Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,88 @@
|
|
1 |
---
|
2 |
license: mit
|
3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
license: mit
|
3 |
---
|
4 |
+
# 0425
|
5 |
+
|
6 |
+
This model is a fine-tuned version of [Qwen/Qwen1.5-7B](https://huggingface.co/Qwen/Qwen1.5-7B) on the alpaca_formatted_ift_eft_Justification dataset.
|
7 |
+
It achieves the following results on the evaluation set:
|
8 |
+
|
9 |
+
- Loss: 0.8213
|
10 |
+
|
11 |
+
## Model description
|
12 |
+
|
13 |
+
Qwen1.5 is the beta version of Qwen2, a transformer-based decoder-only language model pretrained on a large amount of data. In comparison with the previous released Qwen, the improvements include:
|
14 |
+
|
15 |
+
* 8 model sizes, including 0.5B, 1.8B, 4B, 7B, 14B, 32B and 72B dense models, and an MoE model of 14B with 2.7B activated;
|
16 |
+
* Significant performance improvement in Chat models;
|
17 |
+
* Multilingual support of both base and chat models;
|
18 |
+
* Stable support of 32K context length for models of all sizes
|
19 |
+
* No need of `trust_remote_code`.
|
20 |
+
|
21 |
+
For more details, please refer to the [blog post](https://qwenlm.github.io/blog/qwen1.5/) and [GitHub repo](https://github.com/QwenLM/Qwen1.5).
|
22 |
+
|
23 |
+
## Intended uses & limitations
|
24 |
+
|
25 |
+
More information needed
|
26 |
+
|
27 |
+
## Training and evaluation data
|
28 |
+
|
29 |
+
More information needed
|
30 |
+
|
31 |
+
## Training procedure
|
32 |
+
|
33 |
+
### Training hyperparameters
|
34 |
+
|
35 |
+
The following hyperparameters were used during training:
|
36 |
+
|
37 |
+
- learning_rate: 5e-05
|
38 |
+
- train_batch_size: 2
|
39 |
+
- eval_batch_size: 1
|
40 |
+
- seed: 42
|
41 |
+
- distributed_type: multi-GPU
|
42 |
+
- num_devices: 3
|
43 |
+
- gradient_accumulation_steps: 2
|
44 |
+
- total_train_batch_size: 12
|
45 |
+
- total_eval_batch_size: 3
|
46 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
47 |
+
- lr_scheduler_type: cosine
|
48 |
+
- lr_scheduler_warmup_steps: 20
|
49 |
+
- num_epochs: 5.0
|
50 |
+
- mixed_precision_training: Native AMP
|
51 |
+
|
52 |
+
### Training results
|
53 |
+
|
54 |
+
| Training Loss | Epoch | Step | Validation Loss |
|
55 |
+
| :-----------: | :----: | :--: | :-------------: |
|
56 |
+
| 1.0669 | 0.2018 | 100 | 0.8823 |
|
57 |
+
| 0.9156 | 0.4036 | 200 | 0.8593 |
|
58 |
+
| 0.9509 | 0.6054 | 300 | 0.8491 |
|
59 |
+
| 0.8287 | 0.8073 | 400 | 0.8423 |
|
60 |
+
| 0.8772 | 1.0091 | 500 | 0.8390 |
|
61 |
+
| 0.9101 | 1.2109 | 600 | 0.8385 |
|
62 |
+
| 0.8212 | 1.4127 | 700 | 0.8342 |
|
63 |
+
| 0.8721 | 1.6145 | 800 | 0.8327 |
|
64 |
+
| 1.0033 | 1.8163 | 900 | 0.8319 |
|
65 |
+
| 0.9879 | 2.0182 | 1000 | 0.8276 |
|
66 |
+
| 0.964 | 2.2200 | 1100 | 0.8276 |
|
67 |
+
| 0.8409 | 2.4218 | 1200 | 0.8264 |
|
68 |
+
| 0.8055 | 2.6236 | 1300 | 0.8262 |
|
69 |
+
| 1.0026 | 2.8254 | 1400 | 0.8240 |
|
70 |
+
| 0.881 | 3.0272 | 1500 | 0.8241 |
|
71 |
+
| 1.0058 | 3.2291 | 1600 | 0.8226 |
|
72 |
+
| 0.8747 | 3.4309 | 1700 | 0.8205 |
|
73 |
+
| 0.8686 | 3.6327 | 1800 | 0.8215 |
|
74 |
+
| 0.8838 | 3.8345 | 1900 | 0.8208 |
|
75 |
+
| 0.8246 | 4.0363 | 2000 | 0.8218 |
|
76 |
+
| 0.8727 | 4.2381 | 2100 | 0.8216 |
|
77 |
+
| 0.8737 | 4.4400 | 2200 | 0.8214 |
|
78 |
+
| 0.8955 | 4.6418 | 2300 | 0.8214 |
|
79 |
+
| 0.8909 | 4.8436 | 2400 | 0.8215 |
|
80 |
+
|
81 |
+
|
82 |
+
### Framework versions
|
83 |
+
|
84 |
+
- PEFT 0.10.0
|
85 |
+
- Transformers 4.40.0
|
86 |
+
- Pytorch 2.1.0+cu121
|
87 |
+
- Datasets 2.14.5
|
88 |
+
- Tokenizers 0.19.1
|