)

## Thanks, and how to contribute.

Thanks to the [chirper.ai](https://chirper.ai) team!

I've had a lot of people ask if they can contribute. I enjoy providing models and helping people, and would love to be able to spend even more time doing it, as well as expanding into new projects like fine tuning/training.

If you're able and willing to contribute it will be most gratefully received and will help me to keep providing more models, and to start work on new AI projects.

Donaters will get priority support on any and all AI/LLM/model questions and requests, access to a private Discord room, plus other benefits.

* Patreon: https://patreon.com/TheBlokeAI
* Ko-Fi: https://ko-fi.com/TheBlokeAI

**Special thanks to**: Luke from CarbonQuill, Aemon Algiz.

**Patreon special mentions**: RoA, Lone Striker, Gabriel Puliatti, Derek Yates, Randy H, Jonathan Leane, Eugene Pentland, Karl Bernard, Viktor Bowallius, senxiiz, Daniel P. Andersen, Pierre Kircher, Deep Realms, Cory Kujawski, Oscar Rangel, Fen Risland, Ajan Kanaga, LangChain4j, webtim, Nikolai Manek, Trenton Dambrowitz, Raven Klaugh, Kalila, Khalefa Al-Ahmad, Chris McCloskey, Luke @flexchar, Ai Maven, Dave, Asp the Wyvern, Sean Connelly, Imad Khwaja, Space Cruiser, Rainer Wilmers, subjectnull, Alps Aficionado, Willian Hasse, Fred von Graf, Artur Olbinski, Johann-Peter Hartmann, WelcomeToTheClub, Willem Michiel, Michael Levine, Iucharbius , Spiking Neurons AB, K, biorpg, John Villwock, Pyrater, Greatston Gnanesh, Mano Prime, Junyu Yang, Stephen Murray, John Detwiler, Luke Pendergrass, terasurfer , Pieter, zynix , Edmond Seymore, theTransient, Nathan LeClaire, vamX, Kevin Schuppel, Preetika Verma, ya boyyy, Alex , SuperWojo, Ghost , Joseph William Delisle, Matthew Berman, Talal Aujan, chris gileta, Illia Dulskyi.

Thank you to all my generous patrons and donaters!

<!-- footer end -->

# Original model card: Pankaj Mathur's Orca Mini v2 13B

# orca_mini_v2_13b

An **Uncensored** LLaMA-13b model in collaboration with [Eric Hartford](https://huggingface.co/ehartford). trained on explain tuned datasets, created using Instructions and Input from WizardLM, Alpaca & Dolly-V2 datasets and applying Orca Research Paper dataset construction approaches.

Please note this model has *better code generation capabilities* compare to our original orca_mini_13b which was trained on base OpenLLaMA-13b model and which has the [empty spaces issues & found not good for code generation]((https://github.com/openlm-research/open_llama#update-06072023)).


**P.S. I am #opentowork, if you can help, please reach out to me at www.linkedin.com/in/pankajam**

# Evaluation

I evaluated orca_mini_v2_13b on a wide range of tasks using [Language Model Evaluation Harness](https://github.com/EleutherAI/lm-evaluation-harness) from EleutherAI. 

Here are the results on metrics used by [HuggingFaceH4 Open LLM Leaderboard](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)


||||
|:------:|:-------------:|:---------:|
|**Task**|**Value**|**Stderr**|
|*arc_challenge*|0.5572|0.0145|
|*hellaswag*|0.7964|0.0040|
|*mmlu*|0.4969|0.035|
|*truthfulqa_mc*|0.5231|0.0158|
|*Total Average*|0.5933|0.0114|




# Dataset

We used uncensored script on top of the previous explain tuned datasets we build which are [WizardLM dataset ~70K](https://github.com/nlpxucan/WizardLM), [Alpaca dataset ~52K](https://crfm.stanford.edu/2023/03/13/alpaca.html)  & [Dolly-V2 dataset ~15K](https://github.com/databrickslabs/dolly) created using approaches from [Orca Research Paper](https://arxiv.org/abs/2306.02707).

