Robert Sachunsky

> Here, we use `Resolver.workspace_from_url` without `download`, which copies the `mets.xml` and nothing else. Like I ~~already~~ (later) said in #1149, cloning from local workspaces is still fundamentally broken. >...

> One thing I noticed is that pyenv and micromamba interfere with each other, i.e. one has to make sure how to avoid (i.e. document to use just one of)...

Unfortunately, in contrast to [when I first came up with the deps-cuda solution](https://github.com/OCR-D/core/pull/1041#issuecomment-1573439760), the sweet spot where we could easily combine CUDA dependencies for both Torch and TF seems to...

Update: we definitely need to hold at `nvidia-tensorflow==2.15.5+nv22.11`, because all later versions depend on CUDA>11.8, but 11.8 is the **last version** which is supported by Tensorflow 2.x on Python 3.8,...

I think I did find a workable compromise again. Let's see what the CD says. Looking at this horde of CI jobs: shouldn't we rule out these tests if the...

I think we really need this. Even if we don't use the variant images (`core-cuda-tf1`, `core-cuda-tf2`, `core-cuda-torch`) yet: we need the new CUDA compromise (`core-cuda`) and we can reuse the...

I resolved the conflict, but the CI failure seems unrelated – we probably just need to update test assets...

> we probably just need to update test assets... Guessed right.

You mean as in ```XML ocrd-tesserocr-recognize v0.17.0 (tesseract 5.3.1-25-gcf23) OCR-D-BIN OCR-D-BIN-OCR-TESS-frak2021 {"model": "frak2021", "dpi": 0, "padding": 0, "segmentat ion_level": "word", "textequiv_level": "word", "overwrite_segments": false, "overwrite_text": true, "shrink_polygons": false, " block_polygons":...

> A simple solution would be to assume the checksum is related to the raw data that `ocrd resmgr` retrieves, i.e. the models or zipped models as they are downloaded...