Closed BalyshevArtem closed 1 month ago
This pr adds training for FullyConnected kernel: compute gradients for weights (+ bias) and for input tensor and for relu.
for issue https://github.com/Samsung/ONE/issues/12873 from draft: https://github.com/Samsung/ONE/pull/13107
ONE-DCO-1.0-Signed-off-by: Artem Balyshev a.balyshev@samsung.com
This pr adds training for FullyConnected kernel: compute gradients for weights (+ bias) and for input tensor and for relu.
for issue https://github.com/Samsung/ONE/issues/12873 from draft: https://github.com/Samsung/ONE/pull/13107
ONE-DCO-1.0-Signed-off-by: Artem Balyshev a.balyshev@samsung.com