[Apache TVM Discuss] [Questions] LSTM mapped to specific op in custom target

2021-11-03 Thread Sameekbanerjee via Apache TVM Discuss
Is there a way to mark a layer as an LSTM layer after conversion to relay? I see almost RNN ops as broken into individual cells for each operator. I have a custom accelerator with First class support for LSTM at the operator level. Is there an easy way to do this ? Thanks in advance. --

[Apache TVM Discuss] [Questions] Heterogeneous execution on TVM with custom accelerator

2021-03-29 Thread Sameekbanerjee via Apache TVM Discuss
Thanks @max1996 ! Somehow missed/overlooked this. Looks very promising. --- [Visit Topic](https://discuss.tvm.apache.org/t/heterogeneous-execution-on-tvm-with-custom-accelerator/9557/3) to respond. You are receiving this because you enabled mailing list mode. To unsubscribe from these e

[Apache TVM Discuss] [Questions] Heterogeneous execution on TVM with custom accelerator

2021-03-28 Thread Sameekbanerjee via Apache TVM Discuss
I am wondering if I am using a custom accelerator, Can I skip code generation for the sub-graph that have first class support in the accelerator. The accelerator comes with its own SW stack and has its own proprietary code generation which can't be exposed to TVM. However some operator support