Allow SplitDict setitem to replace existing SplitInfo

[lhoestq]

Fix this code provided by @clefourrier

import datasets
import os

token = os.getenv("TOKEN")

results = datasets.load_dataset("gaia-benchmark/results_public", "2023", token=token, download_mode=datasets.DownloadMode.FORCE_REDOWNLOAD)
results["test"] = datasets.Dataset.from_list([row for row in results["test"] if row["model"] != "StateFlow"])
results["test"].push_to_hub("gaia-benchmark/results_public", "2023", token=token, split="test")

ValueError                                Traceback (most recent call last)
Cell In[43], line 1
----> 1 results["test"].push_to_hub("gaia-benchmark/results_public", "2023", token=token, split="test")

File ~/miniconda3/envs/default310/lib/python3.10/site-packages/datasets/arrow_dataset.py:5498, in Dataset.push_to_hub(self, repo_id, config_name, split, private, token, branch, max_shard_size, num_shards, embed_external_files)
   5496         repo_info.dataset_size = (repo_info.dataset_size or 0) + dataset_nbytes
   5497         repo_info.size_in_bytes = repo_info.download_size + repo_info.dataset_size
-> 5498         repo_info.splits[split] = SplitInfo(
   5499             split, num_bytes=dataset_nbytes, num_examples=len(self), dataset_name=dataset_name
   5500         )
   5501         info_to_dump = repo_info
   5502 # create the metadata configs if it was uploaded with push_to_hub before metadata configs existed

File ~/miniconda3/envs/default310/lib/python3.10/site-packages/datasets/splits.py:541, in SplitDict.__setitem__(self, key, value)
    539     raise ValueError(f"Cannot add elem. (key mismatch: '{key}' != '{value.name}')")
    540 if key in self:
--> 541     raise ValueError(f"Split {key} already present")
    542 super().__setitem__(key, value)

ValueError: Split test already present

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[github-actions]

Show benchmarks

PyArrow==8.0.0

Show updated benchmarks!

Benchmark: benchmark_array_xd.json

metric	read_batch_formatted_as_numpy after write_array2d	read_batch_formatted_as_numpy after write_flattened_sequence	read_batch_formatted_as_numpy after write_nested_sequence	read_batch_unformated after write_array2d	read_batch_unformated after write_flattened_sequence	read_batch_unformated after write_nested_sequence	read_col_formatted_as_numpy after write_array2d	read_col_formatted_as_numpy after write_flattened_sequence	read_col_formatted_as_numpy after write_nested_sequence	read_col_unformated after write_array2d	read_col_unformated after write_flattened_sequence	read_col_unformated after write_nested_sequence	read_formatted_as_numpy after write_array2d	read_formatted_as_numpy after write_flattened_sequence	read_formatted_as_numpy after write_nested_sequence	read_unformated after write_array2d	read_unformated after write_flattened_sequence	read_unformated after write_nested_sequence	write_array2d	write_flattened_sequence	write_nested_sequence
new / old (diff)	0.004968 / 0.011353 (-0.006385)	0.003732 / 0.011008 (-0.007276)	0.063672 / 0.038508 (0.025164)	0.027066 / 0.023109 (0.003957)	0.253306 / 0.275898 (-0.022592)	0.283382 / 0.323480 (-0.040098)	0.004217 / 0.007986 (-0.003768)	0.002865 / 0.004328 (-0.001464)	0.048672 / 0.004250 (0.044421)	0.040740 / 0.037052 (0.003688)	0.271848 / 0.258489 (0.013359)	0.293162 / 0.293841 (-0.000679)	0.027410 / 0.128546 (-0.101136)	0.010605 / 0.075646 (-0.065042)	0.210545 / 0.419271 (-0.208726)	0.036085 / 0.043533 (-0.007447)	0.259807 / 0.255139 (0.004668)	0.274056 / 0.283200 (-0.009144)	0.018812 / 0.141683 (-0.122871)	1.116687 / 1.452155 (-0.335468)	1.164276 / 1.492716 (-0.328440)

