Create README.md (#1)
Browse files- Create README.md (3fdf3fca1400c2767dd14fd7f0831d6a848f9a54)
Co-authored-by: YanGPT <Chasell@users.noreply.huggingface.co>
README.md
ADDED
|
@@ -0,0 +1,22 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# OpenLLaMA 7Bv2 Model Card
|
| 2 |
+
|
| 3 |
+
## Model Description
|
| 4 |
+
|
| 5 |
+
OpenLLaMA 7Bv2 is a cutting-edge language model, trained with a focus on delivering high-quality, contextually relevant text predictions. It leverages a diverse composite dataset that includes web-crawled data, scholarly articles, and a wide range of literature and question-answer pairs to ensure broad domain coverage and applicability.
|
| 6 |
+
|
| 7 |
+
## Training Data
|
| 8 |
+
|
| 9 |
+
The model was trained on a composite dataset that includes:
|
| 10 |
+
|
| 11 |
+
- Falcon refined-web dataset
|
| 12 |
+
- starcoder datasets
|
| 13 |
+
- Contributions from Wikipedia for encyclopedic knowledge
|
| 14 |
+
- Academic papers from arXiv for scientific understanding
|
| 15 |
+
- A vast collection of books spanning multiple genres
|
| 16 |
+
- Stack Exchange data curated by RedPajama
|
| 17 |
+
|
| 18 |
+
## Training Procedure
|
| 19 |
+
|
| 20 |
+
- **Learning Rate:** Utilized a maximum learning rate of 3e-4 and a minimum learning rate of 3e-5.
|
| 21 |
+
- **Batch Size:** Employed a batch size of 4 million tokens, optimizing the training process for both efficiency and performance.
|
| 22 |
+
- **Learning Rate Scheduler:** The model's learning rate scheduling closely follows the strategy used in Llama2, ensuring gradual adjustments for optimal convergence.
|