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 |
|---|---|---|---|---|---|---|---|---|
1,783,215,209 |
Adding retry to create duckdb index commit
|
Should fix HfHubHTTPError when creating commit for duckdb-index
Part of https://github.com/huggingface/datasets-server/issues/1462
|
Adding retry to create duckdb index commit: Should fix HfHubHTTPError when creating commit for duckdb-index
Part of https://github.com/huggingface/datasets-server/issues/1462
|
closed
|
2023-06-30T22:16:13Z
|
2023-07-03T12:56:23Z
|
2023-07-03T12:56:22Z
|
AndreaFrancis
|
1,783,075,231 |
Ensure parquet shards are sorted
|
fixes #1397
|
Ensure parquet shards are sorted: fixes #1397
|
closed
|
2023-06-30T19:44:37Z
|
2023-06-30T20:02:40Z
|
2023-06-30T20:02:13Z
|
severo
|
1,782,967,650 |
Change the way we represent ResponseAlreadyComputedError in the cache
|
When a "parallel" step has already been computed, an error is stored in the cache with `ResponseAlreadyComputedError`error_code, and http status 500 (ie: if `split-first-rows-from-streaming` exists, then `split-first-rows-from-parquet` does not need to be computed).
But it makes it hard to monitor the "true" errors. If we follow the analogy with the HTTP status codes, it should be 3xx instead of 5xx, ie: a redirection to another resource.
I don't know how we should change this though. Let's put ideas in the issue.
|
Change the way we represent ResponseAlreadyComputedError in the cache: When a "parallel" step has already been computed, an error is stored in the cache with `ResponseAlreadyComputedError`error_code, and http status 500 (ie: if `split-first-rows-from-streaming` exists, then `split-first-rows-from-parquet` does not need to be computed).
But it makes it hard to monitor the "true" errors. If we follow the analogy with the HTTP status codes, it should be 3xx instead of 5xx, ie: a redirection to another resource.
I don't know how we should change this though. Let's put ideas in the issue.
|
closed
|
2023-06-30T18:13:34Z
|
2024-02-23T09:56:05Z
|
2024-02-23T09:56:04Z
|
severo
|
1,782,821,542 |
Change the structure of parquet files
|
The parquet files will be stored in the `refs/convert/parquet`"branch" with the following structure:
```
[config]/[split]/[shard index: 0000 to 9999].parquet
```
Note that the "partially" converted datasets will use the following (See https://github.com/huggingface/datasets-server/pull/1448):
```
[config]/[split]/partial/[shard index: 0000 to 9999].parquet
```
|
Change the structure of parquet files: The parquet files will be stored in the `refs/convert/parquet`"branch" with the following structure:
```
[config]/[split]/[shard index: 0000 to 9999].parquet
```
Note that the "partially" converted datasets will use the following (See https://github.com/huggingface/datasets-server/pull/1448):
```
[config]/[split]/partial/[shard index: 0000 to 9999].parquet
```
|
closed
|
2023-06-30T16:33:12Z
|
2023-08-17T20:41:16Z
|
2023-08-17T20:41:16Z
|
severo
|
1,782,478,740 |
split-duckdb-index many UnexpectedError in error_code
|
Updated query (Without errors from parent):
```
db.cachedResponsesBlue.aggregate([{$match: {error_code: "UnexpectedError", kind:"split-duckdb-index", "details.copied_from_artifact":{$exists:false}}},{$group: {_id: {cause: "$details.cause_exception"}, count: {$sum: 1}}},{$sort: {count: -1}}])
```
From 128617 records currently existing in cache collection, these are the top kind of UnexpectedErrors:
```
[
{ _id: { cause: 'HfHubHTTPError' }, count: 4429 },
{ _id: { cause: 'HTTPException' }, count: 2570 },
{ _id: { cause: 'Error' }, count: 54 },
{ _id: { cause: 'BinderException' }, count: 41 },
{ _id: { cause: 'CatalogException' }, count: 38 },
{ _id: { cause: 'ParserException' }, count: 29 },
{ _id: { cause: 'InvalidInputException' }, count: 22 },
{ _id: { cause: 'RuntimeError' }, count: 8 },
{ _id: { cause: 'IOException' }, count: 5 },
{ _id: { cause: 'BadRequestError' }, count: 2 },
{ _id: { cause: 'NotPrimaryError' }, count: 2 },
{ _id: { cause: 'EntryNotFoundError' }, count: 2 }
]
```
Since this is a new job runner, most of these should be evaluated in case there is a bug in the code.
|
split-duckdb-index many UnexpectedError in error_code: Updated query (Without errors from parent):
```
db.cachedResponsesBlue.aggregate([{$match: {error_code: "UnexpectedError", kind:"split-duckdb-index", "details.copied_from_artifact":{$exists:false}}},{$group: {_id: {cause: "$details.cause_exception"}, count: {$sum: 1}}},{$sort: {count: -1}}])
```
From 128617 records currently existing in cache collection, these are the top kind of UnexpectedErrors:
```
[
{ _id: { cause: 'HfHubHTTPError' }, count: 4429 },
{ _id: { cause: 'HTTPException' }, count: 2570 },
{ _id: { cause: 'Error' }, count: 54 },
{ _id: { cause: 'BinderException' }, count: 41 },
{ _id: { cause: 'CatalogException' }, count: 38 },
{ _id: { cause: 'ParserException' }, count: 29 },
{ _id: { cause: 'InvalidInputException' }, count: 22 },
{ _id: { cause: 'RuntimeError' }, count: 8 },
{ _id: { cause: 'IOException' }, count: 5 },
{ _id: { cause: 'BadRequestError' }, count: 2 },
{ _id: { cause: 'NotPrimaryError' }, count: 2 },
{ _id: { cause: 'EntryNotFoundError' }, count: 2 }
]
```
Since this is a new job runner, most of these should be evaluated in case there is a bug in the code.
|
closed
|
2023-06-30T12:52:15Z
|
2023-08-11T15:44:16Z
|
2023-08-11T15:44:16Z
|
AndreaFrancis
|
1,781,341,194 |
Disable backfill k8s Job
| null |
Disable backfill k8s Job:
|
closed
|
2023-06-29T19:04:42Z
|
2023-06-29T19:06:03Z
|
2023-06-29T19:06:02Z
|
AndreaFrancis
|
1,781,072,941 |
feat: 🎸 backfill the datasets
|
because dataset-is-valid step version has been increased. Using this to also fix possible issues (see
https://github.com/huggingface/datasets-server/pull/1345 we had put on hold waiting for #1346, that has since be fixed)
|
feat: 🎸 backfill the datasets: because dataset-is-valid step version has been increased. Using this to also fix possible issues (see
https://github.com/huggingface/datasets-server/pull/1345 we had put on hold waiting for #1346, that has since be fixed)
|
closed
|
2023-06-29T15:30:58Z
|
2023-06-29T15:31:05Z
|
2023-06-29T15:31:04Z
|
severo
|
1,780,946,502 |
fix: split-duckdb-index error when indexing columns with spaces
|
After deploying split-duckdb-index, there are some errors because of column names with spaces like:
```
duckdb.BinderException: Binder Error: Referenced column "Mean" not found in FROM clause!
Candidate bindings: "read_parquet.Model"
LINE 1: ...f_index_id, Ranking,User,Model,Results,**Mean Reward**,Std Reward FROM read_parque...
^
```
This PR fixes this error.
|
fix: split-duckdb-index error when indexing columns with spaces: After deploying split-duckdb-index, there are some errors because of column names with spaces like:
```
duckdb.BinderException: Binder Error: Referenced column "Mean" not found in FROM clause!
Candidate bindings: "read_parquet.Model"
LINE 1: ...f_index_id, Ranking,User,Model,Results,**Mean Reward**,Std Reward FROM read_parque...
^
```
This PR fixes this error.
|
closed
|
2023-06-29T14:25:59Z
|
2023-06-29T15:59:00Z
|
2023-06-29T15:54:39Z
|
AndreaFrancis
|
1,780,313,353 |
Improve metrics to hide duplicates
|
Yesterday, a new step (`split-duckcb-index`) was added and run over all the datasets. Here are the metrics:
<img width="771" alt="Capture d’écran 2023-06-29 à 09 46 31" src="https://github.com/huggingface/datasets-server/assets/1676121/5daf2798-c79b-4ff9-a192-03ba38d4b149">
We can see that many new cache entries have been filled and that many of them are errors. But looking into the details, we can see that most of them are copies of the error in a previous step.
It would be useful to show the errors related only to the new step.
|
Improve metrics to hide duplicates: Yesterday, a new step (`split-duckcb-index`) was added and run over all the datasets. Here are the metrics:
<img width="771" alt="Capture d’écran 2023-06-29 à 09 46 31" src="https://github.com/huggingface/datasets-server/assets/1676121/5daf2798-c79b-4ff9-a192-03ba38d4b149">
We can see that many new cache entries have been filled and that many of them are errors. But looking into the details, we can see that most of them are copies of the error in a previous step.
It would be useful to show the errors related only to the new step.
|
closed
|
2023-06-29T07:49:28Z
|
2024-02-06T14:48:11Z
|
2024-02-06T14:48:11Z
|
severo
|
1,780,284,502 |
feat: 🎸 reduce the number of workers back to 20
| null |
feat: 🎸 reduce the number of workers back to 20:
|
closed
|
2023-06-29T07:27:34Z
|
2023-06-29T07:28:07Z
|
2023-06-29T07:27:39Z
|
severo
|
1,779,744,022 |
Disable backfill - ACTION = skip
| null |
Disable backfill - ACTION = skip:
|
closed
|
2023-06-28T20:41:20Z
|
2023-06-28T20:42:36Z
|
2023-06-28T20:42:34Z
|
AndreaFrancis
|
1,779,350,448 |
Update quality target in Makefile for /chart
|
The path to staging environment in the Makefile was outdated (the name was changed from `dev` to `staging`)
|
Update quality target in Makefile for /chart: The path to staging environment in the Makefile was outdated (the name was changed from `dev` to `staging`)
|
closed
|
2023-06-28T17:00:24Z
|
2023-06-28T18:33:54Z
|
2023-06-28T18:33:53Z
|
polinaeterna
|
1,779,215,015 |
Enable backfill one time
|
After this PR is merged and deployed, I will rollback to action=skip to disable the k8s job.
|
Enable backfill one time: After this PR is merged and deployed, I will rollback to action=skip to disable the k8s job.
|
closed
|
2023-06-28T15:36:22Z
|
2023-06-28T15:53:08Z
|
2023-06-28T15:53:07Z
|
AndreaFrancis
|
1,779,204,041 |
Temporaly increase resources for new job runner split-duckdb-index
| null |
Temporaly increase resources for new job runner split-duckdb-index:
|
closed
|
2023-06-28T15:29:19Z
|
2023-06-28T15:31:06Z
|
2023-06-28T15:31:05Z
|
AndreaFrancis
|
1,779,167,702 |
Replace valid with preview and viewer
|
replaces #1450 and #1447
fixes #1446 and #1445
- [x] remove `valid` field from `/valid` endpoint
- [x] replace `valid` with `viewer` and `preview` in `/is-valid`
- [x] update the docs
- [x] update openapi, <strike>rapidapi</strike> (I don't really understand rapidapi anymore), postman
This change is breaking the API for /valid and /is-valid. See https://moon-ci-docs.huggingface.co/docs/datasets-server/pr_1452/en/valid.
