fyang064

Transfer learning issue

1

Hello, I'm wondering how to fix the loading state_dict issue while trying to use pre-trained model. Once the num_classes of the custom dataset has changed from 91, size mismatches for...

Hi Dr. Dai, I have seen your wonderful work on infrared small target detection. I'm wondering if your model could be applied on other infrared images like thermal frames for...

how to train Gemini

1

Just wondering anything special when training Gemini, like is there any differences train on the multimodal data? ## Upvote & Fund - We're using [Polar.sh](https://polar.sh/kyegomez) so you can upvote and...

Limitations of the method

2

Just wondering if any limitations of the Infini-attention like inference speed and model performance. Not too much discussions in the paper.

How about the cost of TUTEL features?

1

I'm wondering the cost of features mentioned in the TUTEL paper. It looks like the dynamic features including top-anything as well as the dynamic capacity factor will introduce the additional...

In the Table 4 of the paper, GPU total number **16384** is not matching with the parallelism group [8, 16, 16, 4]. Is this a mistake in the paper?

I saw the compositional approach adding multimodal capabilities to Llama3 in the report, and am curious about the details about the image encoder and adaptor. Can you please provide any...