Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

More cudf #87

Merged
merged 5 commits into from
Dec 2, 2024
Merged

More cudf #87

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
4133a3a
Add daft and spark
martindurant Oct 28, 2024
40e48f9
cudf updates
Nov 19, 2024
a1524e8
remove unneeded files
Nov 19, 2024
5ea1276
Enable dataframes in cudf
martindurant Nov 28, 2024
4260c10
Merge branch 'main' into more_cudf
martindurant Nov 28, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
83 changes: 41 additions & 42 deletions example/cudf-ak.ipynb
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9,27 +9,27 @@
"```python\n",
"import awkward as ak\n",
"\n",
"def make_data(fn):\n",
"def make_data(fn, N=1000000):\n",
" part = [[[1, 2, 3], [], [4, 5]],\n",
" [[6, 7]]] * 1000000\n",
" [[6, 7]]] * N\n",
" arr = ak.Array({\"a\": part})\n",
" ak.to_parquet(arr, fn, extensionarray=False)\n",
"```"
]
},
{
"cell_type": "code",
"execution_count": 1,
"execution_count": 3,
"id": "cefd8e53-a56f-4b0c-88d2-d662d59849a7",
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"('2.6.9', '2024.8.1.dev29+g9b9f27f.d20240927')"
"('2.7.1', '2024.10.1.dev9+g9f64d31')"
]
},
"execution_count": 1,
"execution_count": 3,
"metadata": {},
"output_type": "execute_result"
}
Expand All @@ -43,35 +43,28 @@
"import subprocess\n",
"\n",
"def gpu_mem():\n",
" return\n",
" print(subprocess.check_output(\"nvidia-smi | grep py\", shell=True).split()[-2].decode())\n",
"\n",
"ak.__version__, akimbo.__version__"
]
},
{
"cell_type": "code",
"execution_count": 2,
"execution_count": 4,
"id": "0490043a-564a-4c11-bb0d-a54fb4c6fb10",
"metadata": {
"scrolled": true
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"160MiB\n"
]
}
],
"outputs": [],
"source": [
"df = cudf.read_parquet(\"/floppy/code/awkward/s.parquet\")\n",
"df = cudf.read_parquet(\"s.parquet\")\n",
"gpu_mem()"
]
},
{
"cell_type": "code",
"execution_count": 3,
"execution_count": 5,
"id": "e29ff9a4-60e4-4260-9a44-c135ad6d7d6b",
"metadata": {},
"outputs": [
Expand All @@ -82,7 +75,7 @@
"dtype: object"
]
},
"execution_count": 3,
"execution_count": 5,
"metadata": {},
"output_type": "execute_result"
}
Expand All @@ -93,7 +86,7 @@
},
{
"cell_type": "code",
"execution_count": 4,
"execution_count": 6,
"id": "58d16a80-041e-4260-8c56-9de932dde557",
"metadata": {},
"outputs": [
Expand All @@ -104,7 +97,7 @@
"Name: 0, dtype: list"
]
},
"execution_count": 4,
"execution_count": 6,
"metadata": {},
"output_type": "execute_result"
}
Expand All @@ -115,7 +108,7 @@
},
{
"cell_type": "code",
"execution_count": 5,
"execution_count": 7,
"id": "c7b65320-e1fa-44b2-a232-6ffb97ba1d18",
"metadata": {
"scrolled": true
Expand Down Expand Up @@ -185,6 +178,7 @@
" 'from_raggedtensor',\n",
" 'from_rdataframe',\n",
" 'from_regular',\n",
" 'from_tensorflow',\n",
" 'from_torch',\n",
" 'full_like',\n",
" 'highlevel',\n",
Expand All @@ -209,6 +203,7 @@
" 'mixin_class',\n",
" 'mixin_class_method',\n",
" 'moment',\n",
" 'named_axis',\n",
" 'nan_to_none',\n",
" 'nan_to_num',\n",
" 'nanargmax',\n",
Expand All @@ -229,6 +224,7 @@
" 'operations',\n",
" 'pad_none',\n",
" 'parameters',\n",
" 'positional_axis',\n",
" 'prettyprint',\n",
" 'prod',\n",
" 'ptp',\n",
Expand Down Expand Up @@ -265,6 +261,7 @@
" 'to_raggedtensor',\n",
" 'to_rdataframe',\n",
" 'to_regular',\n",
" 'to_tensorflow',\n",
" 'to_torch',\n",
" 'tolist',\n",
" 'transform',\n",
Expand All @@ -281,14 +278,16 @@
" 'where',\n",
" 'with_field',\n",
" 'with_name',\n",
" 'with_named_axis',\n",
" 'with_parameter',\n",
" 'without_field',\n",
" 'without_named_axis',\n",
" 'without_parameters',\n",
" 'zeros_like',\n",
" 'zip']"
]
},
"execution_count": 5,
"execution_count": 7,
"metadata": {},
"output_type": "execute_result"
}