- [x] for Autotrain, I think the endpoints are not used in the code (https://github.com/huggingface/autotrain-ui and https://github.com/huggingface/autotrain-backend/). cc @abhishekkrthakur
- [x] for model-evaluator, I opened https://github.com/huggingface/model-evaluator/pull/71 cc @lewtun
|
Replace valid with preview and viewer: replaces #1450 and #1447
fixes #1446 and #1445
- [x] remove `valid` field from `/valid` endpoint
- [x] replace `valid` with `viewer` and `preview` in `/is-valid`
- [x] update the docs
- [x] update openapi, <strike>rapidapi</strike> (I don't really understand rapidapi anymore), postman
This change is breaking the API for /valid and /is-valid. See https://moon-ci-docs.huggingface.co/docs/datasets-server/pr_1452/en/valid.
- [x] for Autotrain, I think the endpoints are not used in the code (https://github.com/huggingface/autotrain-ui and https://github.com/huggingface/autotrain-backend/). cc @abhishekkrthakur
- [x] for model-evaluator, I opened https://github.com/huggingface/model-evaluator/pull/71 cc @lewtun
|
closed
|
2023-06-28T15:09:37Z
|
2023-06-29T14:13:46Z
|
2023-06-29T14:13:14Z
|
severo
|
1,779,124,224 |
Change duckdb commiter key
| null |
Change duckdb commiter key:
|
closed
|
2023-06-28T14:47:37Z
|
2023-06-28T14:50:37Z
|
2023-06-28T14:50:36Z
|
AndreaFrancis
|
1,779,109,875 |
Add preview and viewer to is valid
| null |
Add preview and viewer to is valid:
|
closed
|
2023-06-28T14:41:05Z
|
2023-06-28T15:10:19Z
|
2023-06-28T15:09:50Z
|
severo
|
1,779,028,625 |
Adding debug logs for split-duckdb-index
|
When processing split-duckdb-index in staging env, it is showing this message:
```
DEBUG: 2023-06-28 13:42:51,345 - root - The dataset does not exist on the Hub.
DEBUG: 2023-06-28 13:42:51,349 - root - Directory removed: /duckdb-index/21626898975922-split-duckdb-index-asoria-sample_glue-84f50613
DEBUG: 2023-06-28 13:42:51,349 - root - [split-duckdb-index] the dataset=asoria/sample_glue could not be found, don't update the cache
DEBUG: 2023-06-28 13:42:51,350 - root - [split-duckdb-index] job output with ERROR - JobManager(job_id=649c38d766b00b1fb05b8e4e dataset=asoria/sample_glue job_info={'job_id': '649c38d766b00b1fb05b8e4e', 'type': 'split-duckdb-index', 'params': {'dataset': 'asoria/sample_glue', 'revision': 'cf2d2d9273f1d361831baafab1f29eeb95a6af56', 'config': 'default', 'split': 'test'}, 'priority': <Priority.NORMAL: 'normal'>}
```
Looks like some hub operation does not find the dataset and it is not possible to store the cache result.
Adding a couple of logs for debug.
|
Adding debug logs for split-duckdb-index: When processing split-duckdb-index in staging env, it is showing this message:
```
DEBUG: 2023-06-28 13:42:51,345 - root - The dataset does not exist on the Hub.
DEBUG: 2023-06-28 13:42:51,349 - root - Directory removed: /duckdb-index/21626898975922-split-duckdb-index-asoria-sample_glue-84f50613
DEBUG: 2023-06-28 13:42:51,349 - root - [split-duckdb-index] the dataset=asoria/sample_glue could not be found, don't update the cache
DEBUG: 2023-06-28 13:42:51,350 - root - [split-duckdb-index] job output with ERROR - JobManager(job_id=649c38d766b00b1fb05b8e4e dataset=asoria/sample_glue job_info={'job_id': '649c38d766b00b1fb05b8e4e', 'type': 'split-duckdb-index', 'params': {'dataset': 'asoria/sample_glue', 'revision': 'cf2d2d9273f1d361831baafab1f29eeb95a6af56', 'config': 'default', 'split': 'test'}, 'priority': <Priority.NORMAL: 'normal'>}
```
Looks like some hub operation does not find the dataset and it is not possible to store the cache result.
Adding a couple of logs for debug.
|
closed
|
2023-06-28T14:04:54Z
|
2023-06-28T14:18:18Z
|
2023-06-28T14:18:17Z
|
AndreaFrancis
|
1,778,626,219 |
Stream convert to parquet
|
Allow to have datasets partially converted to parquet, like c4, refinedweb, oscar, etc.
Datasets above 5GB are streamed to generate 5GB (uncompressed) of parquet files.
## Implementation details
I implemented a context manager `limite_parquet_writes` that does some monkeypatching in the `datasets` lib to stop the dataset generation at the right time. This is not the kind of features that would be implemented in the `datasets` library natively, so I'm convinced this is the easiest way to do it. I added a test for this helper to make sure it works as expected and keeps working in the future.
I'm also using the new path scheme `config/split/partial/SSSS.parquet`. I didn't change the scheme for regular parquet conversion though - it can be done later.
close https://github.com/huggingface/datasets-server/issues/1257
|
Stream convert to parquet: Allow to have datasets partially converted to parquet, like c4, refinedweb, oscar, etc.
Datasets above 5GB are streamed to generate 5GB (uncompressed) of parquet files.
## Implementation details
I implemented a context manager `limite_parquet_writes` that does some monkeypatching in the `datasets` lib to stop the dataset generation at the right time. This is not the kind of features that would be implemented in the `datasets` library natively, so I'm convinced this is the easiest way to do it. I added a test for this helper to make sure it works as expected and keeps working in the future.
I'm also using the new path scheme `config/split/partial/SSSS.parquet`. I didn't change the scheme for regular parquet conversion though - it can be done later.
close https://github.com/huggingface/datasets-server/issues/1257
|
closed
|
2023-06-28T10:06:31Z
|
2023-07-03T15:42:26Z
|
2023-07-03T15:40:32Z
|
lhoestq
|
1,778,601,816 |
docs: ✏️ add docs for fields viewer and preview in /valid
| null |
docs: ✏️ add docs for fields viewer and preview in /valid:
|
closed
|
2023-06-28T09:52:14Z
|
2023-06-28T15:10:41Z
|
2023-06-28T15:10:09Z
|
severo
|
1,778,545,555 |
Add fields `viewer` and `preview` to /is-valid
|
For coherence with /valid, we should add the `viewer` and `preview` fields to /is-valid.
We should also consider deprecating the current `valid` field (as in https://github.com/huggingface/datasets-server/issues/1445). Note that it's in use in https://github.com/search?q=org%3Ahuggingface+datasets-server.huggingface.co+repo%3Ahuggingface%2Fnotebooks&type=code and also in the @lewtun's evaluator if I remember correctly.
|
Add fields `viewer` and `preview` to /is-valid: For coherence with /valid, we should add the `viewer` and `preview` fields to /is-valid.
We should also consider deprecating the current `valid` field (as in https://github.com/huggingface/datasets-server/issues/1445). Note that it's in use in https://github.com/search?q=org%3Ahuggingface+datasets-server.huggingface.co+repo%3Ahuggingface%2Fnotebooks&type=code and also in the @lewtun's evaluator if I remember correctly.
|
closed
|
2023-06-28T09:19:56Z
|
2023-06-29T14:13:16Z
|
2023-06-29T14:13:16Z
|
severo
|
1,778,541,141 |
Remove `.valid` from `/valid` endpoint?
|
We recently added to fields to `/valid`:
- `viewer`: all the datasets that have a valid dataset viewer
- `preview`: all the datasets that don't have a valid dataset viewer, but have a dataset preview
And the Hub does not use the original field `valid` anymore. We still fill it with the union of both sets.
Should we remove it, as it doubles the size of the response and increases the response time, with no benefit? cc @huggingface/datasets-server
Note that it's used in the notebooks (https://github.com/search?q=org%3Ahuggingface+datasets-server.huggingface.co+repo%3Ahuggingface%2Fnotebooks&type=code), for example, so it is a breaking change.
I would vote in favor of removing it, and updating the notebooks (and the docs obviously).
|
Remove `.valid` from `/valid` endpoint?: We recently added to fields to `/valid`:
- `viewer`: all the datasets that have a valid dataset viewer
- `preview`: all the datasets that don't have a valid dataset viewer, but have a dataset preview
And the Hub does not use the original field `valid` anymore. We still fill it with the union of both sets.
Should we remove it, as it doubles the size of the response and increases the response time, with no benefit? cc @huggingface/datasets-server
Note that it's used in the notebooks (https://github.com/search?q=org%3Ahuggingface+datasets-server.huggingface.co+repo%3Ahuggingface%2Fnotebooks&type=code), for example, so it is a breaking change.
I would vote in favor of removing it, and updating the notebooks (and the docs obviously).
|
closed
|
2023-06-28T09:17:13Z
|
2023-07-26T15:47:35Z
|
2023-07-26T15:47:35Z
|
severo
|
1,778,524,649 |
Remove the useless mongodb indexes
|
Review the current indexes in the mongodb collections, and ensure all of them are required. Else, remove the redundant ones. It will allow us to reduce the size on the server
|
Remove the useless mongodb indexes: Review the current indexes in the mongodb collections, and ensure all of them are required. Else, remove the redundant ones. It will allow us to reduce the size on the server
|
closed
|
2023-06-28T09:07:19Z
|
2023-08-16T21:23:20Z
|
2023-08-16T21:23:20Z
|
severo
|
1,778,462,851 |
Raise specific errors (and error_code) instead of UnexpectedError
|
The following query on the production database gives the number of datasets with at least one cache entry with error_code "UnexpectedError", grouped by the underlying "cause_exception".
