Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,81 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
library_name: peft
|
| 3 |
+
license: other
|
| 4 |
+
base_model: Qwen/Qwen2.5-3B-Instruct
|
| 5 |
+
tags:
|
| 6 |
+
- generated_from_trainer
|
| 7 |
+
model-index:
|
| 8 |
+
- name: pancho-v1-qw25-3B-UNAMGS
|
| 9 |
+
results: []
|
| 10 |
+
datasets:
|
| 11 |
+
- Magpie-Align/Magpie-Pro-MT-300K-v0.1
|
| 12 |
+
- Magpie-Align/Magpie-Llama-3.1-Pro-MT-300K-Filtered
|
| 13 |
+
language:
|
| 14 |
+
- en
|
| 15 |
+
---
|
| 16 |
+
|
| 17 |
+
# pancho-v1-qw25-3B-UNAMGS
|
| 18 |
+
|
| 19 |
+
This model is a fine-tuned version of [Qwen/Qwen2.5-3B-Instruct](https://huggingface.co/Qwen/Qwen2.5-3B-Instruct):
|
| 20 |
+
It achieves the following results on the evaluation set:
|
| 21 |
+
- Loss: 0.6555
|
| 22 |
+
|
| 23 |
+
[<img src="https://raw.githubusercontent.com/axolotl-ai-cloud/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>](https://github.com/axolotl-ai-cloud/axolotl)
|
| 24 |
+
|
| 25 |
+
## Model description
|
| 26 |
+
Trained with MagPie:
|
| 27 |
+
- Magpie-Align/Magpie-Llama-3.1-Pro-MT-300K-Filtered
|
| 28 |
+
- Magpie-Align/Magpie-Pro-MT-300K-v0.1
|
| 29 |
+
|
| 30 |
+
UNA on MLPs `4, 10, 16, 22, 28`
|
| 31 |
+
|
| 32 |
+
MGS on 3 Scales.
|
| 33 |
+
|
| 34 |
+
Following https://arxiv.org/abs//2410.21228 facts.
|
| 35 |
+
|
| 36 |
+
|
| 37 |
+
### Training hyperparameters
|
| 38 |
+
|
| 39 |
+
The following hyperparameters were used during training:
|
| 40 |
+
- learning_rate: 2e-05
|
| 41 |
+
- seed: 42
|
| 42 |
+
- distributed_type: multi-GPU
|
| 43 |
+
- num_devices: 8
|
| 44 |
+
- total_train_batch_size: 256
|
| 45 |
+
- total_eval_batch_size: 16
|
| 46 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 47 |
+
- num_epochs: 1
|
| 48 |
+
|
| 49 |
+
### Training results
|
| 50 |
+
|
| 51 |
+
| Training Loss | Epoch | Step | Validation Loss |
|
| 52 |
+
|:-------------:|:------:|:----:|:---------------:|
|
| 53 |
+
| 1.2127 | 0.0015 | 1 | 0.8711 |
|
| 54 |
+
| 0.9905 | 0.0509 | 35 | 0.7338 |
|
| 55 |
+
| 0.9685 | 0.1019 | 70 | 0.7114 |
|
| 56 |
+
| 0.9554 | 0.1528 | 105 | 0.6994 |
|
| 57 |
+
| 0.9077 | 0.2037 | 140 | 0.6915 |
|
| 58 |
+
| 0.9149 | 0.2547 | 175 | 0.6859 |
|
| 59 |
+
| 0.9363 | 0.3056 | 210 | 0.6795 |
|
| 60 |
+
| 0.8975 | 0.3566 | 245 | 0.6745 |
|
| 61 |
+
| 0.9095 | 0.4075 | 280 | 0.6709 |
|
| 62 |
+
| 0.9216 | 0.4584 | 315 | 0.6681 |
|
| 63 |
+
| 0.9143 | 0.5094 | 350 | 0.6666 |
|
| 64 |
+
| 0.8879 | 0.5603 | 385 | 0.6645 |
|
| 65 |
+
| 0.9194 | 0.6112 | 420 | 0.6625 |
|
| 66 |
+
| 0.9123 | 0.6622 | 455 | 0.6615 |
|
| 67 |
+
| 0.9056 | 0.7131 | 490 | 0.6591 |
|
| 68 |
+
| 0.9172 | 0.7641 | 525 | 0.6578 |
|
| 69 |
+
| 0.886 | 0.8150 | 560 | 0.6566 |
|
| 70 |
+
| 0.9155 | 0.8659 | 595 | 0.6568 |
|
| 71 |
+
| 0.9029 | 0.9169 | 630 | 0.6560 |
|
| 72 |
+
| 0.8942 | 0.9678 | 665 | 0.6555 |
|
| 73 |
+
|
| 74 |
+
|
| 75 |
+
### Framework versions
|
| 76 |
+
|
| 77 |
+
- PEFT 0.13.2
|
| 78 |
+
- Transformers 4.45.2
|
| 79 |
+
- Pytorch 2.3.0+cu121
|
| 80 |
+
- Datasets 3.0.1
|
| 81 |
+
- Tokenizers 0.20.1#
|