-
Notifications
You must be signed in to change notification settings - Fork 4
/
nerr.proto
23 lines (21 loc) · 1.45 KB
/
nerr.proto
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
syntax = "proto3";
package nrt;
// IMPORTANT! This is a copy of tdrv/nerr.h please keep in sync
enum nerr {
NERR_OK = 0;
NERR_FAIL = 1; // non specific failure, don't use if there is more descriptive type
NERR_INVALID = 2; // e.g. invalid NEFF, bad instruction, bad DMA descriptor, input tensor name/size does not match the model, etc.
NERR_INVALID_HANDLE = 3; // make this one explicit instead of using more generic INVALID_INPUT because it could be a common caller mistake
NERR_RESOURCE = 4; // failed to allocate a resource for requested operation
NERR_TIMEOUT = 5; // operation timed out
NERR_HW_ERROR = 6; // hardware error
NERR_QUEUE_FULL = 7; // not enough space in the inference input queue
NERR_FAILED_STATE = 8; // model was stopped due to unrecoverable error
NERR_RESOURCE_NC = 9; // not enough NCs for requested operation (nn load or eg creation)
NERR_UNSUPPORTED_VERSION = 10; // unsupported version of NEFF
NERR_INFER_PENDING = 1000; // inference request still pending
NERR_INFER_CANCELLED = 1001; // inference request has been canceled
NERR_INFER_BAD_INPUT = 1002; // invalid input has been submitted to infer()
NERR_INFER_COMPLETED_WITH_NUM_ERR = 1003; // inference was completed with numerical errors (NaN)
NERR_INFER_COMPLETED_WITH_ERR = 1004; // inference was completed with non-numerical errors
}