Update docs for nyu_depth_v2 dataset

[awsaf49]

This PR will fix the issue mentioned in #5461. Here is brief overview,

Bug:

Discrepancy between depth map of nyu_depth_v2 dataset here and actual depth map. Depth values somehow got discretized/clipped resulting in depth maps that are different from actual ones. Here is a side-by-side comparison,

Fix:

When I first loaded the datasets from HF I noticed it was 30GB but in DenseDepth data is only 4GB with dtype=uint8. This means data from fast-depth (before loading to HF) must have high precision. So when I tried to dig deeper by directly loading depth_map from h5py, I found depth_map from h5py came with float32. But when the data is processed in HF with datasets.Image() it was directly converted to uint8 from float32 hence the discretized depth map.

datasets/src/datasets/features/image.py

Lines 91 to 93 in c78559c

	elif isinstance(value, np.ndarray):
	# convert the image array to png bytes
	image = PIL.Image.fromarray(value.astype(np.uint8))

cc: @sayakpaul @lhoestq

[awsaf49]

I think I need to create another PR on https://huggingface.co/datasets/huggingface/documentation-images/tree/main/datasets for hosting the images there?

[sayakpaul]

Thanks a lot for working on this! Just a few nits.

[HuggingFaceDocBuilderDev]

The documentation is not available anymore as the PR was closed or merged.

[lhoestq]

Thanks for the update @awsaf49 !

[sayakpaul]

Thanks a lot for the updates!

Just some minor things remain and the we should be good to ship this 🚀

[awsaf49]

Thanks a lot for the updates!

Just some minor things remain and the we should be good to ship this 🚀

@sayakpaul I have updated the minor things. Please approve the workflows

[awsaf49]

I think this PR is good to go..
@sayakpaul @lhoestq

[github-actions]

Show benchmarks

PyArrow==6.0.0

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)	0.009064 / 0.011353 (-0.002289)	0.005262 / 0.011008 (-0.005746)	0.099608 / 0.038508 (0.061100)	0.035015 / 0.023109 (0.011906)	0.296501 / 0.275898 (0.020602)	0.353619 / 0.323480 (0.030139)	0.007903 / 0.007986 (-0.000083)	0.004093 / 0.004328 (-0.000235)	0.075260 / 0.004250 (0.071009)	0.043142 / 0.037052 (0.006089)	0.307755 / 0.258489 (0.049266)	0.336340 / 0.293841 (0.042499)	0.038596 / 0.128546 (-0.089950)	0.011861 / 0.075646 (-0.063786)	0.334226 / 0.419271 (-0.085045)	0.051472 / 0.043533 (0.007940)	0.298539 / 0.255139 (0.043400)	0.316856 / 0.283200 (0.033656)	0.108620 / 0.141683 (-0.033063)	1.434901 / 1.452155 (-0.017254)	1.468368 / 1.492716 (-0.024348)

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)	0.208402 / 0.018006 (0.190395)	0.445799 / 0.000490 (0.445309)	0.003704 / 0.000200 (0.003504)	0.000084 / 0.000054 (0.000030)

Benchmark: benchmark_indices_mapping.json

metric	select	shard	shuffle	sort	train_test_split
new / old (diff)	0.025435 / 0.037411 (-0.011976)	0.105874 / 0.014526 (0.091348)	0.115652 / 0.176557 (-0.060905)	0.150872 / 0.737135 (-0.586263)	0.121705 / 0.296338 (-0.174633)

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)	0.397816 / 0.215209 (0.182607)	3.977766 / 2.077655 (1.900111)	1.850848 / 1.504120 (0.346728)	1.686062 / 1.541195 (0.144867)	1.786277 / 1.468490 (0.317787)	0.696250 / 4.584777 (-3.888527)	3.785255 / 3.745712 (0.039543)	3.355013 / 5.269862 (-1.914849)	1.818232 / 4.565676 (-2.747444)	0.085408 / 0.424275 (-0.338867)	0.012567 / 0.007607 (0.004960)	0.524185 / 0.226044 (0.298140)	5.061975 / 2.268929 (2.793047)	2.299866 / 55.444624 (-53.144758)	1.966709 / 6.876477 (-4.909768)	2.018760 / 2.142072 (-0.123313)	0.841341 / 4.805227 (-3.963886)	0.166374 / 6.500664 (-6.334290)	0.061854 / 0.075469 (-0.013615)

