Skip to content
Problems.

ML experiments are hard to reproduce and track

MLOps4 posts from 4 people4d active+11 post in the last 7 days, 1 the 7 days before (steady)

Posts per day

Posts per day4 posts, Sep 2 to Sep 25

The posts behind it

4, newest first
PostDate
[Bug]: stream_structured_predict / astream_structured_predict raise UnboundLocalError instead of yielding nothing when the partial stream is empty### Bug Description `LLM.stream_structured_predict` (llama-index-core/llama_index/core/llms/llm.py:461) and its async twin `LLM.astream_structured_predict` (llm.py:539) emit their terminal instrumentation event with the loop variable, *after* the loop that binds it: The async...run-llama/llama_indexBlueX888Sep 196 days ago
[BUG] AI Gateway: `send_stream_request` drops the upstream error message unless the body is OpenAI-shaped> [!WARNING] > Before submitting a PR, please make sure that: > - A maintainer has triaged this issue and applied the `ready` label > - This issue has no assignee > - No duplicate PR exists > > PRs not meeting these requirements may be automatically closed. ### Issues Policy...mlflow/mlflowrhamzatovSep 169 days ago
[FR] Link model versions to their source logged model in the Model Registry UI...annoyance, and versions without a run are a dead end. > #### Why is it currently difficult to achieve this use case? The model version REST response only carries the logged model ID inside the `source` string (`ModelVersion.to_proto` does not populate `model_id` yet), and...mlflow/mlflowrriohSep 92 weeks ago
[FR] Expose Helm health probes as configuration variables...and can cause false probe failures during startup or load. > #### Why is it currently difficult to achieve this use case? > The chart template currently hard-codes liveness and readiness probe timings and does not render a startup probe. Supplying similarly named keys through...mlflow/mlflowjeremyederSep 23 weeks ago

Companies and products named

Company or productPosts naming it
MLflow3
GitHub2
LlamaIndex1
OpenAI1
Claude1
Amazon Bedrock1
About this problem

Evidence

4 posts from 4 people in 2 places, about 1 a week over 18 days. Mostly on mlflow/mlflow, run-llama/llama_index. Tools named alongside: MLflow, GitHub, LlamaIndex, OpenAI.

Frustration Frustration 0 of 3· Seen on GitHub issues

How it was grouped

Posts that state a pain and match the "ml-reproducibility" rule. First post Sep 2, 2026, latest Sep 19, 2026. Corroborated: 3 or more posts from 2 or more people or places. Method

History

  • 2026-09-25 Corroboration: unverified to corroborated
  • 2026-09-25 Momentum: new to steady
  • 2026-09-25 Added: ML experiments are hard to reproduce and track (1 post)

Rising problems by email

Mondays: the problems in data, tech and AI that grew fastest that week.

Double opt-in. Unsubscribe any time.