Benchmark: benchmark_getitem_100B.json

metric	get_batch_of_1024_random_rows	get_batch_of_1024_rows	get_first_row	get_last_row
new / old (diff)	0.092874 / 0.018006 (0.074868)	0.355897 / 0.000490 (0.355407)	0.000224 / 0.000200 (0.000024)	0.000051 / 0.000054 (-0.000004)

Benchmark: benchmark_indices_mapping.json

metric	select	shard	shuffle	sort	train_test_split
new / old (diff)	0.018461 / 0.037411 (-0.018950)	0.062061 / 0.014526 (0.047535)	0.072353 / 0.176557 (-0.104203)	0.119162 / 0.737135 (-0.617974)	0.082974 / 0.296338 (-0.213364)

Benchmark: benchmark_iterating.json

metric	read 5000	read 50000	read_batch 50000 10	read_batch 50000 100	read_batch 50000 1000	read_formatted numpy 5000	read_formatted pandas 5000	read_formatted tensorflow 5000	read_formatted torch 5000	read_formatted_batch numpy 5000 10	read_formatted_batch numpy 5000 1000	shuffled read 5000	shuffled read 50000	shuffled read_batch 50000 10	shuffled read_batch 50000 100	shuffled read_batch 50000 1000	shuffled read_formatted numpy 5000	shuffled read_formatted_batch numpy 5000 10	shuffled read_formatted_batch numpy 5000 1000
new / old (diff)	0.291631 / 0.215209 (0.076422)	2.861495 / 2.077655 (0.783841)	1.496753 / 1.504120 (-0.007367)	1.371164 / 1.541195 (-0.170031)	1.415473 / 1.468490 (-0.053018)	0.566778 / 4.584777 (-4.017999)	2.376209 / 3.745712 (-1.369503)	2.812326 / 5.269862 (-2.457535)	1.765640 / 4.565676 (-2.800037)	0.063274 / 0.424275 (-0.361001)	0.004933 / 0.007607 (-0.002674)	0.342345 / 0.226044 (0.116301)	3.407487 / 2.268929 (1.138558)	1.856646 / 55.444624 (-53.587978)	1.590284 / 6.876477 (-5.286193)	1.610068 / 2.142072 (-0.532004)	0.656007 / 4.805227 (-4.149220)	0.118310 / 6.500664 (-6.382354)	0.042596 / 0.075469 (-0.032873)

Benchmark: benchmark_map_filter.json

metric	filter	map fast-tokenizer batched	map identity	map identity batched	map no-op batched	map no-op batched numpy	map no-op batched pandas	map no-op batched pytorch	map no-op batched tensorflow
new / old (diff)	0.991392 / 1.841788 (-0.850395)	11.612397 / 8.074308 (3.538089)	9.627836 / 10.191392 (-0.563556)	0.130575 / 0.680424 (-0.549848)	0.014152 / 0.534201 (-0.520049)	0.289736 / 0.579283 (-0.289548)	0.260041 / 0.434364 (-0.174323)	0.339730 / 0.540337 (-0.200608)	0.447529 / 1.386936 (-0.939407)

PyArrow==latest

Show updated benchmarks!

