You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hello, the model I use is MMedLM2-1_8B. In order to execute the entire process faster, I deleted most of the test cases in the jsonl file in the Test folder, and only kept 30 questions for each language. After I executed inference.py, I got six files ending with _res.txt. I opened these files and checked the generated results. I found that there were a lot of duplicate content in each line of results. This situation is very obvious in the Chinese results. Do you have any suggestions?
Hello! MMedLM2-1_8B is a foundation model that has undergone extended pretraining based on InternLM2-1.8B. Please note that it has not been fine-tuned.
In our paper, all results in the multilingual setting were obtained using a fine-tuned version of the model. The fine-tuning was conducted on the training set of MMedBench, following the process outlined in this code repository.
We recommend fine-tuning the model on the MMedBench training set or another appropriate dataset before using it for downstream tasks or evaluation.
Hello, the model I use is MMedLM2-1_8B. In order to execute the entire process faster, I deleted most of the test cases in the jsonl file in the Test folder, and only kept 30 questions for each language. After I executed
inference.py
, I got six files ending with _res.txt. I opened these files and checked the generated results. I found that there were a lot of duplicate content in each line of results. This situation is very obvious in the Chinese results. Do you have any suggestions?Generated content (Chinese language):
The text was updated successfully, but these errors were encountered: