jade_zhao

I encountered same error as you,and i assumed the err error is generated from rnnt_loss, i have try some ways ,but it didn't work,anyone has fixed it?

I can train the model use multi-gpus by add a decorator @tf.function refer this link https://github.com/tensorflow/tensorflow/issues/29911,and i also add line of “os.environ['CUDA_VISIBLE_DEVICES'] = "{your gpus}” in my code.

It's the multi-gpu training code what i modified,but the loss value from negative to nan after trained some batches. 

You can try modify the in_channels of SECOND backbone from 256 to 128.

I did not familiar with quantization,if anyone can share a tutorial about that how to compress the float32 saved model to int8 and can really work in cpu env with...

I have test the difference between front_x and cross_x that concanate with T: sample 1(cross_x is better): origin front:  the res used front_x:  the res used cross_x: ...

哈哈，求更新

My data is also close to KITTI's data format.

My data only has forward radar scan data, similiar with kitti dataset,so there is a big difference between point range and waymo, especially in the dimension of the x-axis, how...

I encountered the same problem, and I installed the latest version of GraphRAG following https://github.com/Cinnamon/kotaemon#setup-graphrag, but the problem still persists. Has anyone solved it yet?