Update README.md
Browse files
README.md
CHANGED
|
@@ -2,7 +2,7 @@ Eagle Speculative Decoding Model Trained with BaldEagle
|
|
| 2 |
|
| 3 |
BaldEagle Repo: https://github.com/NickL77/BaldEagle/
|
| 4 |
|
| 5 |
-
Learn how the model was trained:
|
| 6 |
|
| 7 |
Achieves 2.06x speed up (50.43 tok/s -> 106.22 tok/s) on Qwen2.5-7B-Instruct.
|
| 8 |
Better than sglang [reported](https://docs.sglang.ai/backend/speculative_decoding.html#Performance-Highlights) improvements (1.54x) with EAGLE 2.
|
|
|
|
| 2 |
|
| 3 |
BaldEagle Repo: https://github.com/NickL77/BaldEagle/
|
| 4 |
|
| 5 |
+
Learn how the model was trained: https://frugalgpu.substack.com/p/how-to-train-your-own-eagle-speculative
|
| 6 |
|
| 7 |
Achieves 2.06x speed up (50.43 tok/s -> 106.22 tok/s) on Qwen2.5-7B-Instruct.
|
| 8 |
Better than sglang [reported](https://docs.sglang.ai/backend/speculative_decoding.html#Performance-Highlights) improvements (1.54x) with EAGLE 2.
|