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)	1.221666 / 1.841788 (-0.620122)	14.373194 / 8.074308 (6.298886)	14.253614 / 10.191392 (4.062222)	0.172979 / 0.680424 (-0.507445)	0.029176 / 0.534201 (-0.505025)	0.447399 / 0.579283 (-0.131884)	0.443663 / 0.434364 (0.009299)	0.537071 / 0.540337 (-0.003267)	0.640539 / 1.386936 (-0.746397)

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)	0.007019 / 0.011353 (-0.004334)	0.005091 / 0.011008 (-0.005917)	0.074588 / 0.038508 (0.036080)	0.032391 / 0.023109 (0.009282)	0.340548 / 0.275898 (0.064650)	0.367159 / 0.323480 (0.043679)	0.005594 / 0.007986 (-0.002392)	0.004003 / 0.004328 (-0.000325)	0.073946 / 0.004250 (0.069695)	0.045921 / 0.037052 (0.008868)	0.340245 / 0.258489 (0.081756)	0.397958 / 0.293841 (0.104117)	0.036539 / 0.128546 (-0.092007)	0.012258 / 0.075646 (-0.063388)	0.087406 / 0.419271 (-0.331865)	0.049276 / 0.043533 (0.005743)	0.345235 / 0.255139 (0.090096)	0.361250 / 0.283200 (0.078050)	0.100757 / 0.141683 (-0.040926)	1.464644 / 1.452155 (0.012489)	1.545852 / 1.492716 (0.053136)

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)	0.222952 / 0.018006 (0.204945)	0.434607 / 0.000490 (0.434117)	0.000438 / 0.000200 (0.000238)	0.000060 / 0.000054 (0.000006)

Benchmark: benchmark_indices_mapping.json

metric	select	shard	shuffle	sort	train_test_split
new / old (diff)	0.028834 / 0.037411 (-0.008577)	0.107523 / 0.014526 (0.092997)	0.122077 / 0.176557 (-0.054479)	0.156574 / 0.737135 (-0.580561)	0.122917 / 0.296338 (-0.173421)

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)	0.417292 / 0.215209 (0.202083)	4.165980 / 2.077655 (2.088325)	1.996731 / 1.504120 (0.492611)	1.802946 / 1.541195 (0.261751)	1.878456 / 1.468490 (0.409966)	0.711035 / 4.584777 (-3.873742)	3.847357 / 3.745712 (0.101644)	2.088354 / 5.269862 (-3.181508)	1.344763 / 4.565676 (-3.220913)	0.086356 / 0.424275 (-0.337919)	0.012530 / 0.007607 (0.004923)	0.511693 / 0.226044 (0.285648)	5.126093 / 2.268929 (2.857165)	2.490023 / 55.444624 (-52.954602)	2.180274 / 6.876477 (-4.696202)	2.221511 / 2.142072 (0.079438)	0.836348 / 4.805227 (-3.968879)	0.169554 / 6.500664 (-6.331110)	0.064555 / 0.075469 (-0.010914)

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)	1.293466 / 1.841788 (-0.548321)	14.785700 / 8.074308 (6.711392)	13.858493 / 10.191392 (3.667101)	0.161777 / 0.680424 (-0.518646)	0.017794 / 0.534201 (-0.516407)	0.426286 / 0.579283 (-0.152997)	0.422517 / 0.434364 (-0.011847)	0.530777 / 0.540337 (-0.009560)	0.634822 / 1.386936 (-0.752114)