Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,64 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
language:
|
| 4 |
+
- en
|
| 5 |
+
- fr
|
| 6 |
+
- de
|
| 7 |
+
- es
|
| 8 |
+
- it
|
| 9 |
+
- pt
|
| 10 |
+
- ru
|
| 11 |
+
- zh
|
| 12 |
+
- ja
|
| 13 |
+
pipeline_tag: text-generation
|
| 14 |
+
tags:
|
| 15 |
+
- chat
|
| 16 |
+
---
|
| 17 |
+
|
| 18 |
+
## This repo contains EXL2 quants of the model. If you need the original weights, please find them [here](https://huggingface.co/anthracite-org/magnum-12b-v2.5-kto).
|
| 19 |
+
## Base repo only contains the measurement file, see revisions for your quant of choice.
|
| 20 |
+
|
| 21 |
+
- [measurement.json](https://huggingface.co/anthracite-org/magnum-12b-v2.5-kto-exl2/tree/main)
|
| 22 |
+
- [3.0bpw](https://huggingface.co/anthracite-org/magnum-12b-v2.5-kto-exl2/tree/4.0bpw)
|
| 23 |
+
- [4.0bpw](https://huggingface.co/anthracite-org/magnum-12b-v2.5-kto-exl2/tree/4.0bpw)
|
| 24 |
+
- [5.0bpw](https://huggingface.co/anthracite-org/magnum-12b-v2.5-kto-exl2/tree/5.0bpw)
|
| 25 |
+
- [6.0bpw](https://huggingface.co/anthracite-org/magnum-12b-v2.5-kto-exl2/tree/6.0bpw)
|
| 26 |
+
- [8.0bpw](https://huggingface.co/anthracite-org/magnum-12b-v2.5-kto-exl2/tree/8.0bpw)
|
| 27 |
+
|
| 28 |
+
|
| 29 |
+
|
| 30 |
+
v2.5 KTO is an experimental release; we are testing a hybrid reinforcement learning strategy of KTO + DPOP, using rejected data sampled from the original model as "rejected". For "chosen", we use data from the original finetuning dataset as "chosen".
|
| 31 |
+
This was done on a limited portion of of primarily instruction following data; we plan to scale up a larger KTO dataset in the future for better generalization.
|
| 32 |
+
|
| 33 |
+
This is the 5th in a series of models designed to replicate the prose quality of the Claude 3 models, specifically Sonnet and Opus. This model is fine-tuned on top of [anthracite-org/magnum-12b-v2](https://huggingface.co/anthracite-org/magnum-12b-v2).
|
| 34 |
+
|
| 35 |
+
## Prompting
|
| 36 |
+
Model has been Instruct tuned with the ChatML formatting. A typical input would look like this:
|
| 37 |
+
|
| 38 |
+
```py
|
| 39 |
+
"""<|im_start|>user
|
| 40 |
+
Hi there!<|im_end|>
|
| 41 |
+
<|im_start|>assistant
|
| 42 |
+
Nice to meet you!<|im_end|>
|
| 43 |
+
<|im_start|>user
|
| 44 |
+
Can I ask a question?<|im_end|>
|
| 45 |
+
<|im_start|>assistant
|
| 46 |
+
"""
|
| 47 |
+
```
|
| 48 |
+
|
| 49 |
+
## Credits
|
| 50 |
+
- Stheno dataset (filtered)
|
| 51 |
+
- [kalomaze/Opus_Instruct_25k](https://huggingface.co/datasets/kalomaze/Opus_Instruct_25k)
|
| 52 |
+
- [Nopm/Opus_WritingStruct](https://huggingface.co/datasets/Nopm/Opus_WritingStruct)
|
| 53 |
+
- [Gryphe/Sonnet3.5-SlimOrcaDedupCleaned](https://huggingface.co/datasets/Gryphe/Sonnet3.5-SlimOrcaDedupCleaned) (A ~16k rows subset)
|
| 54 |
+
- [kalomaze/Opus_Instruct_3k](https://huggingface.co/datasets/kalomaze/Opus_Instruct_3k)
|
| 55 |
+
|
| 56 |
+
This model has been a team effort, and the credits goes to all members of Anthracite.
|
| 57 |
+
|
| 58 |
+
## Training
|
| 59 |
+
The training was done for 2 epochs. We used 8x [NVIDIA H100 Tensor Core](https://www.nvidia.com/en-us/data-center/h100/) GPUs for the full-parameter fine-tuning of the model.
|
| 60 |
+
|
| 61 |
+
[<img src="https://raw.githubusercontent.com/OpenAccess-AI-Collective/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>](https://github.com/OpenAccess-AI-Collective/axolotl)
|
| 62 |
+
|
| 63 |
+
## Safety
|
| 64 |
+
...
|