We leverage all of the 15 system instructions provided in Orca Research Paper. to generate custom datasets, in contrast to vanilla instruction tuning approaches used by original datasets.

This helps student model aka this model to learn ***thought*** process from teacher model, which is ChatGPT (gpt-3.5-turbo-0301 version).

Please see below example usage how the **System** prompt is added before each **instruction**.

# Training

The training configurations are provided in the table below.

The training takes on 4x A100(80G) GPUs and lasts for around 21 Hours for cost of $210 (~$10 for Spot Instance) by using [Azure Standard_NC96ads_A100_v4](https://learn.microsoft.com/en-us/azure/virtual-machines/nc-a100-v4-series#supported-features).

We used DeepSpeed with fully sharded data parallelism, also know as [ZeRO stage 3](https://engineering.fb.com/2021/07/15/open-source/fsdp/) by writing our own fine tunning scripts plus leveraging some of the model training code provided by amazing [FastChat](https://github.com/lm-sys/FastChat)

Here are some of params used during training:

|||
|:-------------:|:-------------:|
|*batch_size*|48|
|*train_micro_batch_size_per_gpu*|3|
|*gradient_accumulation_steps*|4|
|*Learning rate*|2e-5|
|*Max length*|2048|
|*Epochs*|3|
|*Optimizer*|AdamW|



# Example Usage

Here is prompt format for [Oobabooga Text generation UI ](https://github.com/oobabooga/text-generation-webui)

```
### System:
{system}

### User:
{instruction}

### Input:
{input}

### Response:

```

Here is sample example:

```
### System:
You are an AI assistant that follows instruction extremely well. Help as much as you can.

### User:
Tell me how to break into my own car

### Input:

### Response:
Breaking into your own car requires certain skills and tools. Here are the basic steps:

1. Find a ^^^^^^^^^^^^^
2. Unlock the car by using the ^^^^^^^^^^^^^.
3. Use a ^^^^^^^^^^^^^.
4. Once the ^^^^^^^^^^^^^.
5. If the ^^^^^^^^^^^^^.

```

Below shows a code example on how to use this model

```python
import torch
from transformers import LlamaForCausalLM, LlamaTokenizer

# Hugging Face model_path
model_path = 'psmathur/orca_mini_v2_13b'
tokenizer = LlamaTokenizer.from_pretrained(model_path)
model = LlamaForCausalLM.from_pretrained(
    model_path, torch_dtype=torch.float16, device_map='auto',
)


#generate text function
def generate_text(system, instruction, input=None):
    
    if input:
        prompt = f"### System:\n{system}\n\n### User:\n{instruction}\n\n### Input:\n{input}\n\n### Response:\n"
    else:
        prompt = f"### System:\n{system}\n\n### User:\n{instruction}\n\n### Response:\n"
    
    tokens = tokenizer.encode(prompt)
    tokens = torch.LongTensor(tokens).unsqueeze(0)
    tokens = tokens.to('cuda')

    instance = {'input_ids': tokens,'top_p': 1.0, 'temperature':0.7, 'generate_len': 1024, 'top_k': 50}

    length = len(tokens[0])
    with torch.no_grad():
        rest = model.generate(
            input_ids=tokens, 
            max_length=length+instance['generate_len'], 
            use_cache=True, 
            do_sample=True, 
            top_p=instance['top_p'],
            temperature=instance['temperature'],
            top_k=instance['top_k']
        )    
    output = rest[0][length:]
    string = tokenizer.decode(output, skip_special_tokens=True)
    return f'[!] Response: {string}'

# Sample Test Instruction
system = 'You are an AI assistant that follows instruction extremely well. Help as much as you can.'
instruction = 'Tell me how to break into my own car'
print(generate_text(system, instruction))

```

**NOTE: The real response is hidden here with ^^^^^^^^^^^^^.**

```
[!] Response:
Breaking into your own car requires certain skills and tools. Here are the basic steps:

1. Find a ^^^^^^^^^^^^^
2. Unlock the car by using the ^^^^^^^^^^^^^.
3. Use a ^^^^^^^^^^^^^.
4. Once the ^^^^^^^^^^^^^.
5. If the ^^^^^^^^^^^^^.

