mirekphd
mirekphd
@benoitc any progress on this issue?
I thought only Python 3.10 was supported?:) That's what the docs say anyway [here](https://nvidia.github.io/TensorRT-LLM/installation/linux.html) (I know that this Python version is bundled with the current LTS version of Ubuntu, but...
> Also tried adding this argument to the build command, but it is still failing: > > `--extra-backend-cmake-arg=onnxruntime:CMAKE_POLICY_VERSION_MINIMUM=3.5` That `cmake` flag is effective, but without the `onnxruntime:` prefix AND without...
> But maybe we should change the `name` to `code-oss` or something like that. There are two app names and two versions here, so the full info would be two...
We do have different settings of `max user processes`, but I have a larger value (an order of magnitude in fact), even though I'm using the default unchanged settings from...
Unfortunately, the `--checkers` and resulting parallelism ([in multiple places of rclone](https://forum.rclone.org/t/max-read-ahead-what-does-exactly/889/4). or even [in _lots_ of places)](https://github.com/rclone/rclone/issues/4627#issuecomment-699017739), was the reason why the official (containerized) MinIO server was crashing consistently for...
I'd at least document the role of `--checkers` argument as the pretty universal paralellism switch, that needs to be lowered or turned off in case of slow-reacting backends (especially when...
And one of the reasons why the latest `autoawq==0.2.8` requires `transformers
In theory they should be working, because they are all passed to the `options` arg. of `FasterWhisperPipeline` [here](https://github.com/m-bain/whisperX/blob/main/whisperx/asr.py#L412-L415). You can try combinations that must raise errors to verify. It's another...
I'd argue lack of this feature (such as `/v1/logits` or `/v1/rerank_logits` endpoint) is a major shortcoming,.. while not technically a bug, it is limiting the scope of usefullness of this...