Ehsan Azar

Now no code completion works, and no got-to-definition either.

No idea what changed, but it is working today! I see this in the logs ```text Microsoft.Python.LanguageServer[6433]: JsonRpc Error: 10 : Microsoft.Python.LanguageServer[6433]: Exception thrown from request "null" for method initialized:...

@MikhailArkhipov It could be, ssh over wired/wifi over vpn. It asked to "Reload Window" and activated conda environment *twice*. How can we get useful logs next time this issue happens?...

This happens specially when there is an update, the old LS sticks around.

Just removing `Sequence` will not be enough. I ended up using this to fix the exported model, you may need to change depending on the model you have. ```python onnx_model...

One advantage of TensorRT is that it can be used as the Execution Provider for OnnxRuntime. If TRT could support FasterTransformer automatically ONNX would benefit from the optimizations.

I think this is very related to issue #76

@antinucleon Is there a list to know which kernels are not supported in V100? For example, in stable diffusion what is blocking? We could avoid only those kernels, until they...

When I change that UT it passes: ```python class ConvTestCase(unittest.TestCase): def test_fp16(self, batch=1): target = detect_target() X = Tensor( shape=[1, 384, 384, 4], dtype="float16", name="input_0", is_input=True, ) W = Tensor(...

The visualization is pretty simple