nv-samcheng

@NVIDIA

Results 3 comments of


                                            nv-samcheng

Performance Discrepancy between Quantized ONNX Model and FP16 Model

Will fix in TRT 10.0 @zerollzeng Please help to close this bug.

Fix: Handle past_key_values AttributeError in generate Function

I don't think which needed to fix since kvcache doens't be used in JanusFlow

JanusFlow示例代码修复建议

Just want to confirm if the kv cache bu used in Janus Flow?