For the most common ones (`DatasetGenerationError`, `HfHubHTTPError`, `OSError`, etc.) we would benefit from raising a specific error with its error code. It would allow to:
- retry automatically, if needed
- show an adequate error message to the users
- even: adapt the way we show the dataset viewer on the Hub
`null` means it has no `details.cause_exception`. These cache entries should be inspected more closely. See https://github.com/huggingface/datasets-server/issues/1123 in particular, which is one of the cases where no cause exception is reported.
```
db.cachedResponsesBlue.aggregate([
{$match: {error_code: "UnexpectedError"}},
{$group: {_id: {cause: "$details.cause_exception", dataset: "$dataset"}, count: {$sum: 1}}},
{$group: {_id: "$_id.cause", count: {$sum: 1}}},
{$sort: {count: -1}}
])
{ _id: 'DatasetGenerationError', count: 1964 }
{ _id: null, count: 1388 }
{ _id: 'HfHubHTTPError', count: 1154 }
{ _id: 'OSError', count: 433 }
{ _id: 'FileNotFoundError', count: 242 }
{ _id: 'FileExistsError', count: 198 }
{ _id: 'ValueError', count: 186 }
{ _id: 'TypeError', count: 160 }
{ _id: 'ConnectionError', count: 146 }
{ _id: 'RuntimeError', count: 86 }
{ _id: 'NonMatchingSplitsSizesError', count: 83 }
{ _id: 'FileSystemError', count: 62 }
{ _id: 'ClientResponseError', count: 52 }
{ _id: 'ArrowInvalid', count: 45 }
{ _id: 'ParquetResponseEmptyError', count: 43 }
{ _id: 'RepositoryNotFoundError', count: 41 }
{ _id: 'ManualDownloadError', count: 39 }
{ _id: 'IndexError', count: 28 }
{ _id: 'AttributeError', count: 16 }
{ _id: 'KeyError', count: 15 }
{ _id: 'GatedRepoError', count: 13 }
{ _id: 'NotImplementedError', count: 11 }
{ _id: 'ExpectedMoreSplits', count: 9 }
{ _id: 'PermissionError', count: 8 }
{ _id: 'BadRequestError', count: 7 }
{ _id: 'NonMatchingChecksumError', count: 6 }
{ _id: 'AssertionError', count: 4 }
{ _id: 'NameError', count: 4 }
{ _id: 'UnboundLocalError', count: 3 }
{ _id: 'JSONDecodeError', count: 3 }
{ _id: 'ZeroDivisionError', count: 3 }
{ _id: 'InvalidDocument', count: 3 }
{ _id: 'DoesNotExist', count: 3 }
{ _id: 'EOFError', count: 3 }
{ _id: 'ImportError', count: 3 }
{ _id: 'NotADirectoryError', count: 2 }
{ _id: 'RarCannotExec', count: 2 }
{ _id: 'ReadTimeout', count: 2 }
{ _id: 'ChunkedEncodingError', count: 2 }
{ _id: 'ExpectedMoreDownloadedFiles', count: 2 }
{ _id: 'InvalidConfigName', count: 2 }
{ _id: 'ModuleNotFoundError', count: 2 }
{ _id: 'Exception', count: 2 }
{ _id: 'MissingBeamOptions', count: 2 }
{ _id: 'HTTPError', count: 1 }
{ _id: 'BadZipFile', count: 1 }
{ _id: 'OverflowError', count: 1 }
{ _id: 'HFValidationError', count: 1 }
{ _id: 'IsADirectoryError', count: 1 }
{ _id: 'OperationalError', count: 1 }
```
|
Raise specific errors (and error_code) instead of UnexpectedError: The following query on the production database gives the number of datasets with at least one cache entry with error_code "UnexpectedError", grouped by the underlying "cause_exception".
For the most common ones (`DatasetGenerationError`, `HfHubHTTPError`, `OSError`, etc.) we would benefit from raising a specific error with its error code. It would allow to:
- retry automatically, if needed
- show an adequate error message to the users
- even: adapt the way we show the dataset viewer on the Hub
`null` means it has no `details.cause_exception`. These cache entries should be inspected more closely. See https://github.com/huggingface/datasets-server/issues/1123 in particular, which is one of the cases where no cause exception is reported.
```
db.cachedResponsesBlue.aggregate([
{$match: {error_code: "UnexpectedError"}},
{$group: {_id: {cause: "$details.cause_exception", dataset: "$dataset"}, count: {$sum: 1}}},
{$group: {_id: "$_id.cause", count: {$sum: 1}}},
{$sort: {count: -1}}
])
{ _id: 'DatasetGenerationError', count: 1964 }
{ _id: null, count: 1388 }
{ _id: 'HfHubHTTPError', count: 1154 }
{ _id: 'OSError', count: 433 }
{ _id: 'FileNotFoundError', count: 242 }
{ _id: 'FileExistsError', count: 198 }
{ _id: 'ValueError', count: 186 }
{ _id: 'TypeError', count: 160 }
{ _id: 'ConnectionError', count: 146 }
{ _id: 'RuntimeError', count: 86 }
{ _id: 'NonMatchingSplitsSizesError', count: 83 }
{ _id: 'FileSystemError', count: 62 }
{ _id: 'ClientResponseError', count: 52 }
{ _id: 'ArrowInvalid', count: 45 }
{ _id: 'ParquetResponseEmptyError', count: 43 }
{ _id: 'RepositoryNotFoundError', count: 41 }
{ _id: 'ManualDownloadError', count: 39 }
{ _id: 'IndexError', count: 28 }
{ _id: 'AttributeError', count: 16 }
{ _id: 'KeyError', count: 15 }
{ _id: 'GatedRepoError', count: 13 }
{ _id: 'NotImplementedError', count: 11 }
{ _id: 'ExpectedMoreSplits', count: 9 }
{ _id: 'PermissionError', count: 8 }
{ _id: 'BadRequestError', count: 7 }
{ _id: 'NonMatchingChecksumError', count: 6 }
{ _id: 'AssertionError', count: 4 }
{ _id: 'NameError', count: 4 }
{ _id: 'UnboundLocalError', count: 3 }
{ _id: 'JSONDecodeError', count: 3 }
{ _id: 'ZeroDivisionError', count: 3 }
{ _id: 'InvalidDocument', count: 3 }
{ _id: 'DoesNotExist', count: 3 }
{ _id: 'EOFError', count: 3 }
{ _id: 'ImportError', count: 3 }
{ _id: 'NotADirectoryError', count: 2 }
{ _id: 'RarCannotExec', count: 2 }
{ _id: 'ReadTimeout', count: 2 }
{ _id: 'ChunkedEncodingError', count: 2 }
{ _id: 'ExpectedMoreDownloadedFiles', count: 2 }
{ _id: 'InvalidConfigName', count: 2 }
{ _id: 'ModuleNotFoundError', count: 2 }
{ _id: 'Exception', count: 2 }
{ _id: 'MissingBeamOptions', count: 2 }
{ _id: 'HTTPError', count: 1 }
{ _id: 'BadZipFile', count: 1 }
{ _id: 'OverflowError', count: 1 }
{ _id: 'HFValidationError', count: 1 }
{ _id: 'IsADirectoryError', count: 1 }
{ _id: 'OperationalError', count: 1 }
```
|
open
|
2023-06-28T08:28:06Z
|
2024-08-01T11:11:21Z
| null |
severo
|
1,778,427,146 |
Add new dependencies for the job runners
|
Based on statistics about the most needed dependencies (https://github.com/huggingface/datasets-server/issues/1281#issuecomment-1609455781), we should prioritize adding [ir-datasets](https://pypi.org/project/ir-datasets/), [bioc](https://pypi.org/project/bioc/) and [pytorch_ie](https://pypi.org/project/pytorch-ie/).
|
Add new dependencies for the job runners: Based on statistics about the most needed dependencies (https://github.com/huggingface/datasets-server/issues/1281#issuecomment-1609455781), we should prioritize adding [ir-datasets](https://pypi.org/project/ir-datasets/), [bioc](https://pypi.org/project/bioc/) and [pytorch_ie](https://pypi.org/project/pytorch-ie/).
|
closed
|
2023-06-28T08:04:49Z
|
2024-02-02T17:19:24Z
|
2024-02-02T17:19:23Z
|
severo
|
1,777,655,160 |
Adding other processing steps
|
Previously, `split-duckdb-index` was triggered only by `config-split-names-from-info` but when this step finished with error 500 because of ResponseAlreadyComputedError, `split-duckdb-index` never started.
Adding other parents to avoid skipping job compute.
Note that the issue with parallel processing steps will remain and should be fixed on https://github.com/huggingface/datasets-server/issues/1358
|
Adding other processing steps: Previously, `split-duckdb-index` was triggered only by `config-split-names-from-info` but when this step finished with error 500 because of ResponseAlreadyComputedError, `split-duckdb-index` never started.
Adding other parents to avoid skipping job compute.
