Application programming interface to generate a tensor according to a tensor map
Apparatuses, systems, and techniques to cause a first tensor to be translated into a second tensor according to a tensor map without storing information about a memory transaction corresponding to the translation. In at least one embodiment, one or more circuits are to perform an application programming interface (API) to cause a first tensor to be translated into a second tensor according to a tensor map without storing information about one or more memory transactions corresponding to the translation.
1 . A processor, comprising: one or more circuits to perform an application programming interface (API) to cause a first tensor to be translated into a second tensor according to a tensor map without storing information about one or more memory transactions corresponding to the translation.
2 . The processor of claim 1 , wherein the one or more circuits are to perform the API to cause the first tensor to be translated based, at least in part, on asynchronously storing second tensor in one or more second memory locations based, at least in part, on first tensor data stored in one or more first memory locations.
3 . The processor of claim 1 , wherein the one or more circuits are to perform the API to cause the first tensor to be translated based, at least in part, on asynchronously storing second tensor in one or more second memory locations of a graphics processing unit (GPU) based, at least in part, on first tensor data stored in one or more first memory locations of the GPU.
4 . The processor of claim 1 , wherein the one or more circuits are to perform the API to cause the first tensor to be translated based, at least in part, on a data structure that includes a tensor map, and the one or more circuits are to perform the API to cause the second tensor to be asynchronously stored.
5 . The processor of claim 1 , wherein the API is to cause the first tensor to be translated into the second tensor without storing the information by using manual transaction accounting.
6 . The processor of claim 1 , wherein the API is to be performed using one or more asynchronous memory transactions.
7 . The processor of claim 1 , wherein the API is to receive as input a data structure to indicate how to translate the first tensor into the second tensor.
8 . A system, comprising: one or more processors to perform an application programming interface (API) to cause a first tensor to be translated into a second tensor according to a tensor map without storing information about one or more memory transactions corresponding to the translation.
9 . The system of claim 8 , wherein the API is to cause the first tensor to be translated based, at least in part, on asynchronously storing second tensor in one or more second memory locations based, at least in part, on first tensor data stored in one or more first memory locations.
10 . The system of claim 8 , wherein the API is to cause the first tensor to be translated based, at least in part, on asynchronously storing second tensor in one or more second memory locations of a graphics processing unit (GPU) based, at least in part, on first tensor data stored in one or more first memory locations of the GPU.
11 . The system of claim 8 , wherein the API is to cause the first tensor to be translated based, at least in part, on a data structure that includes a tensor map, and the one or more circuits are to perform the API to cause the second tensor to be asynchronously stored.
12 . The system of claim 8 , wherein the API is to cause the first tensor to be translated into the second tensor without storing the information by using manual transaction accounting.
13 . The system of claim 8 , the API is to receive as input a data structure to indicate how to translate the first tensor into the second tensor.
14 . A method, comprising: performing an application programming interface (API) to cause a first tensor to be translated into a second tensor according to a tensor map without storing information about one or more memory transactions corresponding to the translation.
15 . The method of claim 14 , wherein performing the API comprises causing asynchronous storage of second tensor of the second tensor in one or more second memory locations based, at least in part, on first tensor data of the first tensor stored in one or more first memory locations.
16 . The method of claim 14 , wherein performing the API comprises causing asynchronous storage of second tensor in one or more second memory locations of a graphics processing unit (GPU) based, at least in part, on first tensor data stored in one or more first memory locations of the GPU.
17 . The method of claim 14 , performing the API comprises causing the first tensor to be translated based, at least in part, on a data structure that comprises a tensor map.
18 . The method of claim 14 , wherein performing the API comprises using a type of transaction accounting different from automatic transaction accounting.
19 . The method of claim 14 , wherein performing the API to cause the first tensor to be translated into the second tensor without storing the information uses manual transaction accounting.
20 . A non-transitory computer-readable medium having stored thereon a set of instructions, which if performed by one or more processors, cause the one or more processors to at least perform the method of claim 14 .