id int64 959M 2.55B | title stringlengths 3 133 | body stringlengths 1 65.5k ⌀ | description stringlengths 5 65.6k | state stringclasses 2
values | created_at stringlengths 20 20 | updated_at stringlengths 20 20 | closed_at stringlengths 20 20 ⌀ | user stringclasses 174
values |
|---|---|---|---|---|---|---|---|---|
2,180,356,543 | check openapi spec against spectral | also: fix action name
also: update openapi.json detail just to trigger the action | check openapi spec against spectral: also: fix action name
also: update openapi.json detail just to trigger the action | closed | 2024-03-11T22:15:53Z | 2024-03-11T22:29:05Z | 2024-03-11T22:29:05Z | severo |
2,179,805,298 | remove call to the Hub when processing a webhook | Since https://github.com/huggingface/moon-landing/pull/9061 (internal), the webhooks now contain the list of updated refs, which should make the following call redundant:
https://github.com/huggingface/datasets-server/blob/60bcc28d3bcbc27ced08c29966aa332af666d02c/services/api/src/api/routes/webhook.py#L84-L88
We ... | remove call to the Hub when processing a webhook: Since https://github.com/huggingface/moon-landing/pull/9061 (internal), the webhooks now contain the list of updated refs, which should make the following call redundant:
https://github.com/huggingface/datasets-server/blob/60bcc28d3bcbc27ced08c29966aa332af666d02c/ser... | open | 2024-03-11T17:51:17Z | 2024-03-28T13:37:58Z | null | severo |
2,176,481,726 | remove temporary EFS | fixes #2568
- [x] descriptive statistics
- [x] HF datasets
- [x] duckdb index (only for workers)
| remove temporary EFS: fixes #2568
- [x] descriptive statistics
- [x] HF datasets
- [x] duckdb index (only for workers)
| closed | 2024-03-08T17:34:18Z | 2024-03-11T14:04:14Z | 2024-03-11T14:04:04Z | severo |
2,176,445,140 | Use local storage (NVMe) instead of EFS volumes | The worker machines have NVMe, and we should use this local storage instead of EFS for:
- stats cache
- HF datasets cache
- duckdb index cache
We still need to use EFS for parquet metadata since it's shared data, and for duckdb local cache in `services/search`.
Note that `services/admin` and `clean-...` cron ... | Use local storage (NVMe) instead of EFS volumes: The worker machines have NVMe, and we should use this local storage instead of EFS for:
- stats cache
- HF datasets cache
- duckdb index cache
We still need to use EFS for parquet metadata since it's shared data, and for duckdb local cache in `services/search`.
... | closed | 2024-03-08T17:11:32Z | 2024-03-11T14:04:05Z | 2024-03-11T14:04:05Z | severo |
2,176,368,381 | tweak autoscale parameters (again) | I increased the minimum number of workers to what we had nominally before autoscaling. | tweak autoscale parameters (again): I increased the minimum number of workers to what we had nominally before autoscaling. | closed | 2024-03-08T16:27:57Z | 2024-03-08T16:28:32Z | 2024-03-08T16:28:13Z | severo |
2,176,313,155 | increase threshold for heavy workers to 50 | temporary fix for https://github.com/huggingface/datasets-server/issues/889#issuecomment-1985943755
| increase threshold for heavy workers to 50: temporary fix for https://github.com/huggingface/datasets-server/issues/889#issuecomment-1985943755
| closed | 2024-03-08T15:57:42Z | 2024-03-08T15:58:14Z | 2024-03-08T15:57:49Z | severo |
2,176,258,357 | Regularly review logs starting with `ERROR: ` and open issues for them | See https://kibana.elastic.huggingface.tech/app/discover#/?_g=(filters:!(),refreshInterval:(pause:!t,value:0),time:(from:now-25h,to:now))&_a=(columns:!(message,kubernetes.namespace),filters:!(('$state':(store:appState),meta:(alias:!n,disabled:!f,index:de38ff80-ac19-11ec-bb45-ad141ad1c5f8,key:kubernetes.namespace,negate... | Regularly review logs starting with `ERROR: ` and open issues for them: See https://kibana.elastic.huggingface.tech/app/discover#/?_g=(filters:!(),refreshInterval:(pause:!t,value:0),time:(from:now-25h,to:now))&_a=(columns:!(message,kubernetes.namespace),filters:!(('$state':(store:appState),meta:(alias:!n,disabled:!f,in... | open | 2024-03-08T15:27:26Z | 2024-08-22T14:28:07Z | null | severo |
2,176,106,959 | Fix memory leak in `config-parquet-and-info`: do not store pq.ParquetFIle objects in memory | should fix [this](https://github.com/huggingface/datasets-server/issues/2552)
\+ i added test for `fill_builder_info` with multiple parquet files (seems that all other tests for this worker use only 1) | Fix memory leak in `config-parquet-and-info`: do not store pq.ParquetFIle objects in memory: should fix [this](https://github.com/huggingface/datasets-server/issues/2552)
\+ i added test for `fill_builder_info` with multiple parquet files (seems that all other tests for this worker use only 1) | closed | 2024-03-08T14:03:39Z | 2024-03-12T09:55:20Z | 2024-03-12T09:55:19Z | polinaeterna |
2,176,098,851 | -- incorrect pull request ignore this | lol incorrect branch | -- incorrect pull request ignore this: lol incorrect branch | closed | 2024-03-08T13:59:55Z | 2024-03-08T14:04:47Z | 2024-03-08T14:00:30Z | polinaeterna |
2,175,870,914 | Use `revision_exists` (hfh) | Here (and maybe elsewhere):
https://github.com/huggingface/datasets-server/blob/b40606d399731137846f3d67ec3bda964cd25946/services/worker/src/worker/utils.py#L202
Use the new `revision_exists` method (see https://github.com/huggingface/huggingface_hub/releases/tag/v0.21.0) | Use `revision_exists` (hfh): Here (and maybe elsewhere):
https://github.com/huggingface/datasets-server/blob/b40606d399731137846f3d67ec3bda964cd25946/services/worker/src/worker/utils.py#L202
Use the new `revision_exists` method (see https://github.com/huggingface/huggingface_hub/releases/tag/v0.21.0) | open | 2024-03-08T11:38:42Z | 2024-03-08T11:39:03Z | null | severo |
2,175,808,766 | Make CI faster? | See https://github.com/huggingface/huggingface_hub/issues/2062 for reference | Make CI faster?: See https://github.com/huggingface/huggingface_hub/issues/2062 for reference | open | 2024-03-08T11:04:30Z | 2024-03-08T11:04:48Z | null | severo |
2,175,633,391 | tweaking the threshold for autoscaling | 30 jobs is too low for the medium jobs (and for the light ones too). On the other side, I think we can lower the limit for the heavy ones to 20.
<img width="407" alt="Capture d’écran 2024-03-08 à 10 26 57" src="https://github.com/huggingface/datasets-server/assets/1676121/213e1943-3df5-473a-a481-da46396cf3ce">
| tweaking the threshold for autoscaling: 30 jobs is too low for the medium jobs (and for the light ones too). On the other side, I think we can lower the limit for the heavy ones to 20.