Benchmark: benchmark_array_xd.json

metric	read_batch_formatted_as_numpy after write_array2d	read_batch_formatted_as_numpy after write_flattened_sequence	read_batch_formatted_as_numpy after write_nested_sequence	read_batch_unformated after write_array2d	read_batch_unformated after write_flattened_sequence	read_batch_unformated after write_nested_sequence	read_col_formatted_as_numpy after write_array2d	read_col_formatted_as_numpy after write_flattened_sequence	read_col_formatted_as_numpy after write_nested_sequence	read_col_unformated after write_array2d	read_col_unformated after write_flattened_sequence	read_col_unformated after write_nested_sequence	read_formatted_as_numpy after write_array2d	read_formatted_as_numpy after write_flattened_sequence	read_formatted_as_numpy after write_nested_sequence	read_unformated after write_array2d	read_unformated after write_flattened_sequence	read_unformated after write_nested_sequence	write_array2d	write_flattened_sequence	write_nested_sequence
new / old (diff)	0.005315 / 0.011353 (-0.006038)	0.003955 / 0.011008 (-0.007053)	0.049618 / 0.038508 (0.011110)	0.030404 / 0.023109 (0.007295)	0.258727 / 0.275898 (-0.017171)	0.282020 / 0.323480 (-0.041460)	0.004356 / 0.007986 (-0.003629)	0.002866 / 0.004328 (-0.001462)	0.049122 / 0.004250 (0.044872)	0.045534 / 0.037052 (0.008482)	0.269560 / 0.258489 (0.011071)	0.301225 / 0.293841 (0.007384)	0.029786 / 0.128546 (-0.098761)	0.010433 / 0.075646 (-0.065213)	0.058222 / 0.419271 (-0.361049)	0.052968 / 0.043533 (0.009435)	0.256605 / 0.255139 (0.001467)	0.279899 / 0.283200 (-0.003300)	0.018233 / 0.141683 (-0.123450)	1.164060 / 1.452155 (-0.288095)	1.218049 / 1.492716 (-0.274667)

Benchmark: benchmark_getitem_100B.json

metric	get_batch_of_1024_random_rows	get_batch_of_1024_rows	get_first_row	get_last_row
new / old (diff)	0.093646 / 0.018006 (0.075639)	0.288804 / 0.000490 (0.288314)	0.000224 / 0.000200 (0.000024)	0.000051 / 0.000054 (-0.000003)

Benchmark: benchmark_indices_mapping.json

metric	select	shard	shuffle	sort	train_test_split
new / old (diff)	0.022193 / 0.037411 (-0.015219)	0.075507 / 0.014526 (0.060981)	0.086091 / 0.176557 (-0.090465)	0.127433 / 0.737135 (-0.609703)	0.087064 / 0.296338 (-0.209274)

Benchmark: benchmark_iterating.json

metric	read 5000	read 50000	read_batch 50000 10	read_batch 50000 100	read_batch 50000 1000	read_formatted numpy 5000	read_formatted pandas 5000	read_formatted tensorflow 5000	read_formatted torch 5000	read_formatted_batch numpy 5000 10	read_formatted_batch numpy 5000 1000	shuffled read 5000	shuffled read 50000	shuffled read_batch 50000 10	shuffled read_batch 50000 100	shuffled read_batch 50000 1000	shuffled read_formatted numpy 5000	shuffled read_formatted_batch numpy 5000 10	shuffled read_formatted_batch numpy 5000 1000
new / old (diff)	0.292459 / 0.215209 (0.077250)	2.842430 / 2.077655 (0.764776)	1.505824 / 1.504120 (0.001704)	1.377052 / 1.541195 (-0.164143)	1.408757 / 1.468490 (-0.059733)	0.571705 / 4.584777 (-4.013072)	2.459798 / 3.745712 (-1.285914)	2.714826 / 5.269862 (-2.555035)	1.782064 / 4.565676 (-2.783613)	0.063113 / 0.424275 (-0.361162)	0.005099 / 0.007607 (-0.002509)	0.343624 / 0.226044 (0.117579)	3.415806 / 2.268929 (1.146878)	1.853253 / 55.444624 (-53.591371)	1.584392 / 6.876477 (-5.292084)	1.720384 / 2.142072 (-0.421689)	0.646637 / 4.805227 (-4.158590)	0.118072 / 6.500664 (-6.382593)	0.041362 / 0.075469 (-0.034107)

Benchmark: benchmark_map_filter.json

metric	filter	map fast-tokenizer batched	map identity	map identity batched	map no-op batched	map no-op batched numpy	map no-op batched pandas	map no-op batched pytorch	map no-op batched tensorflow
new / old (diff)	1.020086 / 1.841788 (-0.821701)	12.303980 / 8.074308 (4.229672)	10.322869 / 10.191392 (0.131477)	0.140959 / 0.680424 (-0.539465)	0.015372 / 0.534201 (-0.518829)	0.288552 / 0.579283 (-0.290731)	0.278243 / 0.434364 (-0.156121)	0.328399 / 0.540337 (-0.211939)	0.433618 / 1.386936 (-0.953318)