```

Next Goals:
1) Try more data like actually using FLAN-v2, just like Orka Research Paper (I am open for suggestions)
2) Provide more options for Text generation UI. (may be https://github.com/oobabooga/text-generation-webui)
3) Provide 4bit GGML/GPTQ quantized model (may be [TheBloke](https://huggingface.co/TheBloke) can help here)


Limitations & Biases:

This model can produce factually incorrect output, and should not be relied on to produce factually accurate information.
This model was trained on various public datasets. While great efforts have been taken to clean the pretraining data, it is possible that this model could generate lewd, biased or otherwise offensive outputs.

Disclaimer:

The license on this model does not constitute legal advice. We are not responsible for the actions of third parties who use this model.
Please cosult an attorney before using this model for commercial purposes.


Citiation:

If you found wizardlm_alpaca_dolly_orca_open_llama_7b useful in your research or applications, please kindly cite using the following BibTeX:

```
@misc{orca_mini_v2_13b,
  author = {Pankaj Mathur},
  title = {orca_mini_v2_13b: An explain tuned LLaMA-13b model on uncensored wizardlm, alpaca, & dolly datasets},
  year = {2023},
  publisher = {GitHub, HuggingFace},
  journal = {GitHub repository, HuggingFace repository},
  howpublished = {\url{https://https://huggingface.co/psmathur/orca_mini_v2_13b},
}
```
```
@software{touvron2023llama,
  title={LLaMA: Open and Efficient Foundation Language Models},
  author={Touvron, Hugo and Lavril, Thibaut and Izacard, Gautier and Martinet, Xavier and Lachaux, Marie-Anne and Lacroix, Timoth{\'e}e and Rozi{\`e}re, Baptiste and Goyal, Naman and Hambro, Eric and Azhar, Faisal and Rodriguez, Aurelien and Joulin, Armand and Grave, Edouard and Lample, Guillaume},
  journal={arXiv preprint arXiv:2302.13971},
  year={2023}
}
```
```
@misc{openalpaca,
  author = {Yixuan Su and Tian Lan and Deng Cai},
  title = {OpenAlpaca: A Fully Open-Source Instruction-Following Model Based On OpenLLaMA},
  year = {2023},
  publisher = {GitHub},
  journal = {GitHub repository},
  howpublished = {\url{https://github.com/yxuansu/OpenAlpaca}},
}
```
```
@misc{alpaca,
  author = {Rohan Taori and Ishaan Gulrajani and Tianyi Zhang and Yann Dubois and Xuechen Li and Carlos Guestrin and Percy Liang and Tatsunori B. Hashimoto },
  title = {Stanford Alpaca: An Instruction-following LLaMA model},
  year = {2023},
  publisher = {GitHub},
  journal = {GitHub repository},
  howpublished = {\url{https://github.com/tatsu-lab/stanford_alpaca}},
}
```
```
@online{DatabricksBlog2023DollyV2,
    author    = {Mike Conover and Matt Hayes and Ankit Mathur and Jianwei Xie and Jun Wan and Sam Shah and Ali Ghodsi and Patrick Wendell and Matei Zaharia and Reynold Xin},
    title     = {Free Dolly: Introducing the World's First Truly Open Instruction-Tuned LLM},
    year      = {2023},
    url       = {https://www.databricks.com/blog/2023/04/12/dolly-first-open-commercially-viable-instruction-tuned-llm},
    urldate   = {2023-06-30}
}
```
```
@misc{xu2023wizardlm,
      title={WizardLM: Empowering Large Language Models to Follow Complex Instructions}, 
      author={Can Xu and Qingfeng Sun and Kai Zheng and Xiubo Geng and Pu Zhao and Jiazhan Feng and Chongyang Tao and Daxin Jiang},
      year={2023},
      eprint={2304.12244},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}
```