Can I implement block quantization through tensorflow-quantization?

[lqq-feel]

I want to implement block quantization through tensorflow-quantization, what process should I need to follow? Or can you support a simple case?

[lix19937]

Maybe it helps, https://github.com/NVIDIA/TensorRT/tree/release/10.7/tools/tensorflow-quantization

[lqq-feel]

Maybe it helps, https://github.com/NVIDIA/TensorRT/tree/release/10.7/tools/tensorflow-quantization
I have read through it before, but not found a example to implement block quantization. So I do not know that does it really support block quantization?

[lix19937]

trt-llm has do some work on lora, which include Block-wise Quantization.

[lqq-feel]

trt-llm has do some work on lora, which include Block-wise Quantization.
Can trt-llm do block-wise quantization on non-LLM, a graphdef model of tensorflow？