Expand All @@ -300,7 +299,7 @@
},
{
"cell_type": "code",
"execution_count": 6,
"execution_count": 8,
"id": "8ff11e13-8503-4d79-a64c-993028709ca4",
"metadata": {},
"outputs": [
Expand All @@ -310,7 +309,7 @@
"array(28000000)"
]
},
"execution_count": 6,
"execution_count": 8,
"metadata": {},
"output_type": "execute_result"
}
Expand All @@ -321,7 +320,7 @@
},
{
"cell_type": "code",
"execution_count": 7,
"execution_count": 9,
"id": "2dd99fe5-0523-46c9-87ec-1392070f5139",
"metadata": {},
"outputs": [
Expand All @@ -331,7 +330,7 @@
"cupy.ndarray"
]
},
"execution_count": 7,
"execution_count": 9,
"metadata": {},
"output_type": "execute_result"
}
Expand All @@ -343,15 +342,15 @@
},
{
"cell_type": "code",
"execution_count": 8,
"execution_count": 11,
"id": "9d8e55cf-8cf1-40a0-8733-24b7719f431d",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"4.83 ms ± 16 μs per loop (mean ± std. dev. of 7 runs, 100 loops each)\n"
"12.6 ms ± 779 μs per loop (mean ± std. dev. of 7 runs, 10 loops each)\n"
]
}
],
Expand All @@ -362,7 +361,7 @@
},
{
"cell_type": "code",
"execution_count": 9,
"execution_count": 12,
"id": "fae94aea-d9cf-4228-bcab-f843c7cc9c98",
"metadata": {},
"outputs": [
Expand All @@ -383,7 +382,7 @@
"Length: 2000000, dtype: list"
]
},
"execution_count": 9,
"execution_count": 12,
"metadata": {},
"output_type": "execute_result"
}
Expand Down Expand Up @@ -413,7 +412,7 @@
},
{
"cell_type": "code",
"execution_count": 11,
"execution_count": 13,
"id": "558ca2c3-d6c7-4404-bcab-557b9b03f795",
"metadata": {},
"outputs": [
Expand Down Expand Up @@ -445,7 +444,7 @@
},
{
"cell_type": "code",
"execution_count": 12,
"execution_count": 14,
"id": "d240ea54-87b4-4b99-b67f-b2f885a4bf5e",
"metadata": {
"scrolled": true
Expand All @@ -457,7 +456,7 @@
"array([15, 13, 15, ..., 13, 15, 13], dtype=int32)"
]
},
"execution_count": 12,
"execution_count": 14,
"metadata": {},
"output_type": "execute_result"
}
Expand Down Expand Up @@ -485,15 +484,15 @@
},
{
"cell_type": "code",
"execution_count": 13,
"execution_count": 15,
"id": "73a35144-292f-4b1d-bbc0-4ebba2a84b0d",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"6.17 ms ± 118 μs per loop (mean ± std. dev. of 7 runs, 100 loops each)\n"
"16.7 ms ± 233 μs per loop (mean ± std. dev. of 7 runs, 10 loops each)\n"
]
}
],
Expand Down Expand Up @@ -529,7 +528,7 @@
},
{
"cell_type": "code",
"execution_count": 15,
"execution_count": 16,
"id": "d039a508-e77c-4e23-a583-ec7997a88bb1",
"metadata": {},
"outputs": [
Expand All @@ -550,7 +549,7 @@
"Length: 2000000, dtype: list"
]
},
"execution_count": 15,
"execution_count": 16,
"metadata": {},
"output_type": "execute_result"
}
Expand All @@ -562,7 +561,7 @@
},
{
"cell_type": "code",
"execution_count": 16,
"execution_count": 17,
"id": "f149dfaf-c01e-4d0a-8e01-2d20623d216f",
"metadata": {},
"outputs": [
Expand All @@ -583,7 +582,7 @@
"Length: 2000000, dtype: list"
]
},
"execution_count": 16,
"execution_count": 17,
"metadata": {},
"output_type": "execute_result"
}
Expand All @@ -604,9 +603,9 @@
],
"metadata": {
"kernelspec": {
"display_name": "Python [conda env:cuda] *",
"display_name": "Python 3 (ipykernel)",
"language": "python",
"name": "conda-env-cuda-py"
"name": "python3"
},
"language_info": {
"codemirror_mode": {
Expand All @@ -618,7 +617,7 @@
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.10.14"
"version": "3.12.0"
}
},
"nbformat": 4,
Expand Down
14 changes: 10 additions & 4 deletions src/akimbo/cudf.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -47,7 +47,7 @@ def dec_cu(op, match=match_string):
def f(lay, **kwargs):
# op(column, ...)->column
col = op(lay._to_cudf(cudf, None, len(lay)), **kwargs)
return from_cudf(cudf.Series(col)).layout
return from_cudf(cudf.Series._from_column(col)).layout

