Update README.md
Browse files
README.md
CHANGED
@@ -7,15 +7,14 @@ license: apache-2.0
|
|
7 |
Welcome to ***GEM_Testing_Arsenal***, where groundbreaking research meets practical power! This repository unveils a novel architecture for On-Device Language Models (ODLMs), straight from our paper, ["Fragile Mastery: are domain-specific trade-offs undermining On-Device Language Models?"](./link_to_be_insterted). With just a few lines of code, our custom `gem_trainer.py` script lets you train ODLMs that are more accurate than ever, tracking accuracy and loss as you go.
|
8 |
|
9 |
---
|
10 |
-
|
11 |
-
### Highlights:
|
12 |
- **Next-Level ODLMs**: Boosts accuracy with a new architecture from our research.
|
13 |
- **Easy Training**: Call run_gem_pipeline to train on your dataset in minutes.
|
14 |
- **Live Metrics**: Get accuracy and loss results as training unfolds.
|
15 |
- **Flexible Design**: Works with any compatible dataset—plug and play!
|
16 |
|
17 |
---
|
18 |
-
|
19 |
To dive in, you’ll need:
|
20 |
- **Python** `3.8+`
|
21 |
|
@@ -24,11 +23,11 @@ To dive in, you’ll need:
|
|
24 |
- **Git** *(to clone the repo)*
|
25 |
|
26 |
---
|
27 |
-
|
28 |
|
29 |
1. **Clone the repository:**
|
30 |
```bash
|
31 |
-
git clone https://huggingface.co/GEM025/
|
32 |
```
|
33 |
|
34 |
2. **Install Dependencies:**
|
@@ -52,13 +51,54 @@ Create a new python file and execute the code like:
|
|
52 |
```
|
53 |
|
54 |
> ***Boom—your ODLM is training with boosted accuracy!***
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
55 |
|
56 |
---
|
57 |
-
|
58 |
`run_gem_pipeline` keeps it simple, but you can tweak it! Dive into [`gem_trainer.py`](./gem_trainer.py) to adjust epochs, batch size, or other settings to fit your needs.
|
59 |
|
60 |
---
|
61 |
-
|
62 |
Got ideas to make this even better? We’re all ears!
|
63 |
- Fork the repo.
|
64 |
- Branch off (`git checkout -b your-feature`).
|
|
|
7 |
Welcome to ***GEM_Testing_Arsenal***, where groundbreaking research meets practical power! This repository unveils a novel architecture for On-Device Language Models (ODLMs), straight from our paper, ["Fragile Mastery: are domain-specific trade-offs undermining On-Device Language Models?"](./link_to_be_insterted). With just a few lines of code, our custom `gem_trainer.py` script lets you train ODLMs that are more accurate than ever, tracking accuracy and loss as you go.
|
8 |
|
9 |
---
|
10 |
+
## Highlights:
|
|
|
11 |
- **Next-Level ODLMs**: Boosts accuracy with a new architecture from our research.
|
12 |
- **Easy Training**: Call run_gem_pipeline to train on your dataset in minutes.
|
13 |
- **Live Metrics**: Get accuracy and loss results as training unfolds.
|
14 |
- **Flexible Design**: Works with any compatible dataset—plug and play!
|
15 |
|
16 |
---
|
17 |
+
## Prerequisites:
|
18 |
To dive in, you’ll need:
|
19 |
- **Python** `3.8+`
|
20 |
|
|
|
23 |
- **Git** *(to clone the repo)*
|
24 |
|
25 |
---
|
26 |
+
## Quick Start:
|
27 |
|
28 |
1. **Clone the repository:**
|
29 |
```bash
|
30 |
+
git clone https://huggingface.co/GEM025/GEM_Arsenal
|
31 |
```
|
32 |
|
33 |
2. **Install Dependencies:**
|
|
|
51 |
```
|
52 |
|
53 |
> ***Boom—your ODLM is training with boosted accuracy!***
|
54 |
+
---
|
55 |
+
## Running on Colab/Kaggle?
|
56 |
+
|
57 |
+
Well it's pretty similar to the local run.
|
58 |
+
|
59 |
+
```python
|
60 |
+
""" This is very recommended to run for clean ouput during trains...
|
61 |
+
|
62 |
+
import warnings
|
63 |
+
warnings.filterwarnings('ignore')
|
64 |
+
|
65 |
+
"""
|
66 |
+
|
67 |
+
#@ Step 1: Clone the github repo
|
68 |
+
!git clone https://huggingface.co/GEM025/GEM_Arsenal
|
69 |
+
|
70 |
+
#@ Step 2: Install all requirements
|
71 |
+
!pip install -r /content/GEM/requirements.txt #! For colab
|
72 |
+
|
73 |
+
"""
|
74 |
+
|
75 |
+
@! For kaggle:
|
76 |
+
!pip install -r /kaggle/working/GEM/requirements.txt
|
77 |
+
|
78 |
+
"""
|
79 |
+
|
80 |
+
#@ Step 3: Add repo to path
|
81 |
+
import sys
|
82 |
+
sys.path.append('/content/GEM') #! Or /kaggle/working/GEM (for kaggle)
|
83 |
+
|
84 |
+
#@ Step 4: Import and run function
|
85 |
+
from gem_trainer import run_gem_pipeline
|
86 |
+
from datasets import load_dataset
|
87 |
+
|
88 |
+
#@ Rest of the code as above
|
89 |
+
dataset = load_dataset("imdb")
|
90 |
+
|
91 |
+
result = run_gem_pipeline(dataset, num_classes=2, num_epochs=2)
|
92 |
+
|
93 |
+
print(result)
|
94 |
+
```
|
95 |
|
96 |
---
|
97 |
+
## Customizing Training:
|
98 |
`run_gem_pipeline` keeps it simple, but you can tweak it! Dive into [`gem_trainer.py`](./gem_trainer.py) to adjust epochs, batch size, or other settings to fit your needs.
|
99 |
|
100 |
---
|
101 |
+
## Contributing 💓
|
102 |
Got ideas to make this even better? We’re all ears!
|
103 |
- Fork the repo.
|
104 |
- Branch off (`git checkout -b your-feature`).
|