Search Options

Results per page
Sort
Preferred Languages
Advance

Results 1 - 6 of 6 for num_partitions (0.71 sec)

  1. tensorflow/compiler/jit/pjrt_device_compiler_client.cc

      VLOG(2) << "Compiled PJRT executable " << executable->name()
              << " num_replicas " << executable->num_replicas()
              << " num_partitions " << executable->num_partitions();
    
      return std::move(executable);
    }
    
    absl::StatusOr<std::string> PjRtDeviceCompilerClient::SerializeExecutable(
        const xla::PjRtLoadedExecutable& executable) {
    Registered: Sun Jun 16 05:45:23 UTC 2024
    - Last Modified: Wed Feb 21 12:19:41 UTC 2024
    - 3.6K bytes
    - Viewed (0)
  2. tensorflow/compiler/mlir/tensorflow/transforms/xla_call_module_serialization.cc

      // Restore the deserialized stablehlo module's attributes to the reconstructed
      // stablehlo module. The stablehlo module's attributes can contain important
      // information such as SPMD num_replicas and num_partitions.
      auto original_stablehlo_module_attrs =
          op->getAttrOfType<DictionaryAttr>(kStablehloModuleAttrsAttrName);
      if (original_stablehlo_module_attrs) {
    Registered: Sun Jun 16 05:45:23 UTC 2024
    - Last Modified: Thu Apr 25 16:01:03 UTC 2024
    - 9.5K bytes
    - Viewed (0)
  3. tensorflow/compiler/jit/xla_launch_util.cc

      std::vector<std::unique_ptr<xla::PjRtBuffer>> execute_outputs;
      std::optional<xla::PjRtFuture<>> future;
      if (executable->num_replicas() != 1 || executable->num_partitions() != 1) {
        TF_ASSIGN_OR_RETURN(
            execute_outputs,
            executable->ExecuteSharded(
                executable_args, device,
                GetPjRtExecuteOptions(device_type,
    Registered: Sun Jun 16 05:45:23 UTC 2024
    - Last Modified: Thu May 16 00:36:08 UTC 2024
    - 40.4K bytes
    - Viewed (0)
  4. tensorflow/compiler/mlir/tensorflow/ir/tf_ops_n_z.cc

    LogicalResult TPUPartitionedInputV2Op::verify() {
      TPUPartitionedInputV2Op op = *this;
    
      int num_partitions = 1;
      const mlir::ArrayAttr partition_dims = op.getPartitionDims();
      for (const mlir::Attribute &dim : partition_dims) {
        num_partitions *= dim.cast<IntegerAttr>().getInt();
      }
    
      const bool is_packed = op.getIsPacked();
      const bool replicated = partition_dims.empty();
    Registered: Sun Jun 16 05:45:23 UTC 2024
    - Last Modified: Thu May 09 22:07:10 UTC 2024
    - 170.8K bytes
    - Viewed (0)
  5. tensorflow/compiler/mlir/tensorflow/ir/tf_generated_ops.td

    For example:
    
    ```python
        # Scalar partitions.
        partitions = 1
        num_partitions = 2
        data = [10, 20]
        outputs[0] = []  # Empty with shape [0, 2]
        outputs[1] = [[10, 20]]
    
        # Vector partitions.
        partitions = [0, 0, 1, 1, 0]
        num_partitions = 2
        data = [10, 20, 30, 40, 50]
        outputs[0] = [10, 20, 50]
        outputs[1] = [30, 40]
    ```
    
    Registered: Sun Jun 16 05:45:23 UTC 2024
    - Last Modified: Tue Jun 11 23:24:08 UTC 2024
    - 793K bytes
    - Viewed (0)
  6. tensorflow/compiler/jit/xla_launch_util_test.cc

        xla::ExecuteOptions exe_options;
        exe_options.arguments_are_tupled = false;
        exe_options.untuple_result = true;
    
        // TODO(b/257548614): currently PJRT is compiled as portable (num_replica =
        // 1 and num_partition = 1). Support multiple partitions case.
        return executable->ExecutePortable(executable_args, pjrt_device,
                                           exe_options);
      }
    
    Registered: Sun Jun 16 05:45:23 UTC 2024
    - Last Modified: Wed Feb 21 09:53:30 UTC 2024
    - 28.8K bytes
    - Viewed (0)
Back to top