nolanliou opened a new pull request #6546: URL: https://github.com/apache/incubator-tvm/pull/6546
Problems: 1. Model inputs' types always are default types when converting TorchScript(jit) file due to no type information in TorchScript file. related issue: https://github.com/apache/incubator-tvm/issues/6268 2. When converting models with float16 weights, there is some situation to convert from float32 to float16. Solutions: 1. Add input types argument in pytorch frontend API, instead of always using the default type. 2. Support casting float32 to float16 when converting `aten::to` operation. @t-vi @masahi Could you help to review? Thanks a lot. ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected]
