Update README.md
Browse files
README.md
CHANGED
@@ -16,13 +16,62 @@ should probably proofread and complete it, then remove this comment. -->
|
|
16 |
|
17 |
# 06051615
|
18 |
|
19 |
-
This model is a fine-tuned version of [/
|
20 |
It achieves the following results on the evaluation set:
|
21 |
- Loss: 0.9018
|
22 |
|
23 |
## Model description
|
24 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
25 |
More information needed
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
26 |
|
27 |
## Intended uses & limitations
|
28 |
|
|
|
16 |
|
17 |
# 06051615
|
18 |
|
19 |
+
This model is a fine-tuned version of [Qwen/Qwen1.5-7B-Chat](https://huggingface.co/Qwen/Qwen1.5-7B-Chat) on the my own dataset.
|
20 |
It achieves the following results on the evaluation set:
|
21 |
- Loss: 0.9018
|
22 |
|
23 |
## Model description
|
24 |
|
25 |
+
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:
|
26 |
+
* 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;
|
27 |
+
* Significant performance improvement in Chat models;
|
28 |
+
* Multilingual support of both base and chat models;
|
29 |
+
* Stable support of 32K context length for models of all sizes
|
30 |
+
* No need of `trust_remote_code`.
|
31 |
+
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).
|
32 |
+
|
33 |
+
## Intended uses & limitations
|
34 |
+
More information needed
|
35 |
+
## Training and evaluation data
|
36 |
More information needed
|
37 |
+
## Training procedure
|
38 |
+
### Training hyperparameters
|
39 |
+
The following hyperparameters were used during training:
|
40 |
+
- learning_rate: 7e-05
|
41 |
+
- train_batch_size: 2
|
42 |
+
- eval_batch_size: 1
|
43 |
+
- seed: 42
|
44 |
+
- distributed_type: multi-GPU
|
45 |
+
- num_devices: 2
|
46 |
+
- gradient_accumulation_steps: 4
|
47 |
+
- total_train_batch_size: 16
|
48 |
+
- total_eval_batch_size: 2
|
49 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
50 |
+
- lr_scheduler_type: cosine
|
51 |
+
- lr_scheduler_warmup_steps: 13
|
52 |
+
- num_epochs: 5.0
|
53 |
+
- mixed_precision_training: Native AMP
|
54 |
+
|
55 |
+
### Training results
|
56 |
+
|
57 |
+
| Training Loss | Epoch | Step | Validation Loss |
|
58 |
+
| :-----------: | :----: | :--: | :-------------: |
|
59 |
+
| 0.6358 | 0.7619 | 20 | 0.5865 |
|
60 |
+
| 0.6379 | 1.5238 | 40 | 0.5621 |
|
61 |
+
| 0.6067 | 2.2857 | 60 | 0.5561 |
|
62 |
+
| 0.5339 | 3.0476 | 80 | 0.5515 |
|
63 |
+
| 0.6749 | 3.8095 | 100 | 0.5500 |
|
64 |
+
| 0.6351 | 4.5714 | 120 | 0.5497 |
|
65 |
+
|
66 |
+
|
67 |
+
### Framework versions
|
68 |
+
|
69 |
+
- PEFT 0.10.0
|
70 |
+
- Transformers 4.40.0
|
71 |
+
- Pytorch 2.1.0+cu121
|
72 |
+
- Datasets 2.14.5
|
73 |
+
- Tokenizers 0.19.1
|
74 |
+
|
75 |
|
76 |
## Intended uses & limitations
|
77 |
|