Final 500M model (loss=11.2343)
Browse files
README.md
ADDED
|
@@ -0,0 +1,28 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
tags:
|
| 4 |
+
- tritter
|
| 5 |
+
- bitnet
|
| 6 |
+
- code
|
| 7 |
+
- 500m
|
| 8 |
+
---
|
| 9 |
+
|
| 10 |
+
# Tritter 500M BitNet
|
| 11 |
+
|
| 12 |
+
A 500M parameter BitNet b1.58 ternary-quantized model trained for code generation.
|
| 13 |
+
|
| 14 |
+
## Training Details
|
| 15 |
+
|
| 16 |
+
- **Parameters**: 524,372,480
|
| 17 |
+
- **Training tokens**: 118,111,072
|
| 18 |
+
- **Final loss**: 11.2343
|
| 19 |
+
- **Min loss**: 11.0722
|
| 20 |
+
- **Tokens/sec**: 23679.4
|
| 21 |
+
- **Training duration**: 1:23:07.915359
|
| 22 |
+
- **GPU**: NVIDIA GeForce RTX 5080
|
| 23 |
+
|
| 24 |
+
## Checkpoints
|
| 25 |
+
|
| 26 |
+
Intermediate checkpoints available at 10%, 20%, ..., 90% progress.
|
| 27 |
+
|
| 28 |
+
Generated with [Tritter](https://github.com/tzervas/tritter)
|