Note that the issue with parallel processing steps will remain and should be fixed on https://github.com/huggingface/datasets-server/issues/1358
|
closed
|
2023-06-27T19:51:03Z
|
2023-06-30T09:04:08Z
|
2023-06-27T20:01:10Z
|
AndreaFrancis
|
1,777,404,581 |
Try to fix Duckdb extensions
|
Error when computing split-duckdb-index
```
"details": {
"error": "IO Error: Extension \"//.duckdb/extensions/v0.8.1/linux_amd64_gcc4/httpfs.duckdb_extension\" not found.\nExtension \"httpfs\" is an existing extension.\n\nInstall it first using \"INSTALL httpfs\".",
"cause_exception": "IOException",
"cause_message": "IO Error: Extension \"//.duckdb/extensions/v0.8.1/linux_amd64_gcc4/httpfs.duckdb_extension\" not found.\nExtension \"httpfs\" is an existing extension.\n\nInstall it first using \"INSTALL httpfs\".",
"cause_traceback": [
"Traceback (most recent call last):\n",
" File \"/src/services/worker/src/worker/job_manager.py\", line 160, in process\n job_result = self.job_runner.compute()\n",
" File \"/src/services/worker/src/worker/job_runners/split/duckdb_index.py\", line 256, in compute\n compute_index_rows(\n",
" File \"/src/services/worker/src/worker/job_runners/split/duckdb_index.py\", line 148, in compute_index_rows\n con.sql(create_command_sql)\n",
"duckdb.IOException: IO Error: Extension \"//.duckdb/extensions/v0.8.1/linux_amd64_gcc4/httpfs.duckdb_extension\" not found.\nExtension \"httpfs\" is an existing extension.\n\nInstall it first using \"INSTALL httpfs\".\n"
]
},
```
|
Try to fix Duckdb extensions: Error when computing split-duckdb-index
```
"details": {
"error": "IO Error: Extension \"//.duckdb/extensions/v0.8.1/linux_amd64_gcc4/httpfs.duckdb_extension\" not found.\nExtension \"httpfs\" is an existing extension.\n\nInstall it first using \"INSTALL httpfs\".",
"cause_exception": "IOException",
"cause_message": "IO Error: Extension \"//.duckdb/extensions/v0.8.1/linux_amd64_gcc4/httpfs.duckdb_extension\" not found.\nExtension \"httpfs\" is an existing extension.\n\nInstall it first using \"INSTALL httpfs\".",
"cause_traceback": [
"Traceback (most recent call last):\n",
" File \"/src/services/worker/src/worker/job_manager.py\", line 160, in process\n job_result = self.job_runner.compute()\n",
" File \"/src/services/worker/src/worker/job_runners/split/duckdb_index.py\", line 256, in compute\n compute_index_rows(\n",
" File \"/src/services/worker/src/worker/job_runners/split/duckdb_index.py\", line 148, in compute_index_rows\n con.sql(create_command_sql)\n",
"duckdb.IOException: IO Error: Extension \"//.duckdb/extensions/v0.8.1/linux_amd64_gcc4/httpfs.duckdb_extension\" not found.\nExtension \"httpfs\" is an existing extension.\n\nInstall it first using \"INSTALL httpfs\".\n"
]
},
```
|
closed
|
2023-06-27T17:08:56Z
|
2023-06-28T12:04:29Z
|
2023-06-27T19:12:46Z
|
AndreaFrancis
|
1,777,335,026 |
Set Duckdb extensions install directory
|
Error `duckdb.IOException: IO Error: Failed to create directory \"//.duckdb\"!\n `
is shown when computing duckdb index.
|
Set Duckdb extensions install directory: Error `duckdb.IOException: IO Error: Failed to create directory \"//.duckdb\"!\n `
is shown when computing duckdb index.
|
closed
|
2023-06-27T16:24:25Z
|
2023-06-27T16:45:38Z
|
2023-06-27T16:45:36Z
|
AndreaFrancis
|
1,777,242,709 |
Increase chart version
| null |
Increase chart version:
|
closed
|
2023-06-27T15:30:39Z
|
2023-06-27T15:31:42Z
|
2023-06-27T15:31:41Z
|
AndreaFrancis
|
1,777,018,800 |
Use specific stemmer by dataset according to the language
|
Currently, '`porter`' stemmer is used by default for duckdb indexing here https://github.com/huggingface/datasets-server/pull/1296/files#diff-d9a2c828d7feca3b7f9e332e040ef861e842a16d18276b356461d2aa34396a8aR145
See https://duckdb.org/docs/extensions/full_text_search.html for more details about '`stemmer`' parameter.
In the future, we could try to identify the dataset language and use an appropriate stemmer parameter when creating the `fts` index.
|
Use specific stemmer by dataset according to the language: Currently, '`porter`' stemmer is used by default for duckdb indexing here https://github.com/huggingface/datasets-server/pull/1296/files#diff-d9a2c828d7feca3b7f9e332e040ef861e842a16d18276b356461d2aa34396a8aR145
See https://duckdb.org/docs/extensions/full_text_search.html for more details about '`stemmer`' parameter.
In the future, we could try to identify the dataset language and use an appropriate stemmer parameter when creating the `fts` index.
|
open
|
2023-06-27T13:46:44Z
|
2024-08-22T00:45:07Z
| null |
AndreaFrancis
|
1,777,002,036 |
Prevent using cache_subdirectory=None on JobRunnerWithCache's children
|
Currently, all job runners that depend on `JobRunnerWithCache` and use the `cache_subdirectory` field, need to do validation before using the generated value like here https://github.com/huggingface/datasets-server/pull/1296/files#diff-d9a2c828d7feca3b7f9e332e040ef861e842a16d18276b356461d2aa34396a8aR248
We need a better way to prevent using `None` values given that `JobRunnerWithCache` has https://github.com/huggingface/datasets-server/blob/main/services/worker/src/worker/job_runners/_job_runner_with_cache.py#L23 `cache_subdirectory: Optional[Path]`.
If the `pre_compute` method is not run, job runner children could fail because of non-existing directory.
@severo suggested adding a test on each job runner implementation of `JobRunnerWithCache` to ensure the validation is done.
|
Prevent using cache_subdirectory=None on JobRunnerWithCache's children : Currently, all job runners that depend on `JobRunnerWithCache` and use the `cache_subdirectory` field, need to do validation before using the generated value like here https://github.com/huggingface/datasets-server/pull/1296/files#diff-d9a2c828d7feca3b7f9e332e040ef861e842a16d18276b356461d2aa34396a8aR248
We need a better way to prevent using `None` values given that `JobRunnerWithCache` has https://github.com/huggingface/datasets-server/blob/main/services/worker/src/worker/job_runners/_job_runner_with_cache.py#L23 `cache_subdirectory: Optional[Path]`.
If the `pre_compute` method is not run, job runner children could fail because of non-existing directory.
@severo suggested adding a test on each job runner implementation of `JobRunnerWithCache` to ensure the validation is done.
|
open
|
2023-06-27T13:39:56Z
|
2023-08-07T16:35:04Z
| null |
AndreaFrancis
|
1,776,497,349 |
refactor: 💡 remove dead code
| null |
refactor: 💡 remove dead code:
|
closed
|
2023-06-27T09:38:35Z
|
2023-06-27T13:04:02Z
|
2023-06-27T13:04:01Z
|
severo
|
1,776,475,451 |
Unquote path and revision when copying parquet files
|
close https://github.com/huggingface/datasets-server/issues/1433
|
Unquote path and revision when copying parquet files: close https://github.com/huggingface/datasets-server/issues/1433
|
closed
|
2023-06-27T09:26:04Z
|
2023-06-27T12:12:21Z
|
2023-06-27T12:12:20Z
|
lhoestq
|
1,776,453,056 |
Can't copy parquet files with path that can be URL encoded
|
we get this error for bigcode/the-stack in config-parquet-and-info
```
huggingface_hub.utils._errors.EntryNotFoundError: Cannot copy data/c%2B%2B/train-00000-of-00214.parquet at revision 349a71353fd5868fb90b593ef09e311379da498a: file is missing on repo.
```
|
Can't copy parquet files with path that can be URL encoded: we get this error for bigcode/the-stack in config-parquet-and-info
```
huggingface_hub.utils._errors.EntryNotFoundError: Cannot copy data/c%2B%2B/train-00000-of-00214.parquet at revision 349a71353fd5868fb90b593ef09e311379da498a: file is missing on repo.
```
|
closed
|
2023-06-27T09:13:06Z
|
2023-06-27T12:12:21Z
|
2023-06-27T12:12:21Z
|
lhoestq
|
1,775,140,577 |
feat: 🎸 don't insert a new lock when releasing
| null |
feat: 🎸 don't insert a new lock when releasing:
|
closed
|
2023-06-26T16:16:57Z
|
2023-06-26T16:17:04Z
|
2023-06-26T16:17:03Z
|
severo
|
1,775,102,927 |
fix: 🐛 remove the "required" constraint on created_at in Lock
|
the code does not rely on always having a created_at field. And it's not that easy to always ensure it's filled (see `update(upsert=True, ...`)
|
fix: 🐛 remove the "required" constraint on created_at in Lock: the code does not rely on always having a created_at field. And it's not that easy to always ensure it's filled (see `update(upsert=True, ...`)
|
closed
|
2023-06-26T15:56:40Z
|
2023-06-26T15:56:47Z
|
2023-06-26T15:56:46Z
|
severo
|
1,775,088,105 |
Fix auth in rows again
|
for real this time
including a test
|
Fix auth in rows again: for real this time
including a test
|
closed
|
2023-06-26T15:47:13Z
|
2023-06-26T17:23:14Z
|
2023-06-26T17:23:13Z
|
lhoestq
|
1,774,326,535 |
Ignore duckdb files in parquet and info
|
needed for https://github.com/huggingface/datasets-server/pull/1296
see https://github.com/huggingface/datasets-server/pull/1296#issuecomment-1604502957
|
Ignore duckdb files in parquet and info: needed for https://github.com/huggingface/datasets-server/pull/1296
see https://github.com/huggingface/datasets-server/pull/1296#issuecomment-1604502957
|
closed
|
2023-06-26T09:22:03Z
|
2023-06-26T10:40:26Z
|
2023-06-26T10:40:25Z
|
lhoestq
|
1,772,942,026 |
Rename classes to indicate inheritance
|
This commit adds `Document` suffix to classes to indicate inheritance.
cc @severo
Fixes https://github.com/huggingface/datasets-server/issues/1359
|
Rename classes to indicate inheritance: This commit adds `Document` suffix to classes to indicate inheritance.
cc @severo
Fixes https://github.com/huggingface/datasets-server/issues/1359
|
closed
|
2023-06-24T22:06:58Z
|
2023-07-01T16:20:40Z
|
2023-07-01T15:49:33Z
|
geethika-123
|
1,771,743,801 |
Add auth in rows
|
fix
```
aiohttp.client_exceptions.ClientResponseError: 401, message='Unauthorized', url=URL('https://huggingface.co/datasets/bigcode/the-stack-dedup/resolve/refs%2Fconvert%2Fparquet/bigcode--the-stack-dedup/parquet-train-00000-of-05140.parquet')
```
when doing pagination on the bigcode/the-stack-dedup which is gated
|
Add auth in rows: fix
```
aiohttp.client_exceptions.ClientResponseError: 401, message='Unauthorized', url=URL('https://huggingface.co/datasets/bigcode/the-stack-dedup/resolve/refs%2Fconvert%2Fparquet/bigcode--the-stack-dedup/parquet-train-00000-of-05140.parquet')
```
when doing pagination on the bigcode/the-stack-dedup which is gated
|
closed
|
2023-06-23T16:37:55Z
|
2023-06-26T08:21:51Z
|
2023-06-26T08:21:50Z
|
lhoestq
|
1,771,726,572 |
Move dtos to its own file
|
Currently, we have data transfer objects in the `utils.py` file, but this one is growing and growing and sometime we could have duplicated code if don't verify the existing models.
This PR just moves all the abstractions of responses to a `dtos.py` file.