return dec(func=f, match=match, inmode="ak")

Expand All @@ -61,7 +61,7 @@ def f(lay, method=meth, **kwargs):
# this is different from dec_cu, because we need to instantiate StringMethods
# before getting the method from it
col = getattr(
StringMethods(cudf.Series(lay._to_cudf(cudf, None, len(lay)))), method
StringMethods(cudf.Series._from_column(lay._to_cudf(cudf, None, len(lay)))), method
)(**kwargs)
return from_cudf(col).layout

Expand All @@ -87,7 +87,7 @@ def f(lay, method=meth, **kwargs):
else:
# attributes giving components
col = m
return from_cudf(cudf.Series(col)).layout
return from_cudf(cudf.Series._from_column(col)).layout

if isinstance(getattr(DatetimeColumn, meth), property):
setattr(
Expand Down Expand Up @@ -118,7 +118,12 @@ def _to_output(cls, arr):

@classmethod
def to_array(cls, data) -> ak.Array:
return from_cudf(data)
if isinstance(data, cls.series_type):
return from_cudf(data)
out = {}
for col in data.columns:
out[col] = from_cudf(data[col])
return ak.Array(out)

@property
def array(self) -> ak.Array:
Expand Down Expand Up @@ -151,3 +156,4 @@ def ak_property(self):


Series.ak = ak_property # no official register function?
DataFrame.ak = ak_property # no official register function?
7 changes: 7 additions & 0 deletions tests/test_cudf.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -93,3 +93,10 @@ def test_times():
s = akimbo.io.ak_to_series(arr, "cudf")
s2 = s.ak.dt.second
assert s2.ak.to_list() == [[[0, 1, None, 2]], [], [[0, 1, None, 2]]]


def test_dataframe():
df = cudf.DataFrame({"a": [1, 2, 3], "b": [4, 5, 6]})
arr = df.ak.array
assert ak.backend(arr) == "cuda"
assert arr.fields == ["a", "b"]
Loading