botnet26t_256.c1_in1k
| Property | Value |
|---|---|
| Parameter Count | 12.5M |
| Model Type | Image Classification |
| License | Apache-2.0 |
| Image Size | 256 x 256 |
| GMACs | 3.3 |
| Framework | PyTorch (timm) |
What is botnet26t_256.c1_in1k?
botnet26t_256.c1_in1k is a sophisticated image classification model that combines the power of ResNet architecture with self-attention mechanisms. Developed using the timm framework, this model represents a balanced approach to visual recognition, offering 12.5M parameters and optimized performance for 256x256 image inputs.
Implementation Details
The model is built using timm's flexible BYOBNet (Bring-Your-Own-Blocks Network) framework, incorporating elements from both the Bottleneck Transformers paper and ResNet Strikes Back methodology. It utilizes SGD with Nesterov momentum and adaptive gradient clipping, alongside a cosine learning rate schedule with warmup.
- Customizable block and stage layout
- Flexible attention mechanisms integration
- Stochastic depth implementation
- Gradient checkpointing support
- Layer-wise learning rate decay
Core Capabilities
- Image classification with high efficiency
- Feature map extraction across multiple scales
- Image embedding generation
- Per-stage feature extraction
- Support for both training and inference workflows
Frequently Asked Questions
Q: What makes this model unique?
This model uniquely combines ResNet architecture with self-attention blocks, offering a balanced approach between computational efficiency and performance. It's specifically tuned for reduced frequency of self-attention blocks while maintaining reasonable training times.
Q: What are the recommended use cases?
The model excels in image classification tasks, feature extraction, and generating image embeddings. It's particularly well-suited for applications requiring 256x256 image processing with moderate computational resources.
