diff options
Diffstat (limited to 'tensorflow/compiler/xla/xla_data.proto')
-rw-r--r-- | tensorflow/compiler/xla/xla_data.proto | 26 |
1 files changed, 3 insertions, 23 deletions
diff --git a/tensorflow/compiler/xla/xla_data.proto b/tensorflow/compiler/xla/xla_data.proto index eac8f2ff07..06987e0044 100644 --- a/tensorflow/compiler/xla/xla_data.proto +++ b/tensorflow/compiler/xla/xla_data.proto @@ -46,12 +46,6 @@ enum PrimitiveType { // converted to f16 from f32 at arbirary points in the computation. F16 = 10; F32 = 11; - - // Truncated 16 bit floating-point format. This is similar to IEEE's 16 bit - // floating-point format, but uses 1 bit for the sign, 8 bits for the exponent - // and 7 bits for the mantissa. - BF16 = 16; - F64 = 12; // Complex values of fixed width. @@ -69,8 +63,6 @@ enum PrimitiveType { // An opaque type used for passing context specific data to a custom // operation. OPAQUE = 14; - - // Next = 17 } // Describes the value held inside padding elements. @@ -318,10 +310,7 @@ message LiteralProto { repeated double f64s = 9; repeated float c64s = 12; // Stored as interleaved real, imag floats. repeated LiteralProto tuple_literals = 10; - // The F16s and BF16s are encoded in little endian byte order - bytes f16s = 11; - bytes bf16s = 13; - // Next = 14 + bytes f16s = 11; // Note: the F16s are encoded in little endian byte order } message WindowDimension { @@ -836,10 +825,8 @@ message OpSharding { REPLICATED = 0; // This sharding is maximal - one device runs the entire operation. MAXIMAL = 1; - // This sharding is a tuple - only the tuple_shardings field is valid. - TUPLE = 2; - // None of the above; tile_shape and tile_assignment are both used. - OTHER = 3; + // Neither of the above; tile_shape and tile_assignment are both used. + OTHER = 2; } Type type = 1; // The shape of the sharded tile. @@ -851,13 +838,6 @@ message OpSharding { // Flattened list of device IDs. The order of flattening is the same as used // by IndexUtil::MultiToLinearIndex(tile_assignment_shape). repeated int64 tile_assignment_devices = 4; - // If type == TUPLE, the sub-shardings, one per leaf node in the tuple shape, - // in pre-order. The tuple shape could be nested; here we store just a - // flattened list of all leaves in the tuple shape. Note that the tuple shape - // is not stored here; shardings do not store the shapes to which they are - // applied, this is inferred from the instruction this sharding gets attached - // to. - repeated OpSharding tuple_shardings = 5; } message OpRequest { |