Optimize the initialization method of the MonkeyChat class to add the functionality of automatically configuring data types to support different CUDA computing capabilities.

[liwenju0]

No description provided.

[zenosai]

We have merged this PR, thanks for your contribution.

[zenosai]

Thanks again! Also, could you share which GPU you were using during deployment, and whether the deployment was successful? That would help us refine support for different hardware setups and better understand potential issues. Appreciate your input! @liwenju0

[liwenju0]

Thanks again! Also, could you share which GPU you were using during deployment, and whether the deployment was successful? That would help us refine support for different hardware setups and better understand potential issues. Appreciate your input! @liwenju0

I deployed it on my 2080ti with 22G of memory on a single card (magically modified)

[zenosai]

Thanks for your reply!
Just wondering — did you encounter any shared memory errors during deployment? (See: https://github.com/Yuliang-Liu/MonkeyOCR#fix-shared-memory-error-on-rtx-3090--4090---gpus-optional)
This issue tends to occur on RTX 3090 and 4090 GPUs.