Commit
•
f892a31
1
Parent(s):
b41c2b1
uploaded readme
Browse files
README.md
ADDED
|
@@ -0,0 +1,89 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
Quantization made by Richard Erkhov.
|
| 2 |
+
|
| 3 |
+
[Github](https://github.com/RichardErkhov)
|
| 4 |
+
|
| 5 |
+
[Discord](https://discord.gg/pvy7H8DZMG)
|
| 6 |
+
|
| 7 |
+
[Request more models](https://github.com/RichardErkhov/quant_request)
|
| 8 |
+
|
| 9 |
+
|
| 10 |
+
MoMo-72B-lora-1.8.6-DPO - GGUF
|
| 11 |
+
- Model creator: https://huggingface.co/moreh/
|
| 12 |
+
- Original model: https://huggingface.co/moreh/MoMo-72B-lora-1.8.6-DPO/
|
| 13 |
+
|
| 14 |
+
|
| 15 |
+
| Name | Quant method | Size |
|
| 16 |
+
| ---- | ---- | ---- |
|
| 17 |
+
| [MoMo-72B-lora-1.8.6-DPO.Q2_K.gguf](https://huggingface.co/RichardErkhov/moreh_-_MoMo-72B-lora-1.8.6-DPO-gguf/blob/main/MoMo-72B-lora-1.8.6-DPO.Q2_K.gguf) | Q2_K | 25.22GB |
|
| 18 |
+
| [MoMo-72B-lora-1.8.6-DPO.IQ3_XS.gguf](https://huggingface.co/RichardErkhov/moreh_-_MoMo-72B-lora-1.8.6-DPO-gguf/blob/main/MoMo-72B-lora-1.8.6-DPO.IQ3_XS.gguf) | IQ3_XS | 27.88GB |
|
| 19 |
+
| [MoMo-72B-lora-1.8.6-DPO.IQ3_S.gguf](https://huggingface.co/RichardErkhov/moreh_-_MoMo-72B-lora-1.8.6-DPO-gguf/blob/main/MoMo-72B-lora-1.8.6-DPO.IQ3_S.gguf) | IQ3_S | 29.4GB |
|
| 20 |
+
| [MoMo-72B-lora-1.8.6-DPO.Q3_K_S.gguf](https://huggingface.co/RichardErkhov/moreh_-_MoMo-72B-lora-1.8.6-DPO-gguf/blob/main/MoMo-72B-lora-1.8.6-DPO.Q3_K_S.gguf) | Q3_K_S | 29.4GB |
|
| 21 |
+
| [MoMo-72B-lora-1.8.6-DPO.IQ3_M.gguf](https://huggingface.co/RichardErkhov/moreh_-_MoMo-72B-lora-1.8.6-DPO-gguf/blob/main/MoMo-72B-lora-1.8.6-DPO.IQ3_M.gguf) | IQ3_M | 30.98GB |
|
| 22 |
+
| [MoMo-72B-lora-1.8.6-DPO.Q3_K.gguf](https://huggingface.co/RichardErkhov/moreh_-_MoMo-72B-lora-1.8.6-DPO-gguf/blob/main/MoMo-72B-lora-1.8.6-DPO.Q3_K.gguf) | Q3_K | 32.85GB |
|
| 23 |
+
| [MoMo-72B-lora-1.8.6-DPO.Q3_K_M.gguf](https://huggingface.co/RichardErkhov/moreh_-_MoMo-72B-lora-1.8.6-DPO-gguf/blob/main/MoMo-72B-lora-1.8.6-DPO.Q3_K_M.gguf) | Q3_K_M | 32.85GB |
|
| 24 |
+
| [MoMo-72B-lora-1.8.6-DPO.Q3_K_L.gguf](https://huggingface.co/RichardErkhov/moreh_-_MoMo-72B-lora-1.8.6-DPO-gguf/blob/main/MoMo-72B-lora-1.8.6-DPO.Q3_K_L.gguf) | Q3_K_L | 35.85GB |
|
| 25 |
+
| [MoMo-72B-lora-1.8.6-DPO.IQ4_XS.gguf](https://huggingface.co/RichardErkhov/moreh_-_MoMo-72B-lora-1.8.6-DPO-gguf/blob/main/MoMo-72B-lora-1.8.6-DPO.IQ4_XS.gguf) | IQ4_XS | 36.41GB |
|
| 26 |
+
| [MoMo-72B-lora-1.8.6-DPO.Q4_0.gguf](https://huggingface.co/RichardErkhov/moreh_-_MoMo-72B-lora-1.8.6-DPO-gguf/tree/main/) | Q4_0 | 38.19GB |
|
| 27 |
+
| [MoMo-72B-lora-1.8.6-DPO.IQ4_NL.gguf](https://huggingface.co/RichardErkhov/moreh_-_MoMo-72B-lora-1.8.6-DPO-gguf/tree/main/) | IQ4_NL | 38.42GB |
|
| 28 |
+
| [MoMo-72B-lora-1.8.6-DPO.Q4_K_S.gguf](https://huggingface.co/RichardErkhov/moreh_-_MoMo-72B-lora-1.8.6-DPO-gguf/tree/main/) | Q4_K_S | 38.45GB |
|
| 29 |
+
| [MoMo-72B-lora-1.8.6-DPO.Q4_K.gguf](https://huggingface.co/RichardErkhov/moreh_-_MoMo-72B-lora-1.8.6-DPO-gguf/tree/main/) | Q4_K | 40.77GB |
|
| 30 |
+
| [MoMo-72B-lora-1.8.6-DPO.Q4_K_M.gguf](https://huggingface.co/RichardErkhov/moreh_-_MoMo-72B-lora-1.8.6-DPO-gguf/tree/main/) | Q4_K_M | 40.77GB |
|
| 31 |
+
| [MoMo-72B-lora-1.8.6-DPO.Q4_1.gguf](https://huggingface.co/RichardErkhov/moreh_-_MoMo-72B-lora-1.8.6-DPO-gguf/tree/main/) | Q4_1 | 42.32GB |
|
| 32 |
+
| [MoMo-72B-lora-1.8.6-DPO.Q5_0.gguf](https://huggingface.co/RichardErkhov/moreh_-_MoMo-72B-lora-1.8.6-DPO-gguf/tree/main/) | Q5_0 | 46.46GB |
|
| 33 |
+
| [MoMo-72B-lora-1.8.6-DPO.Q5_K_S.gguf](https://huggingface.co/RichardErkhov/moreh_-_MoMo-72B-lora-1.8.6-DPO-gguf/tree/main/) | Q5_K_S | 46.46GB |
|
| 34 |
+
| [MoMo-72B-lora-1.8.6-DPO.Q5_K.gguf](https://huggingface.co/RichardErkhov/moreh_-_MoMo-72B-lora-1.8.6-DPO-gguf/tree/main/) | Q5_K | 47.79GB |
|
| 35 |
+
| [MoMo-72B-lora-1.8.6-DPO.Q5_K_M.gguf](https://huggingface.co/RichardErkhov/moreh_-_MoMo-72B-lora-1.8.6-DPO-gguf/tree/main/) | Q5_K_M | 47.79GB |
|
| 36 |
+
| [MoMo-72B-lora-1.8.6-DPO.Q5_1.gguf](https://huggingface.co/RichardErkhov/moreh_-_MoMo-72B-lora-1.8.6-DPO-gguf/tree/main/) | Q5_1 | 50.59GB |
|
| 37 |
+
| [MoMo-72B-lora-1.8.6-DPO.Q6_K.gguf](https://huggingface.co/RichardErkhov/moreh_-_MoMo-72B-lora-1.8.6-DPO-gguf/tree/main/) | Q6_K | 55.24GB |
|
| 38 |
+
| [MoMo-72B-lora-1.8.6-DPO.Q8_0.gguf](https://huggingface.co/RichardErkhov/moreh_-_MoMo-72B-lora-1.8.6-DPO-gguf/tree/main/) | Q8_0 | 71.55GB |
|
| 39 |
+
|
| 40 |
+
|
| 41 |
+
|
| 42 |
+
|
| 43 |
+
Original model description:
|
| 44 |
+
---
|
| 45 |
+
license: mit
|
| 46 |
+
language:
|
| 47 |
+
- en
|
| 48 |
+
---
|
| 49 |
+
# **Introduction**
|
| 50 |
+
MoMo-72B-lora-1.8.6-DPO is trained via Direct Preference Optimization([DPO](https://arxiv.org/abs/2305.18290)) from [MoMo-72B-LoRA-V1.4](https://huggingface.co/moreh/MoMo-72B-LoRA-V1.4) as its base model, with several optimizations in hyperparameters.
|
| 51 |
+
[MoMo-72B-LoRA-V1.4](https://huggingface.co/moreh/MoMo-72B-LoRA-V1.4) is trained via Supervised Fine-Tuning (SFT) using [LoRA](https://arxiv.org/abs/2106.09685), with the QWEN-72B model as its base-model.
|
| 52 |
+
Note that we did not exploit any form of weight merge.
|
| 53 |
+
For leaderboard submission, the trained weight is realigned for compatibility with llama.
|
| 54 |
+
MoMo-72B is trained using **[Moreh](https://moreh.io/)**'s [MoAI platform](https://moreh.io/product), which simplifies the training of large-scale models, and AMD's MI250 GPU.
|
| 55 |
+
|
| 56 |
+
|
| 57 |
+
## Details
|
| 58 |
+
### Used Librarys
|
| 59 |
+
- torch
|
| 60 |
+
- peft
|
| 61 |
+
### Used Datasets
|
| 62 |
+
- [slimorca](Open-Orca/SlimOrca)
|
| 63 |
+
- [truthy](https://huggingface.co/datasets/jondurbin/truthy-dpo-v0.1)
|
| 64 |
+
- [orca_dpo_pairs](https://huggingface.co/datasets/Intel/orca_dpo_pairs)
|
| 65 |
+
- No other dataset was used
|
| 66 |
+
- No benchmark test set or the training set are used
|
| 67 |
+
- [data contamination check](https://github.com/swj0419/detect-pretrain-code-contamination) result
|
| 68 |
+
|
| 69 |
+
| Model | ARC | MMLU | TruthfulQA | GSM8K |
|
| 70 |
+
|------------------------------|-------|-------|-------|-------|
|
| 71 |
+
| **V1.8.6(result < 0.1, %)**| TBU |TBU | 0.73 | TBU |
|
| 72 |
+
### Used Environments
|
| 73 |
+
- AMD MI250 & MoAI platform
|
| 74 |
+
- Please visit https://moreh.io/product for more information about MoAI platform
|
| 75 |
+
- Or, contact us directly [[email protected]](mailto:[email protected])
|
| 76 |
+
|
| 77 |
+
## How to use
|
| 78 |
+
|
| 79 |
+
```python
|
| 80 |
+
# pip install transformers==4.35.2
|
| 81 |
+
import torch
|
| 82 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
| 83 |
+
|
| 84 |
+
tokenizer = AutoTokenizer.from_pretrained("moreh/MoMo-72B-lora-1.8.6-DPO")
|
| 85 |
+
model = AutoModelForCausalLM.from_pretrained(
|
| 86 |
+
"moreh/MoMo-72B-lora-1.8.6-DPO"
|
| 87 |
+
)
|
| 88 |
+
```
|
| 89 |
+
|