<img width="407" alt="Capture d’écran 2024-03-08 à 10 26 57" src="https://github.com/huggingface/datasets-server/assets/1676121/21... | closed | 2024-03-08T09:27:41Z | 2024-03-08T14:53:57Z | 2024-03-08T14:53:56Z | severo |
2,175,168,890 | Rollback delete duckdb indexes job | Once datasets-server-job-delete-duckdb-indexes job finishes in prod, we can merge this PR in order to avoid analyzing again the datasets to delete old duckdb indexes from refs/convert/parquet | Rollback delete duckdb indexes job: Once datasets-server-job-delete-duckdb-indexes job finishes in prod, we can merge this PR in order to avoid analyzing again the datasets to delete old duckdb indexes from refs/convert/parquet | closed | 2024-03-08T02:42:07Z | 2024-03-08T12:00:55Z | 2024-03-08T12:00:54Z | AndreaFrancis |
2,174,627,617 | Fix: delete duckb indexes failed for datasets without refs/convert/parquet revision | null | Fix: delete duckb indexes failed for datasets without refs/convert/parquet revision: | closed | 2024-03-07T19:34:07Z | 2024-03-07T19:36:23Z | 2024-03-07T19:36:22Z | AndreaFrancis |
2,174,454,038 | Add `/compatible-libraries` route | - Renamed `dataset-loading-tags` -> `dataset-compatible-libraries` in code
- Added migration for queue and cache
- Added route | Add `/compatible-libraries` route: - Renamed `dataset-loading-tags` -> `dataset-compatible-libraries` in code
- Added migration for queue and cache
- Added route | closed | 2024-03-07T18:08:10Z | 2024-03-14T16:22:01Z | 2024-03-14T16:22:01Z | lhoestq |
2,174,354,508 | let's use more workers when needed | null | let's use more workers when needed: | closed | 2024-03-07T17:15:01Z | 2024-03-07T18:04:32Z | 2024-03-07T18:04:32Z | severo |
2,174,324,200 | Sort the rows along a column | See https://github.com/huggingface/moon-landing/issues/9201 (internal)
| Sort the rows along a column: See https://github.com/huggingface/moon-landing/issues/9201 (internal)
| closed | 2024-03-07T16:57:43Z | 2024-05-03T11:14:07Z | 2024-05-03T11:14:07Z | severo |
2,174,065,256 | Default config first | Show the default config first in the Viewer.
I did this by ordering the configs in `dataset-config-names` to have the default config in first position
fix https://github.com/huggingface/datasets-server/issues/2541
No need to recompute all the entries, it's fine to apply this to new / updated datasets imo
cc... | Default config first: Show the default config first in the Viewer.
I did this by ordering the configs in `dataset-config-names` to have the default config in first position
fix https://github.com/huggingface/datasets-server/issues/2541
No need to recompute all the entries, it's fine to apply this to new / upda... | closed | 2024-03-07T14:59:53Z | 2024-03-07T17:09:54Z | 2024-03-07T17:09:53Z | lhoestq |
2,173,843,189 | Determine if a string column is string or category by proportion of unique values | ...instead of by an absolute value. Will solve https://github.com/huggingface/datasets-server/issues/1953
Set to 0.2.
UPD: i also added the maximum value of categories to 1000 (for example, there are ~700 languages in the-stack-v2). Yes, it makes no sense for the viewer to display 1k labels but i think it should be... | Determine if a string column is string or category by proportion of unique values: ...instead of by an absolute value. Will solve https://github.com/huggingface/datasets-server/issues/1953
Set to 0.2.
UPD: i also added the maximum value of categories to 1000 (for example, there are ~700 languages in the-stack-v2). ... | closed | 2024-03-07T13:16:45Z | 2024-03-12T17:17:42Z | 2024-03-12T17:17:41Z | polinaeterna |
2,173,527,327 | `config-parquet-and-info` crashes with OOM at pq files validation, presumably when there are a lot of files | [bigcode/the-stack-v2](https://huggingface.co/datasets/bigcode/the-stack-v2) keeps failing at config-parquet-and-info step and seems that the problem is that there are too many parquet files to validate (there are about 1k of them).
It [starts](https://github.com/huggingface/datasets-server/blob/main/services/worke... | `config-parquet-and-info` crashes with OOM at pq files validation, presumably when there are a lot of files: [bigcode/the-stack-v2](https://huggingface.co/datasets/bigcode/the-stack-v2) keeps failing at config-parquet-and-info step and seems that the problem is that there are too many parquet files to validate (there ... | closed | 2024-03-07T10:30:16Z | 2024-03-12T18:53:10Z | 2024-03-12T09:55:20Z | polinaeterna |
2,171,418,131 | Update openapi specs according to changes in stats #2453 | Some values of numerical results might be null now (rare cases when all values in column are null)
See https://github.com/huggingface/datasets-server/pull/2453 | Update openapi specs according to changes in stats #2453: Some values of numerical results might be null now (rare cases when all values in column are null)
See https://github.com/huggingface/datasets-server/pull/2453 | closed | 2024-03-06T12:41:53Z | 2024-03-08T14:09:29Z | 2024-03-06T15:55:52Z | polinaeterna |
2,170,240,553 | More precise dataset size computation | Currently, the Hub uses the `/size` endpoint's `num_bytes_original_files` value to display the `Size of downloaded dataset files` on a dataset's card page. However, this value does not consider a possible overlap between the configs' data files (and simply [sums](https://github.com/huggingface/datasets-server/blob/e4aa... | More precise dataset size computation: Currently, the Hub uses the `/size` endpoint's `num_bytes_original_files` value to display the `Size of downloaded dataset files` on a dataset's card page. However, this value does not consider a possible overlap between the configs' data files (and simply [sums](https://github.co... | open | 2024-03-05T22:22:24Z | 2024-05-24T20:59:36Z | null | mariosasko |
2,168,915,818 | Fix: Add datasets cache to loading tags job runner | `dataset-hub-cache` is failing because of `dataset-loading-tags` - PermissionError - [Errno 13] Permission denied: '/.cache'
Currently 754 affected datasets.
| Fix: Add datasets cache to loading tags job runner: `dataset-hub-cache` is failing because of `dataset-loading-tags` - PermissionError - [Errno 13] Permission denied: '/.cache'
Currently 754 affected datasets.
| closed | 2024-03-05T11:04:08Z | 2024-03-05T11:17:16Z | 2024-03-05T11:17:15Z | AndreaFrancis |
2,167,946,331 | Don't truncate image bytes | to fix https://huggingface.co/datasets/Major-TOM/Core-S2L2A RowsPostProcessingError | Don't truncate image bytes: to fix https://huggingface.co/datasets/Major-TOM/Core-S2L2A RowsPostProcessingError | closed | 2024-03-04T23:05:06Z | 2024-03-04T23:05:48Z | 2024-03-04T23:05:48Z | lhoestq |
2,167,925,930 | hot fix for images | fix https://github.com/huggingface/datasets-server/pull/2545 | hot fix for images: fix https://github.com/huggingface/datasets-server/pull/2545 | closed | 2024-03-04T22:47:44Z | 2024-03-04T22:47:55Z | 2024-03-04T22:47:55Z | lhoestq |
2,167,880,172 | Revert max_workers for pq validation | revert https://github.com/huggingface/datasets-server/pull/2544 since it made it too slow to process https://huggingface.co/datasets/Major-TOM/Core-S2L2A | Revert max_workers for pq validation: revert https://github.com/huggingface/datasets-server/pull/2544 since it made it too slow to process https://huggingface.co/datasets/Major-TOM/Core-S2L2A | closed | 2024-03-04T22:14:17Z | 2024-03-05T08:58:11Z | 2024-03-04T22:15:35Z | lhoestq |
2,167,762,617 | Support binary columns for images | ... as `datasets` does
fix for https://huggingface.co/datasets/Major-TOM/Core-S2L1C | Support binary columns for images: ... as `datasets` does
fix for https://huggingface.co/datasets/Major-TOM/Core-S2L1C | closed | 2024-03-04T21:11:02Z | 2024-03-04T21:17:11Z | 2024-03-04T21:16:57Z | lhoestq |
2,167,315,911 | Set max workers for `retry_and_validate_get_parquet_file_and_size` in `config-parquet-and-info` to 4 | I assume the fact that we don't restrict the max number of workers (which results in [32 workers](https://github.com/tqdm/tqdm/blob/master/tqdm/contrib/concurrent.py#L44)* from `tqdm.contrib.concurrent.thread_map` which we use at this step) might result in (likely) OOM for `bigcode/the-stack-v2`. Since there are about ... | Set max workers for `retry_and_validate_get_parquet_file_and_size` in `config-parquet-and-info` to 4: I assume the fact that we don't restrict the max number of workers (which results in [32 workers](https://github.com/tqdm/tqdm/blob/master/tqdm/contrib/concurrent.py#L44)* from `tqdm.contrib.concurrent.thread_map` whic... | closed | 2024-03-04T17:00:30Z | 2024-03-04T19:01:57Z | 2024-03-04T18:09:55Z | polinaeterna |
2,167,252,990 | Truncate binary for major tom | Fix viewer for https://huggingface.co/datasets/Major-TOM/Core-S2L2A
The dataset has one row per row group because each row contains multiple binary columns with a lot of data.