Most of the files are changes on the `imports` definition.
|
Move dtos to its own file: Currently, we have data transfer objects in the `utils.py` file, but this one is growing and growing and sometime we could have duplicated code if don't verify the existing models.
This PR just moves all the abstractions of responses to a `dtos.py` file.
Most of the files are changes on the `imports` definition.
|
closed
|
2023-06-23T16:22:40Z
|
2023-06-26T11:04:14Z
|
2023-06-26T11:04:12Z
|
AndreaFrancis
|
1,771,645,506 |
Create missing Jobs when /rows cache does not exists yet
|
Should fix https://github.com/huggingface/datasets-server/issues/1341
|
Create missing Jobs when /rows cache does not exists yet: Should fix https://github.com/huggingface/datasets-server/issues/1341
|
closed
|
2023-06-23T15:22:59Z
|
2023-06-26T11:01:37Z
|
2023-06-26T11:01:36Z
|
AndreaFrancis
|
1,771,454,880 |
Fix regression: use parquet metadata when possible
|
I noticed some datasets have a slow pagination like https://huggingface.co/datasets/mlfoundations/datacomp_1b that times out.
This is because there was a regression in https://github.com/huggingface/datasets-server/pull/1287 where it wouldn't use the parquet metadata because `get_best_response` returns the first successful response
|
Fix regression: use parquet metadata when possible: I noticed some datasets have a slow pagination like https://huggingface.co/datasets/mlfoundations/datacomp_1b that times out.
This is because there was a regression in https://github.com/huggingface/datasets-server/pull/1287 where it wouldn't use the parquet metadata because `get_best_response` returns the first successful response
|
closed
|
2023-06-23T13:20:08Z
|
2023-06-27T08:56:57Z
|
2023-06-23T16:01:53Z
|
lhoestq
|
1,771,416,315 |
Remove too restrictive __all__ definitions
|
Remove too restrictive `__all__` from `libcommon/simple_cache`:
- it contained only one attribute: `DoesNotExist`
- Python considers as private all the module attributes which are not defined in `__all__`
```
from libcommon.simple_cache import upsert_response
Warning: Accessing a protected member of a class or a module:
'upsert_response' is not declared in __all__
```
|
Remove too restrictive __all__ definitions: Remove too restrictive `__all__` from `libcommon/simple_cache`:
- it contained only one attribute: `DoesNotExist`
- Python considers as private all the module attributes which are not defined in `__all__`
```
from libcommon.simple_cache import upsert_response
Warning: Accessing a protected member of a class or a module:
'upsert_response' is not declared in __all__
```
|
closed
|
2023-06-23T12:56:41Z
|
2023-06-26T13:01:27Z
|
2023-06-26T13:01:25Z
|
albertvillanova
|
1,771,026,707 |
Update to datasets 2.13.1
|
This should fix the parquet-and-info job for bigcode/the-stack-dedup.
The patch release includes a fix that makes it ignore non data files (in the case of bigcode/the-stack-dedup there's a license.json file that shouldn't be taken into account in the data of course)
|
Update to datasets 2.13.1: This should fix the parquet-and-info job for bigcode/the-stack-dedup.
The patch release includes a fix that makes it ignore non data files (in the case of bigcode/the-stack-dedup there's a license.json file that shouldn't be taken into account in the data of course)
|
closed
|
2023-06-23T08:15:55Z
|
2023-06-23T12:02:32Z
|
2023-06-23T12:02:30Z
|
lhoestq
|
1,769,777,048 |
/rows returns `null` images for some datasets
|
Reported in https://github.com/huggingface/datasets/issues/2526
See https://huggingface.co/datasets/lombardata/panoptic_2023_06_22 has a working /first-rows but /rows always returns `null` for images.
edit: another one https://huggingface.co/datasets/jonathan-roberts1/RSSCN7
|
/rows returns `null` images for some datasets: Reported in https://github.com/huggingface/datasets/issues/2526
See https://huggingface.co/datasets/lombardata/panoptic_2023_06_22 has a working /first-rows but /rows always returns `null` for images.
edit: another one https://huggingface.co/datasets/jonathan-roberts1/RSSCN7
|
closed
|
2023-06-22T14:13:39Z
|
2023-07-28T12:42:23Z
|
2023-07-28T12:42:23Z
|
lhoestq
|
1,769,268,896 |
Ensure only one job is started for the same unicity_id
|
To avoid multiple job runners getting the same job at the same time, for a given unicity_id (identifies job type + parameters):
- a lock is used during the update of the selected job
- we ensure no other job is already started
- we select the newest (in date order) job from all the waiting jobs and start it (status, started_at)
- we cancel all the other waiting jobs
Should fix #1323.
|
Ensure only one job is started for the same unicity_id: To avoid multiple job runners getting the same job at the same time, for a given unicity_id (identifies job type + parameters):
- a lock is used during the update of the selected job
- we ensure no other job is already started
- we select the newest (in date order) job from all the waiting jobs and start it (status, started_at)
- we cancel all the other waiting jobs
Should fix #1323.
|
closed
|
2023-06-22T09:10:26Z
|
2023-06-26T15:37:28Z
|
2023-06-26T15:37:27Z
|
severo
|
1,769,120,320 |
Use external-secrets to read secrets from AWS
| null |
Use external-secrets to read secrets from AWS:
|
closed
|
2023-06-22T07:37:10Z
|
2023-06-22T07:51:08Z
|
2023-06-22T07:51:07Z
|
rtrompier
|
1,769,074,534 |
Create /filter endpoint
|
Create /filter endpoint.
I have tried to follow roughly the same logic as the /rows endpoint.
TODO:
- [x] e2e tests
- [x] chart: only increase the number of replicas
- ~~docker-compose files~~
- [x] openapi specification
- [x] documentation pages: draft of `filter.mdx'
Subsequent PRs:
- Index all datasets, not only those with text features. See:
- #1854
- Complete the `filter.mdx` docs page
- Give examples for all the supported operators
- Enumerate all the supported column data types
- More complex validation of request parameters (to avoid SQL injection)
- Maybe rename the `search` service to `query`, with the 2 endpoints `search` and `filter`
|
Create /filter endpoint: Create /filter endpoint.
I have tried to follow roughly the same logic as the /rows endpoint.
TODO:
- [x] e2e tests
- [x] chart: only increase the number of replicas
- ~~docker-compose files~~
- [x] openapi specification
- [x] documentation pages: draft of `filter.mdx'
Subsequent PRs:
- Index all datasets, not only those with text features. See:
- #1854
- Complete the `filter.mdx` docs page
- Give examples for all the supported operators
- Enumerate all the supported column data types
- More complex validation of request parameters (to avoid SQL injection)
- Maybe rename the `search` service to `query`, with the 2 endpoints `search` and `filter`
|
closed
|
2023-06-22T07:10:03Z
|
2023-10-05T06:49:47Z
|
2023-10-05T06:49:09Z
|
albertvillanova
|
1,768,066,940 |
keep image and audio untruncated
|
close https://github.com/huggingface/datasets-server/issues/1416
|
keep image and audio untruncated: close https://github.com/huggingface/datasets-server/issues/1416
|
closed
|
2023-06-21T17:15:27Z
|
2023-06-22T12:51:41Z
|
2023-06-22T12:51:40Z
|
lhoestq
|
1,768,045,641 |
Truncated first-rows may crop image URLs
|
see https://huggingface.co/datasets/Antreas/TALI-base
because of that the images are not show in the UI
|
Truncated first-rows may crop image URLs: see https://huggingface.co/datasets/Antreas/TALI-base
because of that the images are not show in the UI
|
closed
|
2023-06-21T17:05:29Z
|
2023-06-22T12:51:41Z
|
2023-06-22T12:51:41Z
|
lhoestq
|
1,767,726,199 |
Test concurrency in parquet and info
|
It adds a test on the concurrency of job runners on the step `config-parquet-and-info`(which creates `refs/convert/parquet` and uploads parquet files), to detect when the lock is not respected, leading to `CreateCommitError`.
Also:
- group all the code that accesses the Hub inside the lock (create the "branch", send the commit, get the list of files)
- retry "hf_api.list_repo_refs" on connection error (I saw it during the tests)
- adapt the sleep intervals
---
First: tests should fail (https://github.com/huggingface/datasets-server/actions/runs/5335283399/jobs/9668277338)
```
FAILED tests/job_runners/config/test_parquet_and_info.py::test_concurrency - huggingface_hub.utils._errors.HfHubHTTPError: 500 Server Error: Internal Server Error for url: https://hub-ci.huggingface.co/api/datasets/__DUMMY_DATASETS_SERVER_USER__/n_configs-16873575569236/branch/refs%2Fconvert%2Fparquet (Request ID: Root=1-64930878-4a2840cd4fc36b1752b0eac0)
```
Then, after rebasing after merging https://github.com/huggingface/datasets-server/pull/1414 into main, the test should pass (https://github.com/huggingface/datasets-server/actions/runs/5335527787/jobs/9669124731)
|
Test concurrency in parquet and info: It adds a test on the concurrency of job runners on the step `config-parquet-and-info`(which creates `refs/convert/parquet` and uploads parquet files), to detect when the lock is not respected, leading to `CreateCommitError`.
Also:
- group all the code that accesses the Hub inside the lock (create the "branch", send the commit, get the list of files)
- retry "hf_api.list_repo_refs" on connection error (I saw it during the tests)
- adapt the sleep intervals
---
First: tests should fail (https://github.com/huggingface/datasets-server/actions/runs/5335283399/jobs/9668277338)
```
FAILED tests/job_runners/config/test_parquet_and_info.py::test_concurrency - huggingface_hub.utils._errors.HfHubHTTPError: 500 Server Error: Internal Server Error for url: https://hub-ci.huggingface.co/api/datasets/__DUMMY_DATASETS_SERVER_USER__/n_configs-16873575569236/branch/refs%2Fconvert%2Fparquet (Request ID: Root=1-64930878-4a2840cd4fc36b1752b0eac0)
```
Then, after rebasing after merging https://github.com/huggingface/datasets-server/pull/1414 into main, the test should pass (https://github.com/huggingface/datasets-server/actions/runs/5335527787/jobs/9669124731)
|
closed
|
2023-06-21T14:22:15Z
|
2023-06-21T15:06:16Z
|
2023-06-21T15:06:15Z
|
severo
|
1,767,622,888 |
Fix the lock
|
Several processes were able to acquire the lock at the same time
|
Fix the lock: Several processes were able to acquire the lock at the same time
|
closed
|
2023-06-21T13:36:15Z
|
2023-06-21T14:25:16Z
|
2023-06-21T14:25:15Z
|
severo
|
1,767,394,602 |
The e2e tests have implicit dependencies
|
As reported by @albertvillanova for example, the following test does not pass
```
$ TEST_PATH=tests/test_11_api.py::test_rows_endpoint make test
```
while this ones pass:
```
$ TEST_PATH="tests/test_11_api.py::test_endpoint tests/test_11_api.py::test_rows_endpoint" make test
```
It's because `test_rows_endpoint` requires the parquet files to exist, and we implicitly rely on the fact that they have been created in `tests/test_11_api.py::test_endpoint`.
