working-group-ideas

57 threads · Page 2 of 2

Maxwell Equations Simulator 41 messages
Multi-GPU
Add Support for <new type> KV Cache Quantization in TorchAO 3 messages
QuantizationSparsity
Optimize Quantization Settings to Fit a Given VRAM Budget 2 messages
QuantizationSparsity
Add an activation sparsity kernel to TorchAO 9 messages
QuantizationSparsity
Develop Fused Quantized GEMM/GEMV with LoRA 14 messages
QuantizationSparsity
Implement an LUT-based n-bit Quantization (nf format) Fused Matmul Kernel 3 messages
QuantizationSparsity
Develop an A16W3 (mixed fp16 x 3-bit) Fused Matmul Kernel 10 messages
QuantizationSparsity