Jiyuan Qian

Support string to IValue dict in stringMapFrom for pytorch

7

## Description I’m planning to use djl mainly for serving pytorch models given it seems to be the most performant option available apart from implementing a service in c++. The...

Support torchtext natively in DJL pytorch engine

1

## Description Currently pytorch models using torchtext modules cannot be loaded with DJL pytorch engine I tried loading the `_torchtext.so` from python site packages in DJL with "PYTORCH_EXTRA_LIBRARY_PATH", but it...

Support easy ways to collapse/hide nested run params

4

## 🚀 Feature Support easy ways to collapse/hide nested run params ### Motivation Deep nested run params can occupy a large number of columns, it would be nice to have...

Error when returning WriteInstructionsResponse with exception from remote server

3

## 🐛 Bug ``` Server is mounted on 0.0.0.0:9999 ERROR:grpc._server:Exception calling application: 0.1 has type float, but expected one of: bytes, unicode Traceback (most recent call last): File "/usr/local/lib/python3.8/dist-packages/aim/ext/transport/server.py", line...

## 🚀 Feature Implement the standard grpc health checkpoint protocol: https://github.com/grpc/grpc/blob/master/doc/health-checking.md ### Motivation Currently when we deploy remote tracker on k8s, there's no way to do meaningful readiness and liveness...

Re-initialize resources when remote tracker returns UnauthorizedRequestError

2

## 🚀 Feature When remote tracker is restarted/redeployed, all its states are gone, and we get UnauthorizedRequestError on the client side if there's an ongoing training run. The client should...

How to scale aim repo to support simultaneous writes from multiple training runs

11

## ❓Question in order to compare multiple training runs side by side, I tried make them write to the same aim repo located on a shared EFS. However when I...

Metrics graphs should be refreshed in UI while a run is in progress

5

## 🚀 Feature When I have an experiment run ongoing, and I click into that run, I can see terminals logs getting refreshed, but not metrics graphs. The only way...

torch.jit.load supports reading an extra files map, see: https://pytorch.org/docs/stable/generated/torch.jit.load.html It's also supported in the C++ api: https://pytorch.org/cppdocs/api/function_namespacetorch_1_1jit_1a05bc7a80919b589891ba0f36ee7be8e0.html#exhale-function-namespacetorch-1-1jit-1a05bc7a80919b589891ba0f36ee7be8e0 Is it possible to add the support in tch-rs?

I have a jit-scripted torch module that takes something like a `List` as input in its `forward`. When I try to construct the input in tch-rs and pass it to...