I fixed this by truncating binary data iteratively on row groups to not fill up the RAM
For now I only enable this logic on Major-TOM d... | Truncate binary for major tom: Fix viewer for https://huggingface.co/datasets/Major-TOM/Core-S2L2A
The dataset has one row per row group because each row contains multiple binary columns with a lot of data.
I fixed this by truncating binary data iteratively on row groups to not fill up the RAM
For now I only e... | closed | 2024-03-04T16:29:30Z | 2024-03-11T21:10:09Z | 2024-03-04T20:26:03Z | lhoestq |
2,166,988,697 | Update datasets and hfh | This should make the viewer faster to get for datasets with images/audio in parquet file.
I also updated hfh - close https://github.com/huggingface/datasets-server/issues/2317 | Update datasets and hfh: This should make the viewer faster to get for datasets with images/audio in parquet file.
I also updated hfh - close https://github.com/huggingface/datasets-server/issues/2317 | closed | 2024-03-04T14:30:53Z | 2024-03-04T20:15:09Z | 2024-03-04T20:15:08Z | lhoestq |
2,166,935,086 | Display default config first in the viewer | Would be nice to have an option to choose which data to show first because sometimes dataset starts with ugly data lol.
example: [bigcode/the-stack-v2](https://huggingface.co/datasets/bigcode/the-stack-v2). (yes, there is only one config with all the data for now but in theory it should be each folder inside `data/` ... | Display default config first in the viewer: Would be nice to have an option to choose which data to show first because sometimes dataset starts with ugly data lol.
example: [bigcode/the-stack-v2](https://huggingface.co/datasets/bigcode/the-stack-v2). (yes, there is only one config with all the data for now but in the... | closed | 2024-03-04T14:06:25Z | 2024-03-07T17:09:54Z | 2024-03-07T17:09:54Z | polinaeterna |
2,166,835,238 | Post-deploy job to delete old duckdb index files | Fix https://github.com/huggingface/datasets-server/issues/2520 and last part of https://github.com/huggingface/datasets-server/issues/1914
I propose a post-deploy execution job to remove old duckdb index files from **refs/convert/parquet branch**.
This script could be removed once all old index files have been remove... | Post-deploy job to delete old duckdb index files: Fix https://github.com/huggingface/datasets-server/issues/2520 and last part of https://github.com/huggingface/datasets-server/issues/1914
I propose a post-deploy execution job to remove old duckdb index files from **refs/convert/parquet branch**.
This script could be... | closed | 2024-03-04T13:19:51Z | 2024-04-11T05:26:24Z | 2024-03-07T19:03:53Z | AndreaFrancis |
2,163,888,456 | Better statistics for URL columns | URLs are stored as strings, so their stat info is their (string) lengths. We could make this info more informative by replacing the lengths with URL domain distribution (or some other stat/metric?). | Better statistics for URL columns: URLs are stored as strings, so their stat info is their (string) lengths. We could make this info more informative by replacing the lengths with URL domain distribution (or some other stat/metric?). | open | 2024-03-01T17:43:58Z | 2024-04-02T22:53:38Z | null | mariosasko |
2,163,406,719 | Enable workers HPA based on queue size (next) | null | Enable workers HPA based on queue size (next): | closed | 2024-03-01T13:22:58Z | 2024-03-01T13:40:40Z | 2024-03-01T13:40:39Z | rtrompier |
2,163,043,503 | Enable workers HPA based on queue size | null | Enable workers HPA based on queue size: | closed | 2024-03-01T10:02:57Z | 2024-03-01T12:21:41Z | 2024-03-01T10:35:00Z | rtrompier |
2,161,865,783 | Support yodas | [espnet/yodas](https://huggingface.co/datasets/espnet/yodas) | Support yodas: [espnet/yodas](https://huggingface.co/datasets/espnet/yodas) | closed | 2024-02-29T18:35:12Z | 2024-02-29T18:37:39Z | 2024-02-29T18:37:38Z | lhoestq |
2,161,540,523 | [Do not merge] Follow croissant 1.0 specs | null | [Do not merge] Follow croissant 1.0 specs: | closed | 2024-02-29T15:36:14Z | 2024-02-29T16:12:58Z | 2024-02-29T16:12:39Z | ccl-core |
2,161,371,736 | use a generic LongStepProfiler | follow-up to #2469 and #2533 | use a generic LongStepProfiler: follow-up to #2469 and #2533 | closed | 2024-02-29T14:16:39Z | 2024-02-29T14:29:02Z | 2024-02-29T14:29:01Z | severo |
2,161,132,089 | Adjust Prometheus hist buckets for `split-descriptive-statistics` job runner | to be more aligned with the default ones.
See https://github.com/huggingface/datasets-server/pull/2469/files#r1507469902 | Adjust Prometheus hist buckets for `split-descriptive-statistics` job runner: to be more aligned with the default ones.
See https://github.com/huggingface/datasets-server/pull/2469/files#r1507469902 | closed | 2024-02-29T12:21:40Z | 2024-02-29T12:36:17Z | 2024-02-29T12:36:17Z | polinaeterna |
2,160,985,334 | Fix worker crash when first heartbeat conflicts with job start | ```
INFO: 2024-02-27 21:49:47,093 - root - [config-split-names] compute JobManager(job_id=65de58fa62c68c8c749fbce3 dataset=ChristophSchuhmann/queries job_info={'job_id': '65de58fa62c68c8c749fbce3', 'type': 'config-split-names', 'params': {'dataset': 'ChristophSchuhmann/queries', 'revision': '641561f9b27193174a37e5a57f... | Fix worker crash when first heartbeat conflicts with job start: ```
INFO: 2024-02-27 21:49:47,093 - root - [config-split-names] compute JobManager(job_id=65de58fa62c68c8c749fbce3 dataset=ChristophSchuhmann/queries job_info={'job_id': '65de58fa62c68c8c749fbce3', 'type': 'config-split-names', 'params': {'dataset': 'Chri... | open | 2024-02-29T11:00:17Z | 2024-06-19T14:27:26Z | null | lhoestq |
2,160,973,054 | Restore missing triggers between steps | fixes #2530
it:
- reverts #2489
- adds the missing triggering step
- adds a check when creating the processing graph, to ensure all the config- and dataset-level steps have at least one triggering step of the same level, to handle the case where we couldn't get the configs or the splits.