Other implicit dependency is when we test `is-valid` and `valid`.
|
The e2e tests have implicit dependencies: As reported by @albertvillanova for example, the following test does not pass
```
$ TEST_PATH=tests/test_11_api.py::test_rows_endpoint make test
```
while this ones pass:
```
$ TEST_PATH="tests/test_11_api.py::test_endpoint tests/test_11_api.py::test_rows_endpoint" make test
```
It's because `test_rows_endpoint` requires the parquet files to exist, and we implicitly rely on the fact that they have been created in `tests/test_11_api.py::test_endpoint`.
Other implicit dependency is when we test `is-valid` and `valid`.
|
open
|
2023-06-21T11:38:20Z
|
2023-08-15T15:13:24Z
| null |
severo
|
1,767,253,537 |
unblock DFKI-SLT/few-nerd
| null |
unblock DFKI-SLT/few-nerd:
|
closed
|
2023-06-21T10:12:26Z
|
2023-06-21T14:21:58Z
|
2023-06-21T14:21:57Z
|
lhoestq
|
1,767,202,799 |
Split-names-from-streaming is incorrect
|
e.g. only returns ["test"] for [Antreas/TALI-base](https://huggingface.co/datasets/Antreas/TALI-base) instead of ['train', 'test', 'val']
|
Split-names-from-streaming is incorrect: e.g. only returns ["test"] for [Antreas/TALI-base](https://huggingface.co/datasets/Antreas/TALI-base) instead of ['train', 'test', 'val']
|
closed
|
2023-06-21T09:44:29Z
|
2023-06-21T09:50:13Z
|
2023-06-21T09:50:13Z
|
lhoestq
|
1,766,207,317 |
fix: 🐛 try to ensure only one job can get the lock
|
As it's difficult to test on multiple MongoDB replicas (we should do it at one point), I'll try to:
- merge to main
- deploy on prod
- refresh the dataset severo/flores_101
- see if all the configs could be processed for the parquet creation, or if we still have an error for "half" of them
|
fix: 🐛 try to ensure only one job can get the lock:
As it's difficult to test on multiple MongoDB replicas (we should do it at one point), I'll try to:
- merge to main
- deploy on prod
- refresh the dataset severo/flores_101
- see if all the configs could be processed for the parquet creation, or if we still have an error for "half" of them
|
closed
|
2023-06-20T21:20:23Z
|
2023-06-20T21:36:30Z
|
2023-06-20T21:26:59Z
|
severo
|
1,766,092,638 |
test: 💍 add a test on lock.git_branch
| null |
test: 💍 add a test on lock.git_branch:
|
closed
|
2023-06-20T20:19:16Z
|
2023-06-20T20:35:13Z
|
2023-06-20T20:35:11Z
|
severo
|
1,765,983,796 |
Add tests on create_commits
|
see #1396
|
Add tests on create_commits: see #1396
|
closed
|
2023-06-20T18:59:08Z
|
2023-06-20T19:08:41Z
|
2023-06-20T19:08:40Z
|
severo
|
1,765,621,130 |
Use EFS instead of NFS for datasets and parquet "local" cache
|
(and duckdb local cache in the PRs)
related to https://github.com/huggingface/datasets-server/issues/1072
internal: see https://github.com/huggingface/infra/issues/605#issue-1758616648
|
Use EFS instead of NFS for datasets and parquet "local" cache: (and duckdb local cache in the PRs)
related to https://github.com/huggingface/datasets-server/issues/1072
internal: see https://github.com/huggingface/infra/issues/605#issue-1758616648
|
closed
|
2023-06-20T15:12:02Z
|
2023-08-11T13:48:34Z
|
2023-08-11T13:48:34Z
|
severo
|
1,765,618,183 |
Use S3 + cloudfront for assets and cached-assets
|
related to #1072
internal: see https://github.com/huggingface/infra/issues/605#issue-1758616648
|
Use S3 + cloudfront for assets and cached-assets: related to #1072
internal: see https://github.com/huggingface/infra/issues/605#issue-1758616648
|
closed
|
2023-06-20T15:10:24Z
|
2023-10-09T17:54:03Z
|
2023-10-09T17:54:03Z
|
severo
|
1,765,617,336 |
Modify TTL index condition
|
Adding a condition TTL index in Job document.
It will only delete those records with a final state of SUCCESS, ERROR or CANCELLED
|
Modify TTL index condition: Adding a condition TTL index in Job document.
It will only delete those records with a final state of SUCCESS, ERROR or CANCELLED
|
closed
|
2023-06-20T15:09:53Z
|
2023-06-20T15:20:00Z
|
2023-06-20T15:19:59Z
|
AndreaFrancis
|
1,765,596,874 |
Retry get_parquet_file_and_size
|
In prod I got an ArrowInvalid when instantiating a pq.ParquetFile for bigcode/the-stack-dedup even though all the parquet files are valid (I ran a script and checked I could get all the pq.ParquetFile objects)
|
Retry get_parquet_file_and_size: In prod I got an ArrowInvalid when instantiating a pq.ParquetFile for bigcode/the-stack-dedup even though all the parquet files are valid (I ran a script and checked I could get all the pq.ParquetFile objects)
|
closed
|
2023-06-20T14:59:10Z
|
2023-06-21T09:54:13Z
|
2023-06-21T09:54:12Z
|
lhoestq
|
1,765,568,744 |
Temporaly remove TTL index
|
Like https://github.com/huggingface/datasets-server/commit/47ea65b2567db4482579cd7000393cf0a15b412e , firs step to modify TTL index is to remove it from the code, then a deploy is needed.
|
Temporaly remove TTL index: Like https://github.com/huggingface/datasets-server/commit/47ea65b2567db4482579cd7000393cf0a15b412e , firs step to modify TTL index is to remove it from the code, then a deploy is needed.
|
closed
|
2023-06-20T14:46:05Z
|
2023-06-20T14:59:09Z
|
2023-06-20T14:59:07Z
|
AndreaFrancis
|
1,765,455,044 |
Update docs with hub parquet endpoint
|
Wait for (internal)https://github.com/huggingface/moon-landing/pull/6695 to be merged and deployed
close https://github.com/huggingface/datasets-server/issues/1400
|
Update docs with hub parquet endpoint: Wait for (internal)https://github.com/huggingface/moon-landing/pull/6695 to be merged and deployed
close https://github.com/huggingface/datasets-server/issues/1400
|
closed
|
2023-06-20T13:49:22Z
|
2023-07-18T15:54:08Z
|
2023-07-18T15:53:36Z
|
lhoestq
|
1,765,413,674 |
fix: 🐛 split the default value to get a list of strings
| null |
fix: 🐛 split the default value to get a list of strings:
|
closed
|
2023-06-20T13:28:14Z
|
2023-06-20T13:28:28Z
|
2023-06-20T13:28:26Z
|
severo
|
1,765,400,845 |
Update docs for hf.co/api/datasets/<dataset>/parquet endpoint
|
To be used instead of the datasets server /parquet endpoint in examples
following https://github.com/huggingface/moon-landing/pull/6695
|
Update docs for hf.co/api/datasets/<dataset>/parquet endpoint: To be used instead of the datasets server /parquet endpoint in examples
following https://github.com/huggingface/moon-landing/pull/6695
|
closed
|
2023-06-20T13:21:04Z
|
2023-07-19T12:02:37Z
|
2023-07-19T12:02:37Z
|
lhoestq
|
1,765,398,521 |
admin-UI stuck for datasets with many configs/splits
|
For example, https://huggingface.co/spaces/datasets-maintainers/datasets-server-admin-ui with `gsarti/flores_101` on "Dataset status" tab takes a lot of time.
|
admin-UI stuck for datasets with many configs/splits: For example, https://huggingface.co/spaces/datasets-maintainers/datasets-server-admin-ui with `gsarti/flores_101` on "Dataset status" tab takes a lot of time.
|
closed
|
2023-06-20T13:19:48Z
|
2024-02-06T14:39:15Z
|
2024-02-06T14:39:14Z
|
severo
|
1,765,309,469 |
Rename parent job runners
|
Based on the new job runner for those that need a cached directory on https://github.com/huggingface/datasets-server/pull/1388
We will need a new split job runner that inherits from the new JobRunnerWithCache to be used as part of https://github.com/huggingface/datasets-server/pull/1199 and https://github.com/huggingface/datasets-server/pull/1296/files.
According to the discussions here, I renamed the parent job runners to:
> CacheDirectoryJobRunner -> JobRunnerWithCache
> DatasetsBasedJobRunner -> JobRunnerWithDatasetsCache
And their children with Dataset/Config/Split + JobRunnerWithCache/JobRunnerWithDatasetsCache
|
Rename parent job runners: Based on the new job runner for those that need a cached directory on https://github.com/huggingface/datasets-server/pull/1388
We will need a new split job runner that inherits from the new JobRunnerWithCache to be used as part of https://github.com/huggingface/datasets-server/pull/1199 and https://github.com/huggingface/datasets-server/pull/1296/files.
According to the discussions here, I renamed the parent job runners to:
> CacheDirectoryJobRunner -> JobRunnerWithCache
> DatasetsBasedJobRunner -> JobRunnerWithDatasetsCache
And their children with Dataset/Config/Split + JobRunnerWithCache/JobRunnerWithDatasetsCache
|
closed
|
2023-06-20T12:31:31Z
|
2023-06-20T18:43:09Z
|
2023-06-20T18:43:08Z
|
AndreaFrancis
|
1,765,303,711 |
Ensure the parquet files in /parquet are sorted by "shard" index
|
And tell it in the docs
|
Ensure the parquet files in /parquet are sorted by "shard" index: And tell it in the docs
|
closed
|
2023-06-20T12:28:36Z
|
2023-06-30T20:02:14Z
|
2023-06-30T20:02:14Z
|
severo
|
1,765,292,247 |
Avoid commit conflicts
|
See https://github.com/huggingface/datasets-server/issues/1163#issuecomment-1598504866 and following comments
- [x] add tests to check if `parent_commit=parent_commit if not commit_infos else commit_infos[-1].oid` is correct. Yes; see https://github.com/huggingface/datasets-server/pull/1408
- [x] add write/read concern for the lock requests to ensure consistency - the [default](https://www.mongodb.com/docs/manual/reference/write-concern/#implicit-default-write-concern) is already write_concern = "majority", so we shouldn't need to change this.
