Update README.md
Browse files
README.md
CHANGED
|
@@ -106,7 +106,7 @@ The current gguf version tends to add hallucinations after translation and the p
|
|
| 106 |
- -e (改行\nをエスケープ)
|
| 107 |
- --temp 0 (最も確率の高いトークンのみを選択)
|
| 108 |
- --repeat-penalty 1.0 (繰り返しペナルティをオフ。指示調整済モデルでこれをするのは、決して良い考えとは言えないとの事。)
|
| 109 |
-
- ~~--no-penalize-nl (改行の繰り返しにはペナルティをあたえない)~~ 最新のllama.cpp
|
| 110 |
|
| 111 |
Adjust the following parameters as needed
|
| 112 |
- Temperature (--temp): Lowering this value will make the model more likely to select more confident (i.e., more common) words.
|
|
@@ -117,4 +117,5 @@ The following are the [recommended parameters](https://huggingface.co/google/gem
|
|
| 117 |
- -e (escape newlines (\n))
|
| 118 |
- --temp 0(pick most probable tokens)
|
| 119 |
- --repeat-penalty 1.0(disable repetition penalty (it's never a good idea to have this with instruction tuned models)~~ latest llama.cpp default behavior, so don't mind.
|
| 120 |
-
- ~~--no-penalize-nl(do not penalize repeating newlines)
|
|
|
|
|
|
| 106 |
- -e (改行\nをエスケープ)
|
| 107 |
- --temp 0 (最も確率の高いトークンのみを選択)
|
| 108 |
- --repeat-penalty 1.0 (繰り返しペナルティをオフ。指示調整済モデルでこれをするのは、決して良い考えとは言えないとの事。)
|
| 109 |
+
- ~~--no-penalize-nl (改行の繰り返しにはペナルティをあたえない)~~ 最新のllama.cppではディフォルト動作になったので指定不要
|
| 110 |
|
| 111 |
Adjust the following parameters as needed
|
| 112 |
- Temperature (--temp): Lowering this value will make the model more likely to select more confident (i.e., more common) words.
|
|
|
|
| 117 |
- -e (escape newlines (\n))
|
| 118 |
- --temp 0(pick most probable tokens)
|
| 119 |
- --repeat-penalty 1.0(disable repetition penalty (it's never a good idea to have this with instruction tuned models)~~ latest llama.cpp default behavior, so don't mind.
|
| 120 |
+
- ~~--no-penalize-nl(do not penalize repeating newlines)~~ latest llama.cpp's default behavior so you need not this option.
|
| 121 |
+
|