Hi,
I’m running a model with TVM on GPU. It is currently failing with “unsupported op: one_hot”. The frontend is TensorFlow. The missing op is: tf.one_hot (https://www.tensorflow.org/api_docs/python/tf/one_hot). One_hot does not seem to exist in CUDA library (cudnn, cublas, etc), so what would be an easy way to add support to this op? Thank you.