// Copyright 2024 Google LLC.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
//      http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.

syntax = "proto3";

package google.gdm.gdmscience.alphagenome.v1main;

option go_package = "google.golang.org/genproto/googleapis/gdm/gdmscience/alphagenome/v1main;alphagenome";
option java_multiple_files = true;
option java_outer_classname = "TensorProto";
option java_package = "com.google.gdm.gdmscience.alphagenome.v1main";

// Protocol buffer representing an arbitrarily large multi-dimensional array of
// data, laid out in row-major format. To support tensors larger than the
// maximum protocol buffer message size, the tensor payload may be split into
// multiple chunks.
message Tensor {
  // The shape of the tensor. If empty, the data will be treated as a scalar.
  repeated int32 shape = 1;

  // Data type for the elements in this tensor.
  DataType data_type = 2;

  // The payload for the tensor.
  oneof payload {
    // The raw data for the tensor. If present, the tensor is not split into
    // chunks.
    TensorChunk array = 3;

    // The number of chunks the tensor is split into.
    int64 chunk_count = 4;
  }
}

// A single chunk of a tensor.
message TensorChunk {
  // Flattened tensor data. Data is laid out in row-major order.
  bytes data = 1;

  // How the data is compressed. Compression is applied to each chunk
  // independently.
  CompressionType compression_type = 2;
}

// The data type of the tensor.
enum DataType {
  // Unspecified data type.
  DATA_TYPE_UNSPECIFIED = 0;

  // 16-bit "Brain Floating Point". See
  // https://en.wikipedia.org/wiki/Bfloat16_floating-point_format for more
  // details.
  DATA_TYPE_BFLOAT16 = 1;

  // 16-bit floating point.
  DATA_TYPE_FLOAT16 = 11;

  // 32-bit floating point.
  DATA_TYPE_FLOAT32 = 2;

  // 64-bit floating point.
  DATA_TYPE_FLOAT64 = 3;

  // 8-bit signed integer.
  DATA_TYPE_INT8 = 4;

  // 32-bit signed integer.
  DATA_TYPE_INT32 = 5;

  // 64-bit signed integer.
  DATA_TYPE_INT64 = 6;

  // 8-bit unsigned integer.
  DATA_TYPE_UINT8 = 7;

  // 32-bit unsigned integer.
  DATA_TYPE_UINT32 = 8;

  // 64-bit unsigned integer.
  DATA_TYPE_UINT64 = 9;

  // 8-bit boolean.
  DATA_TYPE_BOOL = 10;
}

// Compression type for the tensor data.
enum CompressionType {
  // No compression.
  COMPRESSION_TYPE_NONE = 0;

  // ZSTD compression.
  COMPRESSION_TYPE_ZSTD = 1;
}