geohotstan

========== 282 failed, 541 passed, 1811 skipped, 2 warnings in 20.25s ========== ========== 265 failed, 558 passed, 1811 skipped, 2 warnings in 19.62s ========== for ArrayFeatureExtractor, I modified the original...

Tensor.ceil() turns 1.0 into 2

2

Needed Tensor.ceil() in some onnx tests but encountered problems. `In [33]: Tensor([1.0, 2.0, 1.1]).ceil().numpy()` `Out[33]: array([2., 3., 2.], dtype=float32)` I'm not sure if this is expected behavior. 😟

[WIP] onnx full passing

14

Failing tests (53): https://gist.github.com/geohotstan/db8d722fe311db2b06709eb2a5276a00 Todo: - get to 0 failing - clean up code (lots of clean up) - write better implementations for ops (Slice, gather, resize, pad, ...) and...

add asyncio.CancelledError to serialize_decorator

4

while investigating issue #1013 I tried `ctrl + c` to start serializing for a deserialize run, but `Team.serialize` was not being triggered. the traceback looks something like ``` ... File...

move process_message inside BaseLLM

5

**Features** - fixes #1016 - I think moving `process_message` into `BaseLLM` is better since `process_message` is only used in the context of LLM use. **Feature Docs** **Influence** **Result** - This...

conceptually simpler fancy index

4

this was pulled from my setitem attempt. basically while doing setitem I found out we could reduce multiple dims at once and that reduces kernel count (lol didn't know that...

fixes #148

Seeing if this format is acceptable If this is mergable, I'll do observation_serialization too :D

[POC] emit indexing code directly

3

There are a bunch of rough edges here I haven't ironed out, but here's a POC. I wrote a bunch of hacks just to get it working. I've tried doing...