Roadmap

More OP

• Add
• AvgPool2D
• Backend
• Cat
• Causal Mask
• Convolution2D
• Convolution3D
• Division
• Embedding
• GELU
• LayerNorm
• Linear
• Matmul
• MaxPool2D
• Mul
• QuickGELU
• RMSNorm
• ReLU
• ReLU Squared
• RoPE
• Scale
• SiLU
• SoftMax
• Split
• Transpose
• View
• Convelution1D
• Depth-wise Convolution
• Batch Normalization

More Backends

• ARM Neon
• QNN
• OpenCL
• Vulkan
• CUDA

More Models

• Llama
• Alpaca
• Persimmon
• fuyu
• Vit
• ImageBind
• TinyLlama
• Stable Diffusion
• OPT
• CLIP
• PandaGPT
• NextGPT
• LLaVA
• sheared llama
• baby llama
• OpenBuddy
• Falcon
• WizardCoder
• phi-2
• Whisper
• BLIP2
• MiniGPT-4
• baichuan
• Bakllava
• Mistral
• QWen
• Kosmos-2
• NeuralBeagle14-7B
• Yi-VL
• Gemma
• MiniCPM
• MobileVLM
• QWen-VL

Improvement

• Sparsity Inference
• Memory-efficient Inference