lly-zero-one

icon company Facebook icon location Menlo Park icon location work on AI framework, compiler, quantization and performance optimization for inference and distributed training.

Results 2 comments of lly-zero-one

Question on FuseSupportedOps pass

@serhaty I think you saw the same issue and was working on a fix?

[torch_tvm] Support Lowering to TVM even if node cannot be fused

Can you try the flow on the quantized model to see how many subgraph we could get now?