README.md

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

## Dromedary-65B-LoRA GGML

These files are the result of merging the [delta weights of IBM's Dromedary 65B LoRA](https://huggingface.co./zhiqings/dromedary-65b-lora-delta-v0) with the original Llama 65B model.

This repo contains GGML files for for CPU inference using [llama.cpp](https://github.com/ggerganov/llama.cpp).

## Repositories available

* [4bit GPTQ models for GPU inference](https://huggingface.co./TheBloke/dromedary-65B-lora-GPTQ)
* [4bit and 5bit GGML models for CPU inference in llama.cpp](https://huggingface.co./TheBloke/dromedary-65B-lora-GGML)
* [float16 unquantised model for GPU](https://huggingface.co./TheBloke/dromedary-65B-lora-HF)

## THE FILES IN MAIN BRANCH REQUIRES LATEST LLAMA.CPP (May 19th 2023 - commit 2d5db48)!

llama.cpp recently made another breaking change to its quantisation methods - https://github.com/ggerganov/llama.cpp/pull/1508

I have quantised the GGML files in this repo with the latest version. Therefore you will require llama.cpp compiled on May 19th or later (commit `2d5db48` or later) to use them.

For files compatible with the previous version of llama.cpp, please see branch `previous_llama_ggmlv2`.

## Provided files
| Name | Quant method | Bits | Size | RAM required | Use case |
| ---- | ---- | ---- | ---- | ---- | ----- |
`dromedary-lora-65B.ggmlv3.q4_0.bin` | q4_0 | 4bit | 40.8GB | 43GB | 4-bit. |
`dromedary-lora-65B.ggmlv3.q4_1.bin` | q4_1 | 4bit | 44.9GB | 47GB | 4-bit. Higher accuracy than q4_0 but not as high as q5_0. However has quicker inference than q5 models. |
`dromedary-lora-65B.ggmlv3.q5_0.bin` | q5_0 | 5bit | 44.9GB | 47GB | 5-bit. Higher accuracy, higher resource usage and slower inference.  |
`dromedary-lora-65B.ggmlv3.q5_1.bin` | q5_1 | 5bit | 49GB | 51GB | 5-bit. Even higher accuracy, higher resource usage and slower inference. |


# Original Dromedary Model Card
 
See https://github.com/IBM/Dromedary#model-weights for instructions.

## Model details

<div align="center">

<img src="https://raw.githubusercontent.com/IBM/Dromedary/main/assets/images/dromedary_logo.svg" alt="Dromedary Logo"/>

</div>

**Model type:**
Dromedary is an open-source self-aligned language model trained with minimal human supervision.
The base language model is LLaMA-65b, based on the transformer architecture.

**Model date:**
Dromedary was trained between April 2023 and May 2023, but its knowledge only goes up until Sept-2021.

**Organizations developing the model:**
The Dromedary team as a joint effort between CMU and IBM.

**Paper or resources for more information:**
https://mitibmdemos.draco.res.ibm.com/dromedary

**License:**
LLaMA's Non-commercial bespoke license

**Where to send questions or comments about the model:**
https://github.com/IBM/Dromedary/issues

## Intended use
**Primary intended uses:**
The primary use of Dromedary is research on the alignment of large language models.

**Primary intended users:**
The primary intended users of the model are researchers in artificial intelligence.

## Delta weights
We use the following configuration for the LoRA weights:
```
--lora_target_modules='[q_proj,k_proj,v_proj,o_proj]' \
--lora_r=16 \
```

## Training dataset
Fewer than 300 lines of human annotations (including < 200 seed prompts, 16 generic principles, and 5 exemplars for in-context learning),

## Evaluation dataset
We evaluate Dromedary on TruthfulQA and HHH Eval, as well as Vicuna benchmark questions.