Commit
•
661d056
1
Parent(s):
71a6046
Updated README.md with initial information
Browse files
README.md
CHANGED
@@ -1,3 +1,77 @@
|
|
1 |
---
|
2 |
license: apache-2.0
|
3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
license: apache-2.0
|
3 |
---
|
4 |
+
|
5 |
+
# Model Card for FalconAlpaca
|
6 |
+
|
7 |
+
<!-- Provide a quick summary of what the model is/does. -->
|
8 |
+
|
9 |
+
FalconAlpaca is Falcon-7B trained on the [Stanford Alpaca Dataset](https://github.com/tatsu-lab/stanford_alpaca/blob/main/alpaca_data.json)
|
10 |
+
|
11 |
+
## Model Details
|
12 |
+
|
13 |
+
This model was an attempt to influence the learned outputs of Falcon-7B to adapt the outputs to become more information-rich and focused.
|
14 |
+
Trained using [Lit GPT](https://github.com/Lightning-AI/lit-gpt), the model took 2 hours to train on 1 4xA6000 node.
|
15 |
+
|
16 |
+
|
17 |
+
### Model Description
|
18 |
+
|
19 |
+
- **License:** [Apache 2.0]
|
20 |
+
- **Finetuned from model :** [Falcon-7B](https://huggingface.co/tiiuae/falcon-7b)
|
21 |
+
|
22 |
+
### Model Sources
|
23 |
+
|
24 |
+
[Stanford Alpaca Dataset](https://github.com/tatsu-lab/stanford_alpaca/blob/main/alpaca_data.json)
|
25 |
+
|
26 |
+
### Out-of-Scope Use
|
27 |
+
|
28 |
+
This model is not intended for anything but testing purposes. There have been no attempts to control/remove bias, toxicity, or any other form of
|
29 |
+
potentially dangerous or harmful messages.
|
30 |
+
|
31 |
+
## Bias, Risks, and Limitations
|
32 |
+
|
33 |
+
No effort was made to remove any wrong or harmful information from Falcon-7B or the Alpaca dataset. Any risks and limitations from either of
|
34 |
+
those datasets/models carry over to this project as well.
|
35 |
+
|
36 |
+
## How to Get Started with the Model
|
37 |
+
|
38 |
+
Download and install libraries for [Lit GPT](https://github.com/Lightning-AI/lit-gpt)
|
39 |
+
|
40 |
+
```sh
|
41 |
+
python generate/adapter_v2.py \
|
42 |
+
--adapter_path path/to/model/lit_model_adapter_finetuned.pth \
|
43 |
+
--checkpoint_dir path/to/model \
|
44 |
+
--prompt "What temperature should I cook pork at to ensure it is safe?"
|
45 |
+
```
|
46 |
+
|
47 |
+
This uses around 14GB of VRAM. If you need to use less you can read [this](https://lightning.ai/pages/blog/falcon-a-guide-to-finetune-and-inference/)
|
48 |
+
|
49 |
+
### Training Data
|
50 |
+
|
51 |
+
[Stanford Alpaca Dataset](https://github.com/tatsu-lab/stanford_alpaca/blob/main/alpaca_data.json)
|
52 |
+
|
53 |
+
### Training Procedure
|
54 |
+
|
55 |
+
Followed the guide [here](https://lightning.ai/pages/blog/falcon-a-guide-to-finetune-and-inference/)
|
56 |
+
|
57 |
+
|
58 |
+
#### Training Hyperparameters
|
59 |
+
|
60 |
+
The defaults were as follows
|
61 |
+
```
|
62 |
+
learning_rate = 9e-3
|
63 |
+
batch_size = 32
|
64 |
+
micro_batch_size = 2
|
65 |
+
gradient_accumulation_iters = 16
|
66 |
+
epoch_size = 50000
|
67 |
+
num_epochs = 5
|
68 |
+
max_iters = 125000
|
69 |
+
weight_decay = 0.02
|
70 |
+
warmup_iters = 50000
|
71 |
+
```
|
72 |
+
|
73 |
+
|
74 |
+
|
75 |
+
## Evaluation
|
76 |
+
|
77 |
+
[More Information Needed]
|