| Restore missing triggers between steps: fixes #2530
it:
- reverts #2489
- adds the missing triggering step
- adds a check when creating the processing graph, to ensure all the config- and dataset-level steps have at least one triggering step of the same level, to handle the case where we couldn't get the configs ... | closed | 2024-02-29T10:54:13Z | 2024-02-29T11:56:15Z | 2024-02-29T11:56:14Z | severo |
2,160,858,215 | Missing triggers between steps | possibly due to https://github.com/huggingface/datasets-server/pull/2489
We get errors like:
- https://huggingface.co/datasets/jmc255/aphantasia_drawing_dataset/discussions/1
- https://huggingface.co/datasets/yzeng58/CoBSAT/discussions/1
where the first step `dataset-config-names` gives an error, and no other e... | Missing triggers between steps: possibly due to https://github.com/huggingface/datasets-server/pull/2489
We get errors like:
- https://huggingface.co/datasets/jmc255/aphantasia_drawing_dataset/discussions/1
- https://huggingface.co/datasets/yzeng58/CoBSAT/discussions/1
where the first step `dataset-config-names... | closed | 2024-02-29T09:54:07Z | 2024-02-29T11:56:15Z | 2024-02-29T11:56:15Z | severo |
2,160,824,291 | statistics fail on 2A2I/Arabic-OpenHermes-2.5 | https://huggingface.co/datasets/2A2I/Arabic-OpenHermes-2.5
the error:
```
TypeError: unsupported operand type(s) for *: 'NoneType' and 'float'
```
Full trace:
```
{
"error": "unsupported operand type(s) for *: 'NoneType' and 'float'",
"cause_exception": "TypeError",
"cause_message": "unsuppo... | statistics fail on 2A2I/Arabic-OpenHermes-2.5: https://huggingface.co/datasets/2A2I/Arabic-OpenHermes-2.5
the error:
```
TypeError: unsupported operand type(s) for *: 'NoneType' and 'float'
```
Full trace:
```
{
"error": "unsupported operand type(s) for *: 'NoneType' and 'float'",
"cause_exceptio... | closed | 2024-02-29T09:39:58Z | 2024-02-29T11:13:27Z | 2024-02-29T11:13:20Z | severo |
2,160,804,825 | run the backfill on retryable errors every 30 min (not every 10 min) | null | run the backfill on retryable errors every 30 min (not every 10 min): | closed | 2024-02-29T09:31:19Z | 2024-02-29T10:12:44Z | 2024-02-29T09:36:49Z | severo |
2,159,385,240 | Don't unquote path_in_repo | Support paths containing `%` | Don't unquote path_in_repo: Support paths containing `%` | closed | 2024-02-28T16:43:35Z | 2024-04-08T15:04:13Z | 2024-04-08T15:04:13Z | lhoestq |
2,158,955,584 | add metrics on jobs count per worker type | See #2095.
Add metrics `worker_size_jobs_count{pid="10",worker_size="heavy"}: 12` to prometheus (value for worker_size: `heavy`, `medium` and `light`).
It will be used to autoscale the workers
Also:
- difficulty_min is now excluded (and 0 is not a valid difficulty anymore)
- in prod, medium and heavy worker... | add metrics on jobs count per worker type: See #2095.
Add metrics `worker_size_jobs_count{pid="10",worker_size="heavy"}: 12` to prometheus (value for worker_size: `heavy`, `medium` and `light`).
It will be used to autoscale the workers
Also:
- difficulty_min is now excluded (and 0 is not a valid difficulty a... | closed | 2024-02-28T13:18:54Z | 2024-02-28T16:52:45Z | 2024-02-28T16:52:45Z | severo |
2,158,648,183 | Use hf_transfer to speed up downloads from the Hub | In /search, in particular
https://github.com/huggingface/hf_transfer/ or `HF_HUB_ENABLE_HF_TRANSFER` in huggingface_hub
| Use hf_transfer to speed up downloads from the Hub: In /search, in particular
https://github.com/huggingface/hf_transfer/ or `HF_HUB_ENABLE_HF_TRANSFER` in huggingface_hub
| closed | 2024-02-28T10:40:31Z | 2024-04-15T11:37:07Z | 2024-04-15T11:37:07Z | severo |
2,157,709,268 | Compute croissant in workers | The croissant metadata is computed by workers and stored in the cache
When served by the API service, and if `?full=false` is passed in the query, the columns are limited to 1,000 in the response.
fix #2521 | Compute croissant in workers: The croissant metadata is computed by workers and stored in the cache
When served by the API service, and if `?full=false` is passed in the query, the columns are limited to 1,000 in the response.
fix #2521 | closed | 2024-02-27T22:18:45Z | 2024-03-29T10:03:48Z | 2024-03-29T10:03:48Z | severo |
2,157,406,547 | Don't augment difficulty of children jobs | Some easy jobs have unexpected high difficulties because of that
| Don't augment difficulty of children jobs: Some easy jobs have unexpected high difficulties because of that
| closed | 2024-02-27T19:07:44Z | 2024-02-27T21:07:42Z | 2024-02-27T21:07:42Z | lhoestq |
2,157,306,761 | Fix backfill wrong difficulties, causing OOMs | The backfill was creating duckdb indexing jobs for big datasets without adjusting the difficulty.
Because of that the medium workers were processing them and OOMing and every retry. | Fix backfill wrong difficulties, causing OOMs: The backfill was creating duckdb indexing jobs for big datasets without adjusting the difficulty.
Because of that the medium workers were processing them and OOMing and every retry. | closed | 2024-02-27T18:09:42Z | 2024-02-27T20:54:33Z | 2024-02-27T20:54:32Z | lhoestq |
2,157,258,083 | Move /croissant to the cache | As it's called on every dataset page, we should cache it | Move /croissant to the cache: As it's called on every dataset page, we should cache it | closed | 2024-02-27T17:40:10Z | 2024-03-29T10:03:49Z | 2024-03-29T10:03:49Z | severo |
2,156,847,921 | Delete all .duckdb files from refs/convert/parquet | see https://github.com/huggingface/datasets-server/issues/1914 | Delete all .duckdb files from refs/convert/parquet: see https://github.com/huggingface/datasets-server/issues/1914 | closed | 2024-02-27T15:10:53Z | 2024-03-08T11:58:17Z | 2024-03-08T11:58:17Z | severo |
2,156,796,234 | backfill retryable errors every 10 minutes | fixes #2439.
For example, if a Parquet file commit to the Hub fails, we want to retry quickly (but not immediately, in case the Hub has an issue). So... checking and retrying every 10 minutes seems a good compromise. | backfill retryable errors every 10 minutes: fixes #2439.
For example, if a Parquet file commit to the Hub fails, we want to retry quickly (but not immediately, in case the Hub has an issue). So... checking and retrying every 10 minutes seems a good compromise. | closed | 2024-02-27T14:52:38Z | 2024-02-28T18:08:13Z | 2024-02-28T18:08:13Z | severo |
2,156,610,693 | Tweak the production resources | See https://www.notion.so/huggingface2/Infrastructure-b4fd07f015e04a84a41ec6472c8a0ff5 and https://huggingface.slack.com/archives/C04L6P8KNQ5/p1702553107305559 (internal links) | Tweak the production resources: See https://www.notion.so/huggingface2/Infrastructure-b4fd07f015e04a84a41ec6472c8a0ff5 and https://huggingface.slack.com/archives/C04L6P8KNQ5/p1702553107305559 (internal links) | closed | 2024-02-27T13:33:50Z | 2024-02-28T17:55:50Z | 2024-02-28T17:55:49Z | severo |
2,156,042,660 | Understand why simple steps get high difficulty | It's weird that jobs like `split-is-valid` and `config-duckdb-index-size`, which are very simple, get a high difficulty value.
