search

combust / mleap

MLeap: Deploy ML Pipelines to Production
https://combust.github.io/mleap-docs/
Apache License 2.0
1.5k stars 310 forks source link

mleap enhancments #814

Closed mtsol closed 2 years ago

mtsol commented 2 years ago
mullerhai commented 1 year ago

when support pytorch model save and load ？

jsleight commented 1 year ago

Supporting pytorch is difficult because pytorch does not have any java bindings (to my knowledge at least). The best option for supporting pytorch is probably to support something like ONNX and then have folks convert their pytorch models into ONNX before serializing.

mullerhai commented 1 year ago

Supporting pytorch is difficult because pytorch does not have any java bindings (to my knowledge at least). The best option for supporting pytorch is probably to support something like ONNX and then have folks convert their pytorch models into ONNX before serializing.

ok, if pytorch has java bindings,do you know how to save pytorch save and load, maybe I could do it ,just tell me the method is ok

jsleight commented 1 year ago

There currently isn't any pytorch support in mleap. You'd need to create:

Basically look at everything which exists for tensorflow and need a new variant of that for pytorch.

mullerhai commented 1 year ago

There currently isn't any pytorch support in mleap. You'd need to create:

  • a new mleap transformer
  • a new mleap op
  • a new spark transformer
  • a new spark op

Basically look at everything which exists for tensorflow and need a new variant of that for pytorch.

thanks, If I success ,please pick up me code