[Apache TVM Discuss] [Questions] Does Ansor support TensorCore INT8?

[Chenfan via Apache TVM Discuss <no...@discuss.tvm.ai>]

Sorry for that. Though we have a strong will to support TensorCore in Ansor, currently I don't have extra bandwidth to work on this topic.

As far as I know, some guys are working on the new TensorIR, and based on which, TVM will get a new infrastructure to combine the current AutoTVM and AutoScheduler together. The auto schedule search with TensorCore support will be fully supported then.

p.s. The [repo](https://github.com/jcf94/tvm/tree/tensorcore_sketch_support/python/tvm/auto_scheduler/test_sketch) you got is a good example to write extra sketch rules. Have fun with these codes.





---
[Visit Topic](https://discuss.tvm.apache.org/t/does-ansor-support-tensorcore-int8/9941/2) to respond.

You are receiving this because you enabled mailing list mode.

To unsubscribe from these emails, [click here](https://discuss.tvm.apache.org/email/unsubscribe/999d33ab6559c6b28569366f9429d4d405f1a0c3e33a130620efa7bad7459dd8).

[Apache TVM Discuss] [Questions] Does Ansor support TensorCore INT8?

[Subject_No_i via Apache TVM Discuss <no...@discuss.tvm.ai>]

Hello, I check this repo and I'm testing this feature, I wonder how to port this feature into the official Auto Scheduler tutorial code? I set the target to `cuda -arch=sm75` and found that none of the generated sketches contain wmma instructions. 
But there exists several samples that trigger CUDA_ERROR_MISALIGNED_ADDRESS, I wonder is it triggered by wmma instructions?





---
[Visit Topic](https://discuss.tvm.apache.org/t/does-ansor-support-tensorcore-int8/9941/3) to respond.

You are receiving this because you enabled mailing list mode.

To unsubscribe from these emails, [click here](https://discuss.tvm.apache.org/email/unsubscribe/68b8b6e6cca8a1b3e12f4a9941a03a4d2b4532ff5d4adf957462ec88443e5f17).