auto operands = adaptor.getOperands();
  auto lookup_attr = operands[0].dyn_cast_or_null<DenseIntElementsAttr>();
  auto value_attr = operands[1].dyn_cast_or_null<DenseElementsAttr>();
  if (!lookup_attr || !value_attr) {
    return {};
  }

  // Hybrid lookup is not supported.
  if (value_attr.getType().getElementType() != getType().getElementType()) {
    return {};
  }

      params.push_back(idx);
      continue;
    }

    TF_ASSIGN_OR_RETURN(auto value_attr,
                        ConvertTensor(xla_arg.constant_value, &builder));
    // TODO(hinsu): Use the actual location of the constant.
    auto constant = builder.create<mlir::TF::ConstOp>(
        mlir::UnknownLoc::get(module_op.getContext()), value_attr);
    mlir_arg.replaceAllUsesWith(constant);
    args_to_erase.push_back(idx);
  }

                             : nullptr;
    se::Stream* const dst_device_to_device_stream =
        should_use_substream
            ? device_to_device_master_stream->GetOrCreateSubStream().value_or(
                  nullptr)
            : dst_xla_context->GetDeviceToDeviceStream();
    TF_RET_CHECK(dst_device_to_device_stream != nullptr);
    auto return_substream = gtl::MakeCleanup(

                 Const(scope, *end - start, TensorShape({1})));
  } else {
    return Slice(scope, tensor, Const(scope, start, TensorShape({1})),
                 Add(scope, Shape(scope, tensor), end.value_or(0) - start));
  }
}

// Returns reduced subscripts and their corresponding dimensions and axes.
//
// Given a set of axis labels, returns their concatenated subscript, their

        op.getLoc(), DenseElementsAttr::get(k_ty, k));
    // Compute new result types.
    auto values_ty = mlir::dyn_cast<ShapedType>(values.getType());
    auto indices_ty = mlir::dyn_cast<ShapedType>(indices.getType());
    auto shape = std::vector<int64_t>();
    for (auto d : values_ty.getShape().drop_back()) {
      shape.push_back(d);
    }
    shape.push_back(static_cast<int64_t>(k));

  RankedTensorType value_ty =
      mlir::dyn_cast<RankedTensorType>(op.getValue().getType());
  RankedTensorType bias_ty =
      mlir::dyn_cast<RankedTensorType>(op.getBias().getType());
  if (!bias_ty || !value_ty) return success();

  int64_t feature_dim_idx =
      tensorflow::GetTensorFeatureDimIndex(value_ty.getRank(), format);
  int64_t feature_dim = value_ty.getDimSize(feature_dim_idx);

            std::make_unique<xla::PjRtStreamExecutorBuffer>(
                device_shape, std::move(device_buffer), pjrt_client,
                pjrt_device,
                pjrt_device->default_memory_space().value_or(nullptr));
        owned_args->push_back(std::move(pjrt_buffer));
        args->push_back(owned_args->back().get());
      }
    } else {
      if (av_tensor->GetBuffer() == nullptr) {

if(event.name){rows.push({name:'Name',value:event.name});}
rows.push({name:'Start',value:tr.v.ui.createScalarSpan(event.start,{unit:tr.b.Unit.byName.timeStampInMs})});if(event.duration){rows.push({name:'Wall Duration',value:tr.v.ui.createScalarSpan(event.duration,{unit:tr.b.Unit.byName.timeDurationInMs})});}
if(event.cpuDuration){rows.push({name:'CPU Duration',value:tr.v.ui.createScalarSpan(event.cpuDuration,{unit:tr.b.Unit.byName.timeDurationInMs})});}

var(--tw-ring-offset-color)","--tw-ring-shadow":`var(--tw-ring-inset) 0 0 0 calc(${s} + var(--tw-ring-offset-width)) var(--tw-ring-color)`,"box-shadow":["var(--tw-ring-offset-shadow)","var(--tw-ring-shadow)","var(--tw-shadow, 0 0 #0000)"].join(", ")})},{values:r("ringWidth"),type:"length"}),t({".ring-inset":{"@defaults ring-width":{},"--tw-ring-inset":"inset"}})},ringColor:({matchUtilities:i,theme:e,corePlugins:t})=>{i({ring:r=>t("ringOpacity")?se({color:r,property:"--tw-ring-color",variable:"--tw-ring-...