|
Mila 0.13.48
Deep Neural Network Library
|
Process-wide shared cos/sin cache registry for CudaRopeOp. More...
#include <cuda_runtime.h>#include <mutex>#include <unordered_map>#include <bit>#include <cstdint>import Cuda.Error;import Dnn.TensorDataType;Classes | |
| struct | Mila::Dnn::Compute::Cuda::Rope::RopeCacheRegistry::AcquireResult |
| struct | Mila::Dnn::Compute::Cuda::Rope::RopeCacheRegistry::CacheEntry |
| struct | Mila::Dnn::Compute::Cuda::Rope::RopeCacheRegistry::CacheKey |
| struct | Mila::Dnn::Compute::Cuda::Rope::RopeCacheRegistry::CacheKeyHash |
| class | Mila::Dnn::Compute::Cuda::Rope::RopeCacheRegistry |
| Process-wide shared cache for RoPE cos/sin frequency tables. More... | |
Namespaces | |
| namespace | Mila |
| Mila main API namespace. | |
| namespace | Mila::Dnn |
| namespace | Mila::Dnn::Compute |
| namespace | Mila::Dnn::Compute::Cuda |
| namespace | Mila::Dnn::Compute::Cuda::Rope |
Process-wide shared cos/sin cache registry for CudaRopeOp.
Provides RopeCacheRegistry, an implementation detail of Compute.CudaRopeOp. Not exported to consumers of the module.