Update README.md
Browse files
README.md
CHANGED
|
@@ -20,7 +20,7 @@ The GGUFs uploaded are full FP16 precision.
|
|
| 20 |
- 40M parameters
|
| 21 |
- 8 attention heads
|
| 22 |
- 28 layers
|
| 23 |
-
- 1536
|
| 24 |
|
| 25 |
## Prompt Format:
|
| 26 |
```
|
|
@@ -38,8 +38,8 @@ Ensure clarity and practicality, allowing readers to easily follow and apply the
|
|
| 38 |
## Training Details:
|
| 39 |
- 1x RTX 3070 8GB
|
| 40 |
- 1x Ryzen 3 3700x
|
| 41 |
-
-
|
| 42 |
-
- Approx
|
| 43 |
- Training data = 1 billion tokens
|
| 44 |
|
| 45 |
## Notes:
|
|
|
|
| 20 |
- 40M parameters
|
| 21 |
- 8 attention heads
|
| 22 |
- 28 layers
|
| 23 |
+
- 4096 context (upgraded from 1536, please expect a temporary performance drop)
|
| 24 |
|
| 25 |
## Prompt Format:
|
| 26 |
```
|
|
|
|
| 38 |
## Training Details:
|
| 39 |
- 1x RTX 3070 8GB
|
| 40 |
- 1x Ryzen 3 3700x
|
| 41 |
+
- 7660 iterations
|
| 42 |
+
- Approx 100 million tokens/120k samples (>0.01 epoches)
|
| 43 |
- Training data = 1 billion tokens
|
| 44 |
|
| 45 |
## Notes:
|