<img width="476" alt="Capture d’écran 2024-02-27 à 10 01 10" src="https://github.com/huggingface/datasets-server/assets/1676121/0c69cadd-78a0-4478-a37f-5abff642627d">
| Understand why simple steps get high difficulty: It's weird that jobs like `split-is-valid` and `config-duckdb-index-size`, which are very simple, get a high difficulty value.
<img width="476" alt="Capture d’écran 2024-02-27 à 10 01 10" src="https://github.com/huggingface/datasets-server/assets/1676121/0c69cadd-78... | closed | 2024-02-27T09:02:28Z | 2024-02-29T10:58:38Z | 2024-02-29T10:58:37Z | severo |
2,155,884,924 | Update orjson to 3.9.15 to fix vulnerability | Update `orjson` to 3.9.15 to fix vulnerability: https://github.com/ijl/orjson/releases/tag/3.9.15
- Affected versions: < 3.9.15
This will fix 11 Dependabot alerts. | Update orjson to 3.9.15 to fix vulnerability: Update `orjson` to 3.9.15 to fix vulnerability: https://github.com/ijl/orjson/releases/tag/3.9.15
- Affected versions: < 3.9.15
This will fix 11 Dependabot alerts. | closed | 2024-02-27T07:41:45Z | 2024-02-27T09:22:05Z | 2024-02-27T09:22:04Z | albertvillanova |
2,155,254,955 | Bump the pip group group in /libs/libapi with 1 update | Bumps the pip group group in /libs/libapi with 1 update: [orjson](https://github.com/ijl/orjson).
Updates `orjson` from 3.9.7 to 3.9.15
<details>
<summary>Release notes</summary>
<p><em>Sourced from <a href="https://github.com/ijl/orjson/releases">orjson's releases</a>.</em></p>
<blockquote>
<h2>3.9.15</h2>
<h3>Fixed<... | Bump the pip group group in /libs/libapi with 1 update: Bumps the pip group group in /libs/libapi with 1 update: [orjson](https://github.com/ijl/orjson).
Updates `orjson` from 3.9.7 to 3.9.15
<details>
<summary>Release notes</summary>
<p><em>Sourced from <a href="https://github.com/ijl/orjson/releases">orjson's releas... | closed | 2024-02-26T22:09:26Z | 2024-02-27T09:23:21Z | 2024-02-27T09:23:20Z | dependabot[bot] |
2,155,254,659 | Bump orjson from 3.9.7 to 3.9.15 in /services/sse-api | Bumps [orjson](https://github.com/ijl/orjson) from 3.9.7 to 3.9.15.
<details>
<summary>Release notes</summary>
<p><em>Sourced from <a href="https://github.com/ijl/orjson/releases">orjson's releases</a>.</em></p>
<blockquote>
<h2>3.9.15</h2>
<h3>Fixed</h3>
<ul>
<li>Implement recursion limit of 1024 on <code>orjson.loads... | Bump orjson from 3.9.7 to 3.9.15 in /services/sse-api: Bumps [orjson](https://github.com/ijl/orjson) from 3.9.7 to 3.9.15.
<details>
<summary>Release notes</summary>
<p><em>Sourced from <a href="https://github.com/ijl/orjson/releases">orjson's releases</a>.</em></p>
<blockquote>
<h2>3.9.15</h2>
<h3>Fixed</h3>
<ul>
<li>... | closed | 2024-02-26T22:09:14Z | 2024-02-27T09:23:13Z | 2024-02-27T09:23:12Z | dependabot[bot] |
2,155,253,810 | Bump orjson from 3.9.2 to 3.9.15 in /services/search | Bumps [orjson](https://github.com/ijl/orjson) from 3.9.2 to 3.9.15.
<details>
<summary>Release notes</summary>
<p><em>Sourced from <a href="https://github.com/ijl/orjson/releases">orjson's releases</a>.</em></p>
<blockquote>
<h2>3.9.15</h2>
<h3>Fixed</h3>
<ul>
<li>Implement recursion limit of 1024 on <code>orjson.loads... | Bump orjson from 3.9.2 to 3.9.15 in /services/search: Bumps [orjson](https://github.com/ijl/orjson) from 3.9.2 to 3.9.15.
<details>
<summary>Release notes</summary>
<p><em>Sourced from <a href="https://github.com/ijl/orjson/releases">orjson's releases</a>.</em></p>
<blockquote>
<h2>3.9.15</h2>
<h3>Fixed</h3>
<ul>
<li>I... | closed | 2024-02-26T22:08:40Z | 2024-02-27T09:22:57Z | 2024-02-27T09:22:55Z | dependabot[bot] |
2,155,225,470 | Bump orjson from 3.9.1 to 3.9.15 in /services/rows | Bumps [orjson](https://github.com/ijl/orjson) from 3.9.1 to 3.9.15.
<details>
<summary>Release notes</summary>
<p><em>Sourced from <a href="https://github.com/ijl/orjson/releases">orjson's releases</a>.</em></p>
<blockquote>
<h2>3.9.15</h2>
<h3>Fixed</h3>
<ul>
<li>Implement recursion limit of 1024 on <code>orjson.loads... | Bump orjson from 3.9.1 to 3.9.15 in /services/rows: Bumps [orjson](https://github.com/ijl/orjson) from 3.9.1 to 3.9.15.
<details>
<summary>Release notes</summary>
<p><em>Sourced from <a href="https://github.com/ijl/orjson/releases">orjson's releases</a>.</em></p>
<blockquote>
<h2>3.9.15</h2>
<h3>Fixed</h3>
<ul>
<li>Imp... | closed | 2024-02-26T21:51:53Z | 2024-02-27T09:22:52Z | 2024-02-27T09:22:50Z | dependabot[bot] |
2,155,222,681 | Bump orjson from 3.9.0 to 3.9.15 in /services/worker | Bumps [orjson](https://github.com/ijl/orjson) from 3.9.0 to 3.9.15.
<details>
<summary>Release notes</summary>
<p><em>Sourced from <a href="https://github.com/ijl/orjson/releases">orjson's releases</a>.</em></p>
<blockquote>
<h2>3.9.15</h2>
<h3>Fixed</h3>
<ul>
<li>Implement recursion limit of 1024 on <code>orjson.loads... | Bump orjson from 3.9.0 to 3.9.15 in /services/worker: Bumps [orjson](https://github.com/ijl/orjson) from 3.9.0 to 3.9.15.
<details>
<summary>Release notes</summary>
<p><em>Sourced from <a href="https://github.com/ijl/orjson/releases">orjson's releases</a>.</em></p>
<blockquote>
<h2>3.9.15</h2>
<h3>Fixed</h3>
<ul>
<li>I... | closed | 2024-02-26T21:50:41Z | 2024-02-27T09:23:06Z | 2024-02-27T09:23:05Z | dependabot[bot] |
2,155,219,047 | Bump orjson from 3.9.0 to 3.9.15 in /front/admin_ui | Bumps [orjson](https://github.com/ijl/orjson) from 3.9.0 to 3.9.15.
<details>
<summary>Release notes</summary>
<p><em>Sourced from <a href="https://github.com/ijl/orjson/releases">orjson's releases</a>.</em></p>
<blockquote>
<h2>3.9.15</h2>
<h3>Fixed</h3>
<ul>
<li>Implement recursion limit of 1024 on <code>orjson.loads... | Bump orjson from 3.9.0 to 3.9.15 in /front/admin_ui: Bumps [orjson](https://github.com/ijl/orjson) from 3.9.0 to 3.9.15.
