Update README.md
Browse files
README.md
CHANGED
@@ -1,9 +1,55 @@
|
|
1 |
---
|
2 |
library_name: peft
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
---
|
4 |
-
## Training procedure
|
5 |
|
6 |
-
###
|
7 |
|
|
|
|
|
8 |
|
9 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
library_name: peft
|
3 |
+
tags:
|
4 |
+
- meta-llama
|
5 |
+
- code
|
6 |
+
- instruct
|
7 |
+
- WizardLM
|
8 |
+
- Mistral-7B-v0.1
|
9 |
+
datasets:
|
10 |
+
- WizardLM/WizardLM_evol_instruct_70k
|
11 |
+
base_model: mistralai/Mistral-7B-v0.1
|
12 |
+
license: apache-2.0
|
13 |
---
|
|
|
14 |
|
15 |
+
### Finetuning Overview:
|
16 |
|
17 |
+
**Model Used:** mistralai/Mistral-7B-v0.1
|
18 |
+
**Dataset:** WizardLM/WizardLM_evol_instruct_70k
|
19 |
|
20 |
+
#### Dataset Insights:
|
21 |
+
|
22 |
+
The WizardLM/WizardLM_evol_instruct_70k dataset, tailored specifically for enhancing interactive capabilities, provides valuable instruction-based content. (Note: Additional insights about the dataset, its origin, content, and contributors can be provided here if required).
|
23 |
+
|
24 |
+
#### Finetuning Details:
|
25 |
+
|
26 |
+
With the utilization of [MonsterAPI](https://monsterapi.ai)'s [LLM finetuner](https://docs.monsterapi.ai/fine-tune-a-large-language-model-llm), this finetuning:
|
27 |
+
|
28 |
+
- Was achieved with great cost-effectiveness.
|
29 |
+
- Completed in a total duration of 5hrs 18mins for 1 epoch using an A6000 48GB GPU.
|
30 |
+
- Costed `$10` for the entire epoch.
|
31 |
+
|
32 |
+
#### Hyperparameters & Additional Details:
|
33 |
+
|
34 |
+
- **Epochs:** 1
|
35 |
+
- **Cost Per Epoch:** $10
|
36 |
+
- **Total Finetuning Cost:** $10
|
37 |
+
- **Model Path:** mistralai/Mistral-7B-v0.1
|
38 |
+
- **Learning Rate:** 0.0002
|
39 |
+
- **Data Split:** 90% train 10% validation
|
40 |
+
- **Gradient Accumulation Steps:** 4
|
41 |
+
|
42 |
+
---
|
43 |
+
```
|
44 |
+
### INSTRUCTION:
|
45 |
+
[instruction]
|
46 |
+
|
47 |
+
### RESPONSE:
|
48 |
+
[output]
|
49 |
+
```
|
50 |
+
Training loss :
|
51 |
+
![training loss](train-loss.png "Training loss")
|
52 |
+
|
53 |
+
---
|
54 |
+
|
55 |
+
license: apache-2.0
|