Python open issues, still unclaimed

Description With the PyTorch backend now supporting jit (see issue #4), it is crucial to establish a comprehensive benchmark suite. This suite will be used to evaluate and compare the performance of the PyTorch backend agaisnt TensorFlow, and JAX backends across different…

Description The SklearnSimilarityIndex class in the FiftyOne Brain module applies mean-centering to embeddings in the _build_dists and _build_neighbors methods before computing nearest neighbors. However, there are issues with the current implementation that need to be addressed.…

soft_errors job option is declared as a string parameter: openeo-geopyspark-driver/openeogeotrellis/job_options.py Lines 74 to 77 in 1e8f9ef soft_errors: str = field( default=get_backend_config().default_soft_errors, metadata={ "description": "Ratio of soft-errors to allow in…

Problem KernelCI Debos rootfs images currently use a generic shell prompt: / # Some newer images may instead show: ~ # KernelCI lists these strings as expected login prompts in its LAVA job templates. LAVA warns that these prompts are not distinctive and may accidentally match…

Documentation location https://deepinv.github.io/deepinv/user_guide/reconstruction/pretrained-models.html#pretrained-weights https://deepinv.github.io/deepinv/api/stubs/deepinv.models.DRUNet.html#deepinv.models.DRUNet Describe the issue or suggestion According to this page, there…

The OSS avoid to store redundant object data by computing a 64-bits Murmur hash using std::hash. Looks like 64-bits murmur is mostly OK, although there is a small risk of collision. That's about 39% probability of a collision on 4B items, but only 0.00000271% on 1M items. Not…

It appears that we are adding the same jars twice when passing them to the spark_context for the unit tests. ('spark.jars', '/var/lib/jenkins/workspace/opyspark-driver-python311_master/tests/../jars/geotrellis-extensions-2.5.0_2.12-SNAPSHOT.jar,…

Currently it seems that all autovacuum logs after working the database, table, and duration. It would be great also log how many rows and/or files were compacted/removed and totals left when finished.

Description Describe your idea. Please be detailed. If a feature request, please describe the desired behavior, what scenario it enables, and how it would be used. AWS IAM users can have static credentials associated with them that are tied to a specific AWS service e.g. Bedrock.…

This issue serves to track the runnable examples in the latest release of Ianvs. Ianvs includes a wide range of examples showcasing its benchmarking capabilities. So far, I have verified the following examples as runnable and also updated their quick-start documentation guide for…

Prior to capabilities support, scaler only have a message queue size for each worker, but now since we have capabilities (tag to a numerical capabilities), like {"cpu": 1000, "gpu": 1000}, we should make old queue size default capabilities as {"default": 1000} and remove old…

I fine-tuned the Qwen3-VL-8B-Instruct model using Unsloth. My code is 99% identical to the official guide; the only change I made was replacing the 8B model in the guide with the 2B model for fine-tuning. After fine-tuning, I confirmed that the QLoRA adapter was saved correctly.…

Please note that the reproduceability of this behaviour is not 100%, most likely due to a nuance I'm missing on the steps, therefore it's been difficult to find well defined and concise steps to reproduce. On 16.16 latest release build. The computation's that uses 2 data items is…

Summary We should have a way to ensure whether code in our docs is working and producing the expected output. This could either be done by setting up doctests (with some custom logic for our shell commands etc.), or by referencing tested examples via literalinclude instead of…

In Abakus, you should be able to be engaged on many different fronts. In interest groups, it is usually the leader of the group who takes the initiative and organizes activities. We should reward the leaders for their work by giving them a unique badge on their profile. I think…

VideoDecoder can handle variable FPS videos in some scenarios, but doesn't handle them well in other scenarios. We should have some unit tests which use a variable FPS video. We should create a small variable FPS video and check it in. A command that can create one: ffmpeg -i…

I find myself always needing to make sure the annotations and images line up as expected, its such a common error. We should just auto log five images before training, I think the cost to the user is very low and it sets up well for tensorboard or comet or wandb. I can imagine a…

We should use absl::InlinedVector for vectors that typically have a small number of elements. This avoids heap allocation for the common case while gracefully handling larger collections. Currently, we have 2k+ uses of std::vector but only 4 uses of absl::InlinedVector in…

When the point graphic is used the masks can often have several adjacent pixels included seemingly at random. Looking at the code, this is due to it using an elliptical mask with diameter 1.2 pixels centered on the point. Would a 1 pixel x 1 pixel rectangle mask be a better…

Observe on usim. When right clicking a camera display panel that doesn't have an image, the option doesn't appear. This makes the tool confusing, as if it would only eliminate the contents, then when used on a camera display panel that does have an image, it shouldn't eliminate…

ND: 2-if you're aligning with a lattice then typically one can adjust the location of a far out spot to line up, but trying to align the innermost spot to make the outer patches land in the right location is not possible with the mouse UI coarseness. If one could adjust the mouse…

Right now we have a guest price in backend for the event model, and after requesting you get either the member price or guest price (defined by if you are a member or not). If you are not a member the price will therefor default to 0 because you can't set the guest price in…

Currently operations on (reads) a Databricks Unity Catalog on GCP doesn't work with the Catalog abstraction for Delta tables. The reason for this is because internally we use the delta-rs library to write to delta tables, and delta-rs relies on an older version of object_store…