<details>
<summary>Release notes</summary>
<p><em>Sourced from <a href="https://github.com/ijl/orjson/releases">orjson's releases</a>.</em></p>
<blockquote>
<h2>3.9.15</h2>
<h3>Fixed</h3>
<ul>
<li>Im... | closed | 2024-02-26T21:47:21Z | 2024-02-27T09:22:59Z | 2024-02-27T09:22:58Z | dependabot[bot] |
2,155,218,297 | Bump orjson from 3.9.0 to 3.9.15 in /libs/libcommon | Bumps [orjson](https://github.com/ijl/orjson) from 3.9.0 to 3.9.15.
<details>
<summary>Release notes</summary>
<p><em>Sourced from <a href="https://github.com/ijl/orjson/releases">orjson's releases</a>.</em></p>
<blockquote>
<h2>3.9.15</h2>
<h3>Fixed</h3>
<ul>
<li>Implement recursion limit of 1024 on <code>orjson.loads... | Bump orjson from 3.9.0 to 3.9.15 in /libs/libcommon: Bumps [orjson](https://github.com/ijl/orjson) from 3.9.0 to 3.9.15.
<details>
<summary>Release notes</summary>
<p><em>Sourced from <a href="https://github.com/ijl/orjson/releases">orjson's releases</a>.</em></p>
<blockquote>
<h2>3.9.15</h2>
<h3>Fixed</h3>
<ul>
<li>Im... | closed | 2024-02-26T21:46:51Z | 2024-02-27T09:22:51Z | 2024-02-27T09:22:50Z | dependabot[bot] |
2,155,218,017 | Bump orjson from 3.9.0 to 3.9.15 in /jobs/cache_maintenance | Bumps [orjson](https://github.com/ijl/orjson) from 3.9.0 to 3.9.15.
<details>
<summary>Release notes</summary>
<p><em>Sourced from <a href="https://github.com/ijl/orjson/releases">orjson's releases</a>.</em></p>
<blockquote>
<h2>3.9.15</h2>
<h3>Fixed</h3>
<ul>
<li>Implement recursion limit of 1024 on <code>orjson.loads... | Bump orjson from 3.9.0 to 3.9.15 in /jobs/cache_maintenance: Bumps [orjson](https://github.com/ijl/orjson) from 3.9.0 to 3.9.15.
<details>
<summary>Release notes</summary>
<p><em>Sourced from <a href="https://github.com/ijl/orjson/releases">orjson's releases</a>.</em></p>
<blockquote>
<h2>3.9.15</h2>
<h3>Fixed</h3>
<ul... | closed | 2024-02-26T21:46:41Z | 2024-02-27T09:23:35Z | 2024-02-27T09:22:48Z | dependabot[bot] |
2,155,218,013 | Bump orjson from 3.9.0 to 3.9.15 in /jobs/cache_maintenance | Bumps [orjson](https://github.com/ijl/orjson) from 3.9.0 to 3.9.15.
<details>
<summary>Release notes</summary>
<p><em>Sourced from <a href="https://github.com/ijl/orjson/releases">orjson's releases</a>.</em></p>
<blockquote>
<h2>3.9.15</h2>
<h3>Fixed</h3>
<ul>
<li>Implement recursion limit of 1024 on <code>orjson.loads... | Bump orjson from 3.9.0 to 3.9.15 in /jobs/cache_maintenance: Bumps [orjson](https://github.com/ijl/orjson) from 3.9.0 to 3.9.15.
<details>
<summary>Release notes</summary>
<p><em>Sourced from <a href="https://github.com/ijl/orjson/releases">orjson's releases</a>.</em></p>
<blockquote>
<h2>3.9.15</h2>
<h3>Fixed</h3>
<ul... | closed | 2024-02-26T21:46:40Z | 2024-02-27T09:23:35Z | 2024-02-27T09:23:33Z | dependabot[bot] |
2,155,217,989 | Bump orjson from 3.9.0 to 3.9.15 in /services/admin | Bumps [orjson](https://github.com/ijl/orjson) from 3.9.0 to 3.9.15.
<details>
<summary>Release notes</summary>
<p><em>Sourced from <a href="https://github.com/ijl/orjson/releases">orjson's releases</a>.</em></p>
<blockquote>
<h2>3.9.15</h2>
<h3>Fixed</h3>
<ul>
<li>Implement recursion limit of 1024 on <code>orjson.loads... | Bump orjson from 3.9.0 to 3.9.15 in /services/admin: Bumps [orjson](https://github.com/ijl/orjson) from 3.9.0 to 3.9.15.
<details>
<summary>Release notes</summary>
<p><em>Sourced from <a href="https://github.com/ijl/orjson/releases">orjson's releases</a>.</em></p>
<blockquote>
<h2>3.9.15</h2>
<h3>Fixed</h3>
<ul>
<li>Im... | closed | 2024-02-26T21:46:39Z | 2024-02-27T09:22:50Z | 2024-02-27T09:22:49Z | dependabot[bot] |
2,155,217,978 | Bump orjson from 3.9.0 to 3.9.15 in /services/api | Bumps [orjson](https://github.com/ijl/orjson) from 3.9.0 to 3.9.15.
<details>
<summary>Release notes</summary>
<p><em>Sourced from <a href="https://github.com/ijl/orjson/releases">orjson's releases</a>.</em></p>
<blockquote>
<h2>3.9.15</h2>
<h3>Fixed</h3>
<ul>
<li>Implement recursion limit of 1024 on <code>orjson.loads... | Bump orjson from 3.9.0 to 3.9.15 in /services/api: Bumps [orjson](https://github.com/ijl/orjson) from 3.9.0 to 3.9.15.
<details>
<summary>Release notes</summary>
<p><em>Sourced from <a href="https://github.com/ijl/orjson/releases">orjson's releases</a>.</em></p>
<blockquote>
<h2>3.9.15</h2>
<h3>Fixed</h3>
<ul>
<li>Impl... | closed | 2024-02-26T21:46:39Z | 2024-02-27T09:23:34Z | 2024-02-27T09:23:33Z | dependabot[bot] |
2,155,217,892 | Bump orjson from 3.9.0 to 3.9.15 in /jobs/mongodb_migration | Bumps [orjson](https://github.com/ijl/orjson) from 3.9.0 to 3.9.15.
<details>
<summary>Release notes</summary>
<p><em>Sourced from <a href="https://github.com/ijl/orjson/releases">orjson's releases</a>.</em></p>
<blockquote>
<h2>3.9.15</h2>
<h3>Fixed</h3>
<ul>
<li>Implement recursion limit of 1024 on <code>orjson.loads... | Bump orjson from 3.9.0 to 3.9.15 in /jobs/mongodb_migration: Bumps [orjson](https://github.com/ijl/orjson) from 3.9.0 to 3.9.15.
<details>
<summary>Release notes</summary>
<p><em>Sourced from <a href="https://github.com/ijl/orjson/releases">orjson's releases</a>.</em></p>
<blockquote>
<h2>3.9.15</h2>
<h3>Fixed</h3>
<ul... | closed | 2024-02-26T21:46:36Z | 2024-02-27T09:22:49Z | 2024-02-27T09:22:48Z | dependabot[bot] |
2,154,767,928 | Support parquet datasets with >10k files | By respecting the 10k files limit per folder when copying them in `refs/convert/parquet` .
To do so I'm using a directory scheme like config/train-part0/0000.parquet up to config/train-part9/9999.parquet when there are >10k files
Close https://github.com/huggingface/datasets-server/issues/2498
cc @guipenedo | Support parquet datasets with >10k files: By respecting the 10k files limit per folder when copying them in `refs/convert/parquet` .
