In order to build our latency prediction model, We test three types of hardware devices, NVIDIA V100, NVIDIA GTX 2080, and NVIDIA GTX 1080. Their respective properties are presented in Table 6. It shows that the server GPU V100 is the most powerful hardware device with the most processing engines (#PE). We map the operations to hardware. These split tiles are assigned to multiple PEs.

We provide the details of the proposed latency-driven fast slimming in Alg. 1. Formulations of the Our major conclusions and speed analysis can be found in Sec. 3 and Figure 1. Compared to non-overlap large-kernel patch embedding (V5 in Tab. MHSA with the global receptive field is an essential contribution to model performance. By comparing V1 and V2 in Tab. 3, we can observe that the GN We explore ReLU and HardSwish (V3 and V4 in Tab. 3) in addition to GeLU We draw a conclusion that the activation function can be selected on a case-by-case basis depending on the specific hardware and compiler. In this work, we use GeLU to provide better performance than ReLU while executing faster.

The tokenizer, as one of the fundamental components of large models, has long been overlooked or even misunderstood in visual tasks. One key factor of the great comprehension power of the large language model is that natural language tokenizers utilize meaningful words or subwords as the basic elements of language. In contrast, mainstream visual tokenizers, represented by patch-based methods such as Patch Embed, rely on meaningless rectangular patches as basic elements of vision, which cannot serve as effectively as words or subwords in language. Starting from the essence of the tokenizer, we defined semantically independent regions (SIRs) for vision. We designed a simple HOmogeneous visual tOKenizer: HOOK. HOOK mainly consists of two modules: the Object Perception Module (OPM) and the Object Vectorization Module (OVM). To achieve homogeneity, the OPM splits the image into 4*4 pixel seeds and then utilizes the attention mechanism to perceive SIRs. The OVM employs cross-attention to merge seeds within the same SIR. To achieve adaptability, the OVM defines a variable number of learnable vectors as cross-attention queries, allowing for the adjustment of token quantity. We conducted experiments on the NWPU-RESISC45, WHU-RS19 classification dataset, and GID5 segmentation dataset for sparse and dense tasks. The results demonstrate that the visual tokens obtained by HOOK correspond to individual objects, which demonstrates homogeneity. HOOK outperformed Patch Embed by 6\% and 10\% in the two tasks and achieved state-of-the-art performance compared to the baselines used for comparison. Compared to Patch Embed, which requires more than one hundred tokens for one image, HOOK requires only 6 and 8 tokens for sparse and dense tasks, respectively, resulting in efficiency improvements of 1.5 to 2.8 times. The code is available at https://github.com/GeoX-Lab/Hook.