- [ ] ???
|
Avoid commit conflicts: See https://github.com/huggingface/datasets-server/issues/1163#issuecomment-1598504866 and following comments
- [x] add tests to check if `parent_commit=parent_commit if not commit_infos else commit_infos[-1].oid` is correct. Yes; see https://github.com/huggingface/datasets-server/pull/1408
- [x] add write/read concern for the lock requests to ensure consistency - the [default](https://www.mongodb.com/docs/manual/reference/write-concern/#implicit-default-write-concern) is already write_concern = "majority", so we shouldn't need to change this.
- [ ] ???
|
closed
|
2023-06-20T12:23:06Z
|
2023-06-21T15:04:04Z
|
2023-06-21T15:04:04Z
|
severo
|
1,765,084,367 |
Increase max job duration
|
Currently bigcode/the-stack-dedup seems to take more than 20min to copy the parquet files.
This is mostly to test that it works - we can decide later if we keep this value or if we need to make this value depend on the job.
|
Increase max job duration: Currently bigcode/the-stack-dedup seems to take more than 20min to copy the parquet files.
This is mostly to test that it works - we can decide later if we keep this value or if we need to make this value depend on the job.
|
closed
|
2023-06-20T10:11:37Z
|
2023-06-20T13:52:23Z
|
2023-06-20T13:52:22Z
|
lhoestq
|
1,765,010,399 |
Remove unused code from /rows API endpoint
|
Remove unused code from /rows API endpoint.
|
Remove unused code from /rows API endpoint: Remove unused code from /rows API endpoint.
|
closed
|
2023-06-20T09:26:25Z
|
2023-06-21T14:22:46Z
|
2023-06-21T14:22:44Z
|
albertvillanova
|
1,764,960,588 |
Raise retryable error on hfhubhttperror
|
see https://github.com/huggingface/datasets-server/issues/1163
|
Raise retryable error on hfhubhttperror: see https://github.com/huggingface/datasets-server/issues/1163
|
closed
|
2023-06-20T08:58:43Z
|
2023-06-20T12:36:46Z
|
2023-06-20T12:36:45Z
|
severo
|
1,763,437,683 |
feat: 🎸 10x the size of supported images
| null |
feat: 🎸 10x the size of supported images:
|
closed
|
2023-06-19T12:27:34Z
|
2023-06-19T12:36:01Z
|
2023-06-19T12:36:00Z
|
severo
|
1,762,949,715 |
Fix typo in error message
|
I already suggested this typo fix:
- https://github.com/huggingface/datasets-server/pull/1371/files#r1230650727
while reviewing PR:
- #1371
And normally it was taken into account with commit:
- https://github.com/huggingface/datasets-server/pull/1371/commits/1c55c9e6d55e178062eb6b85c33e7c6e71dc13ec
However a subsequent force-push removed this commit:
- https://github.com/huggingface/datasets-server/compare/1c55c9e6d55e178062eb6b85c33e7c6e71dc13ec..d29012b12274ea9d58d3a6c3cdbb2dd64097f202
|
Fix typo in error message: I already suggested this typo fix:
- https://github.com/huggingface/datasets-server/pull/1371/files#r1230650727
while reviewing PR:
- #1371
And normally it was taken into account with commit:
- https://github.com/huggingface/datasets-server/pull/1371/commits/1c55c9e6d55e178062eb6b85c33e7c6e71dc13ec
However a subsequent force-push removed this commit:
- https://github.com/huggingface/datasets-server/compare/1c55c9e6d55e178062eb6b85c33e7c6e71dc13ec..d29012b12274ea9d58d3a6c3cdbb2dd64097f202
|
closed
|
2023-06-19T07:45:23Z
|
2023-06-19T08:55:14Z
|
2023-06-19T08:55:12Z
|
albertvillanova
|
1,762,411,257 |
Add Docker internal to extra_hosts
|
This is required to connect to the local DB instance on Linux; it is already added to `tools/docker-compose-dev-datasets-server.yml`
|
Add Docker internal to extra_hosts: This is required to connect to the local DB instance on Linux; it is already added to `tools/docker-compose-dev-datasets-server.yml`
|
closed
|
2023-06-18T18:26:33Z
|
2023-06-19T10:39:36Z
|
2023-06-19T10:39:36Z
|
baskrahmer
|
1,762,407,543 |
Small typos
|
Fix closing brackets and GH action link
|
Small typos: Fix closing brackets and GH action link
|
closed
|
2023-06-18T18:19:07Z
|
2023-06-19T08:51:26Z
|
2023-06-19T08:51:25Z
|
baskrahmer
|
1,761,088,254 |
New parent job runner for cached data
|
Currently, we have datasets_based_job_runner but we need a new one that only creates a chance folder without modifying the datasets library config.
Context https://github.com/huggingface/datasets-server/pull/1296#discussion_r1232512427
|
New parent job runner for cached data: Currently, we have datasets_based_job_runner but we need a new one that only creates a chance folder without modifying the datasets library config.
Context https://github.com/huggingface/datasets-server/pull/1296#discussion_r1232512427
|
closed
|
2023-06-16T18:05:01Z
|
2023-06-20T12:21:33Z
|
2023-06-20T12:21:32Z
|
AndreaFrancis
|
1,761,052,990 |
fix: 🐛 support bigger images
|
fixes https://github.com/huggingface/datasets-server/issues/1361
|
fix: 🐛 support bigger images: fixes https://github.com/huggingface/datasets-server/issues/1361
|
closed
|
2023-06-16T17:40:57Z
|
2023-06-19T11:21:41Z
|
2023-06-19T11:21:40Z
|
severo
|
1,760,975,012 |
Detect flaky hosting platforms and propose to host on the Hub
|
Many datasets are hosted on Zenodo or GDrive, and loaded using a loading script. But we have a lot of issues with them, it's not very reliable. @albertvillanova has fixed a lot of them, see https://huggingface.co/datasets/medal/discussions/2#648856b01927b18ced79d8b7 for example.
In case of errors, we could detect if the hosting platform is the issue, and create a specific message to propose the user to move the hosting to the Hub.
|
Detect flaky hosting platforms and propose to host on the Hub: Many datasets are hosted on Zenodo or GDrive, and loaded using a loading script. But we have a lot of issues with them, it's not very reliable. @albertvillanova has fixed a lot of them, see https://huggingface.co/datasets/medal/discussions/2#648856b01927b18ced79d8b7 for example.
In case of errors, we could detect if the hosting platform is the issue, and create a specific message to propose the user to move the hosting to the Hub.
|
closed
|
2023-06-16T16:39:23Z
|
2024-06-19T14:15:38Z
|
2024-06-19T14:15:38Z
|
severo
|
1,760,678,709 |
Uncaught error on config-parquet-and-info on big datasets
|
For some datasets that require copying original parquet files to `refs/convert/parquet` in multiple commits under a lock (see https://github.com/huggingface/datasets-server/issues/1349), we get:
https://datasets-server.huggingface.co/parquet?dataset=bigcode/the-stack&config=bigcode--the-stack
https://datasets-server.huggingface.co/parquet?dataset=lhoestq/tmp-lots-of-lfs-files&config=lhoestq--tmp-lots-of-lfs-files
```
{"error":"Give up after 6 attempts with <class 'huggingface_hub.utils._errors.EntryNotFoundError'>"}
```
with error_code: `UnexpectedError`
|
Uncaught error on config-parquet-and-info on big datasets: For some datasets that require copying original parquet files to `refs/convert/parquet` in multiple commits under a lock (see https://github.com/huggingface/datasets-server/issues/1349), we get:
https://datasets-server.huggingface.co/parquet?dataset=bigcode/the-stack&config=bigcode--the-stack
https://datasets-server.huggingface.co/parquet?dataset=lhoestq/tmp-lots-of-lfs-files&config=lhoestq--tmp-lots-of-lfs-files
```
{"error":"Give up after 6 attempts with <class 'huggingface_hub.utils._errors.EntryNotFoundError'>"}
```
with error_code: `UnexpectedError`
|
closed
|
2023-06-16T13:49:29Z
|
2023-07-17T16:40:53Z
|
2023-07-17T16:40:52Z
|
severo
|
1,760,673,656 |
Uncaught error in /rows on big datasets
|
For some datasets, for which the parquet files have been uploaded (or copied) with multiple commits (see https://github.com/huggingface/datasets-server/issues/1349) like `atom-in-the-universe/zlib-books-1k-50k`:
```
tiiuae/falcon-refinedweb
marianna13/zlib-books-1k-500k
atom-in-the-universe/zlib-books-1k-50k
atom-in-the-universe/zlib-books-1k-100k
atom-in-the-universe/zlib-books-1k-500k
atom-in-the-universe/zlib-books-1k-1000k
mlfoundations/datacomp_1b
```
the parquet files exist, all the steps are OK, but we get an uncaught error on /rows (in services/api):
https://datasets-server.huggingface.co/rows?dataset=atom-in-the-universe/zlib-books-1k-50k&config=atom-in-the-universe--zlib-books-1k-50k&split=train&offset=500&length=100
```
{"error":"Unexpected error."}
```
|
Uncaught error in /rows on big datasets: For some datasets, for which the parquet files have been uploaded (or copied) with multiple commits (see https://github.com/huggingface/datasets-server/issues/1349) like `atom-in-the-universe/zlib-books-1k-50k`:
```
tiiuae/falcon-refinedweb
marianna13/zlib-books-1k-500k
atom-in-the-universe/zlib-books-1k-50k
atom-in-the-universe/zlib-books-1k-100k
atom-in-the-universe/zlib-books-1k-500k
atom-in-the-universe/zlib-books-1k-1000k
mlfoundations/datacomp_1b
```
the parquet files exist, all the steps are OK, but we get an uncaught error on /rows (in services/api):
https://datasets-server.huggingface.co/rows?dataset=atom-in-the-universe/zlib-books-1k-50k&config=atom-in-the-universe--zlib-books-1k-50k&split=train&offset=500&length=100
```
{"error":"Unexpected error."}
```
|
closed
|
2023-06-16T13:46:20Z
|
2023-07-17T16:40:28Z
|
2023-07-17T16:40:28Z
|
severo
|
1,760,254,971 |
Rename dev to staging, and use staging mongodb cluster
| null |
Rename dev to staging, and use staging mongodb cluster:
|
closed
|
2023-06-16T09:22:22Z
|
2023-06-19T12:12:20Z
|
2023-06-19T12:12:18Z
|
severo
|
1,760,249,622 |
Change "dev" environment to "staging"
|
It makes more sense to call it "staging".