To do so I'm using a directory scheme like config/train-part0/0000.parquet up to config/train-part9/9999.parquet when there are >10k files
Close https://github.com/huggingface/datas... | closed | 2024-02-26T17:51:02Z | 2024-02-27T16:52:34Z | 2024-02-27T16:52:33Z | lhoestq |
2,151,869,571 | Number of cache entries in metrics is incorrect | Currently 4.22M, while it should be 3.7M
see #2496. | Number of cache entries in metrics is incorrect: Currently 4.22M, while it should be 3.7M
see #2496. | closed | 2024-02-23T21:59:05Z | 2024-02-23T22:00:54Z | 2024-02-23T22:00:53Z | severo |
2,151,858,600 | reduce resources to 4/10/4 | null | reduce resources to 4/10/4: | closed | 2024-02-23T21:49:40Z | 2024-02-23T21:50:19Z | 2024-02-23T21:49:48Z | severo |
2,151,844,666 | move the backfill hour + refresh cache metrics every 3 hours | null | move the backfill hour + refresh cache metrics every 3 hours: | closed | 2024-02-23T21:37:25Z | 2024-02-23T21:38:04Z | 2024-02-23T21:37:31Z | severo |
2,151,525,679 | Fix extensions directory search | While deploying https://github.com/huggingface/datasets-server/pull/2482 in staging, this error appeared:
```
INFO: 2024-02-23 17:03:36,387 - root - /search dataset='asoria/bolivian-recipes' config='default' split='same' query='beef' offset=0 length=100
DEBUG: 2024-02-23 17:03:36,413 - root - connect to index file /... | Fix extensions directory search: While deploying https://github.com/huggingface/datasets-server/pull/2482 in staging, this error appeared:
```
INFO: 2024-02-23 17:03:36,387 - root - /search dataset='asoria/bolivian-recipes' config='default' split='same' query='beef' offset=0 length=100
DEBUG: 2024-02-23 17:03:36,413... | closed | 2024-02-23T17:39:22Z | 2024-02-23T17:50:18Z | 2024-02-23T17:50:18Z | AndreaFrancis |
2,151,427,623 | Support datasets with >10k Parquet files | Fix this ([internal](https://huggingface.slack.com/archives/C02V51Q3800/p1708702334005179?thread_ts=1708606229.359269&cid=C02V51Q3800))
I'll use a directory scheme like `config/train-part0/0000.parquet` up... | Support datasets with >10k Parquet files: Fix this ([internal](https://huggingface.slack.com/archives/C02V51Q3800/p1708702334005179?thread_ts=1708606229.359269&cid=C02V51Q3800))
I'll use a directory scheme... | closed | 2024-02-23T16:34:19Z | 2024-02-27T16:52:34Z | 2024-02-27T16:52:34Z | lhoestq |
2,151,199,156 | Fix metrics | fixes #2496 | Fix metrics: fixes #2496 | closed | 2024-02-23T14:24:27Z | 2024-02-23T14:37:04Z | 2024-02-23T14:37:04Z | severo |
2,150,985,120 | Cache and Queue metrics should not include obsolete values | For example, after deleting the step "config-split-names-from-streaming", we still had a lot of entries for it in the metrics. We should check and delete them automatically, as they are not present anymore in the cache collection.
```
db.cacheTotalMetric.deleteMany({kind: "config-split-names-from-streaming"})
{ ac... | Cache and Queue metrics should not include obsolete values: For example, after deleting the step "config-split-names-from-streaming", we still had a lot of entries for it in the metrics. We should check and delete them automatically, as they are not present anymore in the cache collection.
```
db.cacheTotalMetric.d... | closed | 2024-02-23T12:27:03Z | 2024-02-23T15:06:25Z | 2024-02-23T14:37:05Z | severo |
2,150,980,402 | Cache/Queue metrics should not be negative | In the cache metrics collection, we have entries like:
```
{ _id: ObjectId("65d60409acf7b3719825797e"),
error_code: 'PreviousStepStillProcessingError',
http_status: 500,
kind: 'split-duckdb-index-010',
total: -67 }
```
The total should never be lower than 0
| Cache/Queue metrics should not be negative: In the cache metrics collection, we have entries like:
```
{ _id: ObjectId("65d60409acf7b3719825797e"),
error_code: 'PreviousStepStillProcessingError',
http_status: 500,
kind: 'split-duckdb-index-010',
total: -67 }
```
The total should never be lower than ... | open | 2024-02-23T12:23:58Z | 2024-03-08T15:42:26Z | null | severo |
2,150,839,878 | remove MetricsMongoResource and METRICS_MONGOENGINE_ALIAS | fixes https://github.com/huggingface/datasets-server/issues/2210 | remove MetricsMongoResource and METRICS_MONGOENGINE_ALIAS: fixes https://github.com/huggingface/datasets-server/issues/2210 | closed | 2024-02-23T10:53:03Z | 2024-02-23T11:24:01Z | 2024-02-23T11:24:00Z | severo |
2,150,067,161 | set updated_at field to the same value for all steps of a dataset | After the migrations:
- `_20240221103200_cache_merge_config_split_names`
- `_20240221160700_cache_merge_split_first_rows`
the updated_at field is not accurate anymore, leading to backfilling a lot of them (in the daily cronjob), leading to unnecessary jobs.
To fix that, we take the updated_at value for the root... | set updated_at field to the same value for all steps of a dataset: After the migrations:
- `_20240221103200_cache_merge_config_split_names`
- `_20240221160700_cache_merge_split_first_rows`
the updated_at field is not accurate anymore, leading to backfilling a lot of them (in the daily cronjob), leading to unnecess... | closed | 2024-02-22T22:43:05Z | 2024-02-23T12:57:07Z | 2024-02-23T12:57:06Z | severo |
2,150,032,390 | Bump the pip group across 2 directories with 1 update | Bumps the pip group with 1 update in the /front/admin_ui directory: [gradio](https://github.com/gradio-app/gradio).
Updates `gradio` from 4.18.0 to 4.19.2
<details>
<summary>Release notes</summary>
<p><em>Sourced from <a href="https://github.com/gradio-app/gradio/releases">gradio's releases</a>.</em></p>
<blockquote>
... | Bump the pip group across 2 directories with 1 update: Bumps the pip group with 1 update in the /front/admin_ui directory: [gradio](https://github.com/gradio-app/gradio).
