Update README.md
Browse files
README.md
CHANGED
|
@@ -190,10 +190,12 @@ del AkaLlama-llama3-70b-v0.1.Q8_0.00001-of-00002.gguf AkaLlama-llama3-70b-v0.1.Q
|
|
| 190 |
|
| 191 |
## Evaluation
|
| 192 |
|
| 193 |
-
| Model | #Parameter | Qunatized? | LogicKor |
|
| 194 |
-
|
| 195 |
-
| AkaLlama-llama3-70b-v0.1-GGUF.Q4 | 70B |
|
| 196 |
-
| AkaLlama-llama3-70b-v0.1-GGUF.Q8 | 70B |
|
|
|
|
|
|
|
| 197 |
|
| 198 |
## Training Details
|
| 199 |
### Training Procedure
|
|
@@ -343,14 +345,14 @@ print(output.shape)
|
|
| 343 |
|
| 344 |
You can find more examples at [our project page](https://yonsei-mir.github.io/AkaLLaMA-page)
|
| 345 |
|
| 346 |
-
## Warning
|
| 347 |
-
|
| 348 |
-
Although AKALlama-70B has significant potential, its responses can sometimes be inaccurate, biased, or misaligned, presenting risks if used without additional testing and refinement. Furthermore, the quality of the model's output is greatly influenced by the system prompt and decoding strategy. Changes in these areas could result in less precise outputs. Therefore, we strongly recommend handling our model with considerable caution.
|
| 349 |
-
|
| 350 |
## Special Thanks
|
| 351 |
|
| 352 |
- Data Center of the Department of Artificial Intelligence and Jeong Mee Koh at Yonsei University for the computation resources
|
| 353 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 354 |
## Comments
|
| 355 |
|
| 356 |
- Title image generated by DALL·E 3
|
|
|
|
| 190 |
|
| 191 |
## Evaluation
|
| 192 |
|
| 193 |
+
| Model | #Parameter | Qunatized? | LogicKor* |
|
| 194 |
+
|:--------------------------------:|:----------:|------------|:---------:|
|
| 195 |
+
| AkaLlama-llama3-70b-v0.1-GGUF.Q4 | 70B | 4bit | 6.56 |
|
| 196 |
+
| AkaLlama-llama3-70b-v0.1-GGUF.Q8 | 70B | 8bit | 6.34 |
|
| 197 |
+
|
| 198 |
+
*mean over 3 random seeds
|
| 199 |
|
| 200 |
## Training Details
|
| 201 |
### Training Procedure
|
|
|
|
| 345 |
|
| 346 |
You can find more examples at [our project page](https://yonsei-mir.github.io/AkaLLaMA-page)
|
| 347 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 348 |
## Special Thanks
|
| 349 |
|
| 350 |
- Data Center of the Department of Artificial Intelligence and Jeong Mee Koh at Yonsei University for the computation resources
|
| 351 |
|
| 352 |
+
## Warning
|
| 353 |
+
|
| 354 |
+
Although AKALlama-70B has significant potential, its responses can sometimes be inaccurate, biased, or misaligned, presenting risks if used without additional testing and refinement. Furthermore, the quality of the model's output is greatly influenced by the system prompt and decoding strategy. Changes in these areas could result in less precise outputs. Therefore, we strongly recommend handling our model with considerable caution.
|
| 355 |
+
|
| 356 |
## Comments
|
| 357 |
|
| 358 |
- Title image generated by DALL·E 3
|