And it will use the mongo atlas staging cluster
|
Change "dev" environment to "staging": It makes more sense to call it "staging".
And it will use the mongo atlas staging cluster
|
closed
|
2023-06-16T09:18:33Z
|
2023-06-20T18:05:49Z
|
2023-06-20T18:05:49Z
|
severo
|
1,760,105,155 |
Upgrade prod mongo from v5 to v6
|
This is needed for `$in` function in TTL index: https://github.com/huggingface/datasets-server/pull/1325/files#diff-44fa7cb2645881e55953db64dafa198b2e007a2e531f70acaeebfc50ffa67953R141
See https://www.mongodb.com/docs/manual/release-notes/6.0/#indexes
---
to upgrade: https://www.mongodb.com/docs/atlas/tutorial/major-version-change/
|
Upgrade prod mongo from v5 to v6: This is needed for `$in` function in TTL index: https://github.com/huggingface/datasets-server/pull/1325/files#diff-44fa7cb2645881e55953db64dafa198b2e007a2e531f70acaeebfc50ffa67953R141
See https://www.mongodb.com/docs/manual/release-notes/6.0/#indexes
---
to upgrade: https://www.mongodb.com/docs/atlas/tutorial/major-version-change/
|
closed
|
2023-06-16T07:37:29Z
|
2023-06-20T15:08:11Z
|
2023-06-20T15:08:11Z
|
severo
|
1,760,089,716 |
Revert "Delete ttl index from queue.py code (#1378)"
|
This reverts commit 47ea65b2567db4482579cd7000393cf0a15b412e.
|
Revert "Delete ttl index from queue.py code (#1378)": This reverts commit 47ea65b2567db4482579cd7000393cf0a15b412e.
|
closed
|
2023-06-16T07:29:39Z
|
2023-06-16T07:29:55Z
|
2023-06-16T07:29:54Z
|
severo
|
1,759,657,559 |
Rollback TTL index
| null |
Rollback TTL index:
|
closed
|
2023-06-15T23:08:40Z
|
2023-06-15T23:20:25Z
|
2023-06-15T23:20:24Z
|
AndreaFrancis
|
1,759,573,668 |
Delete ttl index from queue.py code
|
First part of https://github.com/huggingface/datasets-server/issues/1326
|
Delete ttl index from queue.py code: First part of https://github.com/huggingface/datasets-server/issues/1326
|
closed
|
2023-06-15T21:28:16Z
|
2023-06-15T22:08:08Z
|
2023-06-15T22:08:07Z
|
AndreaFrancis
|
1,759,467,470 |
[docs] Add build notebook workflow
|
Enables the doc-builder to build Colab notebooks :)
|
[docs] Add build notebook workflow: Enables the doc-builder to build Colab notebooks :)
|
closed
|
2023-06-15T19:58:06Z
|
2023-06-15T20:22:51Z
|
2023-06-15T20:22:50Z
|
stevhliu
|
1,759,417,823 |
[docs] Improvements
|
Based on @mishig25's [feedback](https://huggingface.slack.com/archives/C0311GZ7R6K/p1684484245379859), this adds:
- a response to the code snippets in the Quickstart
- an end-to-end example of using `/parquet` to get a dataset, analyze it, and plot the results
- button to open in a Colab notebook to run examples right away
|
[docs] Improvements: Based on @mishig25's [feedback](https://huggingface.slack.com/archives/C0311GZ7R6K/p1684484245379859), this adds:
- a response to the code snippets in the Quickstart
- an end-to-end example of using `/parquet` to get a dataset, analyze it, and plot the results
- button to open in a Colab notebook to run examples right away
|
closed
|
2023-06-15T19:23:55Z
|
2023-06-16T16:10:35Z
|
2023-06-16T16:10:04Z
|
stevhliu
|
1,759,223,139 |
Fix fill_builder_info
|
close https://github.com/huggingface/datasets-server/issues/1374
NamedSplit peut pas être convertit par orjson
|
Fix fill_builder_info: close https://github.com/huggingface/datasets-server/issues/1374
NamedSplit peut pas être convertit par orjson
|
closed
|
2023-06-15T16:57:35Z
|
2023-06-15T20:37:36Z
|
2023-06-15T20:37:35Z
|
lhoestq
|
1,759,049,916 |
Truncated cells seem to prevent conversion to parquet
|
See https://github.com/huggingface/datasets-server/pull/1372#issuecomment-1593249655
```
Traceback (most recent call last):
File "/src/services/worker/src/worker/job_manager.py", line 167, in process
if len(orjson_dumps(content)) > self.worker_config.content_max_bytes:
File "/src/libs/libcommon/src/libcommon/utils.py", line 79, in orjson_dumps
return orjson.dumps(content, option=orjson.OPT_UTC_Z, default=orjson_default)
TypeError: Dict key must be str
```
on https://huggingface.co/datasets/philippemo/dummy_dataset_without_schema_12_06
|
Truncated cells seem to prevent conversion to parquet: See https://github.com/huggingface/datasets-server/pull/1372#issuecomment-1593249655
```
Traceback (most recent call last):
File "/src/services/worker/src/worker/job_manager.py", line 167, in process
if len(orjson_dumps(content)) > self.worker_config.content_max_bytes:
File "/src/libs/libcommon/src/libcommon/utils.py", line 79, in orjson_dumps
return orjson.dumps(content, option=orjson.OPT_UTC_Z, default=orjson_default)
TypeError: Dict key must be str
```
on https://huggingface.co/datasets/philippemo/dummy_dataset_without_schema_12_06
|
closed
|
2023-06-15T15:11:19Z
|
2023-06-15T20:37:36Z
|
2023-06-15T20:37:36Z
|
severo
|
1,758,633,320 |
Refac hub_datasets fixture
|
This way no need to setup all the hub datasets fixtures just to run one single test
close #921
|
Refac hub_datasets fixture: This way no need to setup all the hub datasets fixtures just to run one single test
close #921
|
closed
|
2023-06-15T11:35:02Z
|
2023-06-15T20:53:44Z
|
2023-06-15T20:53:43Z
|
lhoestq
|
1,758,454,961 |
Update datasets dependency to 2.13.0 version
|
After 2.13.0 datasets release, update dependencies on it.
Note that I have also removed the explicit dependency on `datasets` from `services/api`,
- see commit: https://github.com/huggingface/datasets-server/commit/a2c0cd908b45a7065d936964c4d0477143146d6c
This is analogous to what was previously done on `services/worker`.
- See discussion: https://github.com/huggingface/datasets-server/pull/1147#issuecomment-1539883057
- See commit: https://github.com/huggingface/datasets-server/pull/1147/commits/4163a18ec561ddf93f93c096d76c2b06cf652f4d
Fix #1370.
|
Update datasets dependency to 2.13.0 version: After 2.13.0 datasets release, update dependencies on it.
Note that I have also removed the explicit dependency on `datasets` from `services/api`,
- see commit: https://github.com/huggingface/datasets-server/commit/a2c0cd908b45a7065d936964c4d0477143146d6c
This is analogous to what was previously done on `services/worker`.
- See discussion: https://github.com/huggingface/datasets-server/pull/1147#issuecomment-1539883057
- See commit: https://github.com/huggingface/datasets-server/pull/1147/commits/4163a18ec561ddf93f93c096d76c2b06cf652f4d
Fix #1370.
|
closed
|
2023-06-15T09:48:39Z
|
2023-06-15T20:49:23Z
|
2023-06-15T15:59:03Z
|
albertvillanova
|
1,757,309,822 |
Adding limit for number of configs
|
Closes https://github.com/huggingface/datasets-server/issues/1367
|
Adding limit for number of configs: Closes https://github.com/huggingface/datasets-server/issues/1367
|
closed
|
2023-06-14T16:56:59Z
|
2023-06-15T14:58:43Z
|
2023-06-15T14:58:42Z
|
AndreaFrancis
|
1,757,272,674 |
Update datasets to 2.13.0
|
https://github.com/huggingface/datasets/releases/tag/2.13.0
Related to the datasets server:
- Better row group size in push_to_hub by @lhoestq in https://github.com/huggingface/datasets/pull/5935
- Make get_from_cache use custom temp filename that is locked by @albertvillanova in https://github.com/huggingface/datasets/pull/5938
Did I miss something @huggingface/datasets-server ?
|
Update datasets to 2.13.0: https://github.com/huggingface/datasets/releases/tag/2.13.0
Related to the datasets server:
- Better row group size in push_to_hub by @lhoestq in https://github.com/huggingface/datasets/pull/5935
- Make get_from_cache use custom temp filename that is locked by @albertvillanova in https://github.com/huggingface/datasets/pull/5938
Did I miss something @huggingface/datasets-server ?
|
closed
|
2023-06-14T16:31:24Z
|
2023-06-15T15:59:05Z
|
2023-06-15T15:59:05Z
|
severo
|
1,757,265,256 |
Remove duplicates in cache
|
We have cache entries for `bigscience/P3` and `BigScience/P3` for example: they resolve to the same dataset.
|
Remove duplicates in cache: We have cache entries for `bigscience/P3` and `BigScience/P3` for example: they resolve to the same dataset.
|
closed
|
2023-06-14T16:26:07Z
|
2023-07-17T16:41:02Z
|
2023-07-17T16:41:02Z
|
severo
|
1,757,157,046 |
feat: 🎸 reduce the resources
| null |
feat: 🎸 reduce the resources:
|
closed
|
2023-06-14T15:23:53Z
|
2023-06-14T15:25:10Z
|
2023-06-14T15:25:09Z
|
severo
|
1,757,098,647 |
Set a limit on the number of configs
|
Dataset https://huggingface.co/datasets/Muennighoff/flores200 has more than 40,000 configs. It's too much for our infrastructure for now. We should set a limit on it.
|
Set a limit on the number of configs: Dataset https://huggingface.co/datasets/Muennighoff/flores200 has more than 40,000 configs. It's too much for our infrastructure for now. We should set a limit on it.
|
closed
|
2023-06-14T14:55:05Z
|
2023-06-15T14:58:44Z
|
2023-06-15T14:58:43Z
|
severo
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.