Updates `gradio` from 4.18.0 to 4.19.2
<details>
<summary>Release notes</summary>
<p><em>Sourced from <a href="https://github.com/gradio-app/gradio/... | closed | 2024-02-22T22:15:37Z | 2024-02-23T07:21:37Z | 2024-02-22T22:53:16Z | dependabot[bot] |
2,149,961,591 | retry error PreviousStepStillProcessingError - at least every day | null | retry error PreviousStepStillProcessingError - at least every day: | closed | 2024-02-22T21:22:26Z | 2024-02-22T21:22:36Z | 2024-02-22T21:22:35Z | severo |
2,149,587,793 | upgrade ruff to >=2.1.0 | fixes #2479 | upgrade ruff to >=2.1.0: fixes #2479 | closed | 2024-02-22T17:42:40Z | 2024-02-22T18:00:18Z | 2024-02-22T18:00:17Z | severo |
2,149,560,027 | remove unnecessary triggers from processing graph | null | remove unnecessary triggers from processing graph: | closed | 2024-02-22T17:26:30Z | 2024-02-22T17:59:25Z | 2024-02-22T17:59:24Z | severo |
2,149,253,370 | count is deprecated | null | count is deprecated: | closed | 2024-02-22T14:51:00Z | 2024-02-22T14:52:55Z | 2024-02-22T14:52:54Z | severo |
2,149,206,724 | use the new job runner version when migrating | null | use the new job runner version when migrating: | closed | 2024-02-22T14:30:45Z | 2024-02-22T14:32:10Z | 2024-02-22T14:32:09Z | severo |
2,148,702,361 | Add schema.org/license and schema.org/identifier to Croissant. | null | Add schema.org/license and schema.org/identifier to Croissant.: | closed | 2024-02-22T10:15:01Z | 2024-02-22T13:57:29Z | 2024-02-22T13:57:29Z | marcenacp |
2,148,397,951 | Update cryptography to 42.0.4 to fix vulnerability | Update `cryptography` to 42.0.4 to fix vulnerability:
- Affected versions: >= 38.0.0, < 42.0.4
This will fix 12 Dependabot alerts. | Update cryptography to 42.0.4 to fix vulnerability: Update `cryptography` to 42.0.4 to fix vulnerability:
- Affected versions: >= 38.0.0, < 42.0.4
This will fix 12 Dependabot alerts. | closed | 2024-02-22T07:26:13Z | 2024-02-22T09:14:57Z | 2024-02-22T09:14:56Z | albertvillanova |
2,147,809,473 | Increase resources | Same as https://github.com/huggingface/datasets-server/pull/2468, increasing resources to 10/80/10 since we have like 28K pending jobs for split-duckdb-index-010 | Increase resources: Same as https://github.com/huggingface/datasets-server/pull/2468, increasing resources to 10/80/10 since we have like 28K pending jobs for split-duckdb-index-010 | closed | 2024-02-21T22:18:48Z | 2024-02-21T22:29:03Z | 2024-02-21T22:29:02Z | AndreaFrancis |
2,147,421,947 | Update pandas version to 2.2 | Part of https://github.com/huggingface/datasets-server/issues/1914
Depends on https://github.com/huggingface/datasets-server/pull/2482 | Update pandas version to 2.2: Part of https://github.com/huggingface/datasets-server/issues/1914
Depends on https://github.com/huggingface/datasets-server/pull/2482 | closed | 2024-02-21T18:23:51Z | 2024-02-23T16:47:44Z | 2024-02-23T16:47:43Z | AndreaFrancis |
2,147,388,275 | Switch to duckdb 010 in search/filter service | Part of https://github.com/huggingface/datasets-server/issues/1914
Following https://github.com/huggingface/datasets-server/pull/2463
**WARNING:** This PR should be merged only once all "split-duckdb-index-010" entries have been computed. And workers should be scaled down to 0 before deploying.
Changes done:
- Mi... | Switch to duckdb 010 in search/filter service: Part of https://github.com/huggingface/datasets-server/issues/1914
Following https://github.com/huggingface/datasets-server/pull/2463
**WARNING:** This PR should be merged only once all "split-duckdb-index-010" entries have been computed. And workers should be scaled d... | closed | 2024-02-21T18:04:22Z | 2024-02-23T16:27:30Z | 2024-02-23T16:27:29Z | AndreaFrancis |
2,146,938,324 | temporary hardcode filter target revision | null | temporary hardcode filter target revision: | closed | 2024-02-21T14:47:22Z | 2024-02-21T14:47:40Z | 2024-02-21T14:47:40Z | AndreaFrancis |
2,146,890,292 | fix search/filter temporary hardcode target revision | null | fix search/filter temporary hardcode target revision: | closed | 2024-02-21T14:25:54Z | 2024-02-21T14:37:33Z | 2024-02-21T14:37:32Z | AndreaFrancis |
2,146,883,011 | Upgrade ruff? | My VSCode plugin says:
```
Ruff >=0.2.1 required, but found 0.1.3 at /home/slesage/hf/datasets-server/services/worker/.venv/bin/ruff
``` | Upgrade ruff?: My VSCode plugin says:
```
Ruff >=0.2.1 required, but found 0.1.3 at /home/slesage/hf/datasets-server/services/worker/.venv/bin/ruff
``` | closed | 2024-02-21T14:23:16Z | 2024-02-22T18:00:19Z | 2024-02-22T18:00:19Z | severo |
2,146,813,357 | Remove logic for "manual download" here? | Restriction to manual downloads is a feature for script-based datasets. Should we stop handling this in the code? | Remove logic for "manual download" here?: Restriction to manual downloads is a feature for script-based datasets. Should we stop handling this in the code? | closed | 2024-02-21T13:51:55Z | 2024-07-31T08:23:00Z | 2024-07-31T08:23:00Z | severo |
2,144,730,928 | remove parallel steps | With this PR, we remove the weird concept of parallel steps (which do not exist in DAG software/theory, I think).
Benefits:
- multiple parts of the code are now simpler
- fixes #2460
Drawbacks:
- we might end up computing from "streaming" twice when the Parquet has any error. Anyway, it should be 1. exception... | remove parallel steps: With this PR, we remove the weird concept of parallel steps (which do not exist in DAG software/theory, I think).
Benefits:
- multiple parts of the code are now simpler
- fixes #2460
Drawbacks:
- we might end up computing from "streaming" twice when the Parquet has any error. Anyway, it... | closed | 2024-02-20T16:07:15Z | 2024-02-22T14:03:59Z | 2024-02-22T14:03:58Z | severo |
2,144,471,994 | Remove unnecessary dependencies once we remove support for all script based datasets | Normally: services/worker should only depend on `datasets[audio,image]` to be able to load all the datasets. No need for specific ones like `pdf2image`, `kenlm` etc. | Remove unnecessary dependencies once we remove support for all script based datasets: Normally: services/worker should only depend on `datasets[audio,image]` to be able to load all the datasets. No need for specific ones like `pdf2image`, `kenlm` etc. | closed | 2024-02-20T14:11:17Z | 2024-05-29T09:18:37Z | 2024-05-29T09:18:37Z | severo |
2,144,467,936 | remove dependency 'trec-car-tools' from worker | fixes #2474 | remove dependency 'trec-car-tools' from worker: fixes #2474 | closed | 2024-02-20T14:09:21Z | 2024-02-20T14:57:32Z | 2024-02-20T14:57:31Z | severo |
2,144,460,242 | Remove vendored dependency `trec-car-tools` | See https://github.com/huggingface/datasets-server/tree/main/services/worker/vendors/trec-car-tools
It's not required anymore since we don't allow datasets with script. | Remove vendored dependency `trec-car-tools`: See https://github.com/huggingface/datasets-server/tree/main/services/worker/vendors/trec-car-tools
It's not required anymore since we don't allow datasets with script. | closed | 2024-02-20T14:05:35Z | 2024-02-20T14:57:32Z | 2024-02-20T14:57:32Z | severo |
2,144,064,689 | Remove the parallel steps? | I think it would be simpler to delete the concept of "parallel" steps.
Currently, we have two pairs of parallel steps:
1. `config-split-names-from-streaming` and `config-split-names-from-info`
2. `split-first-rows-from-streaming` and `split-first-rows-from-parquet`
It generates some complications:
- we don't... | Remove the parallel steps?: I think it would be simpler to delete the concept of "parallel" steps.
Currently, we have two pairs of parallel steps:
1. `config-split-names-from-streaming` and `config-split-names-from-info`
2. `split-first-rows-from-streaming` and `split-first-rows-from-parquet`
It generates som... | closed | 2024-02-20T10:37:24Z | 2024-02-22T14:04:00Z | 2024-02-22T14:04:00Z | severo |
2,143,176,730 | Revert "increase resources to 10/80/10 (#2468)" | This reverts commit f495e2c437689519f6bed507fb43b5a6f2d9b39c. | Revert "increase resources to 10/80/10 (#2468)": This reverts commit f495e2c437689519f6bed507fb43b5a6f2d9b39c. | closed | 2024-02-19T21:34:36Z | 2024-02-19T21:35:09Z | 2024-02-19T21:34:55Z | severo |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.