Grigori Fursin
commited on
Add version 2.0 with benchmark name, version, and timestamp
Browse files- README.md +54 -54
- data.json +0 -0
- data.parquet +2 -2
- processor.py → process.py +53 -1
README.md
CHANGED
@@ -1,54 +1,54 @@
|
|
1 |
-
---
|
2 |
-
license: apache-2.0
|
3 |
-
---
|
4 |
-
|
5 |
-
# Preparing OpenMLPerf dataset
|
6 |
-
|
7 |
-
To process the semi-raw MLPerf data into the OpenMLPerf dataset, run the following command:
|
8 |
-
|
9 |
-
```bash
|
10 |
-
# Untar raw files
|
11 |
-
|
12 |
-
bzip2 -d semi-raw-mlperf-data.tar.bz2
|
13 |
-
tar xvf semi-raw-mlperf-data.tar
|
14 |
-
|
15 |
-
# Create a virtual environment
|
16 |
-
python -m venv .venv
|
17 |
-
|
18 |
-
# Activate the virtual environment
|
19 |
-
source .venv/bin/activate
|
20 |
-
|
21 |
-
# Install the required packages
|
22 |
-
pip install -r requirements.txt
|
23 |
-
|
24 |
-
# Run the processing script
|
25 |
-
python process.py
|
26 |
-
```
|
27 |
-
|
28 |
-
The processed dataset will be saved both as `data.json` and `data.parquet` in the `OpenMLPerf-dataset` directory.
|
29 |
-
The `data.json` file is a JSON file containing the processed data, while the `data.parquet` file is a Parquet file containing the same data in a more efficient format for storage and processing.
|
30 |
-
|
31 |
-
# Preprocessing raw MLPerf results using MLCommons CMX
|
32 |
-
|
33 |
-
We preprocess official raw MLPerf data, such as [inference v5.0](https://github.com/mlcommons/inference_results_v5.0),
|
34 |
-
into semi-raw format compatible with the `process.py` script, using the [MLCommons CM/CMX automation framework](https://arxiv.org/abs/2406.16791).
|
35 |
-
This is done using through the ["import mlperf results"](https://github.com/mlcommons/ck/tree/master/cmx4mlops/repo/flex.task/import-mlperf-results)
|
36 |
-
automation action, which we plan to document in more detail soon.
|
37 |
-
|
38 |
-
# License and Copyright
|
39 |
-
|
40 |
-
This project is licensed under the [Apache License 2.0](LICENSE.md).
|
41 |
-
|
42 |
-
© 2025 FlexAI
|
43 |
-
|
44 |
-
Portions of the data were adapted from the following MLCommons repositories,
|
45 |
-
which are also licensed under the Apache 2.0 license:
|
46 |
-
|
47 |
-
* [mlcommons@inference_results_v5.0](https://github.com/mlcommons/inference_results_v5.0)
|
48 |
-
* [mlcommons@inference_results_v4.1](https://github.com/mlcommons/inference_results_v4.1)
|
49 |
-
* [mlcommons@inference_results_v4.0](https://github.com/mlcommons/inference_results_v4.0)
|
50 |
-
* [mlcommons@inference_results_v3.1](https://github.com/mlcommons/inference_results_v3.1)
|
51 |
-
|
52 |
-
# Authors and maintaners
|
53 |
-
|
54 |
-
[Daniel Altunay](https://www.linkedin.com/in/daltunay) and [Grigori Fursin](https://cKnowledge.org/gfursin) (FCS Labs)
|
|
|
1 |
+
---
|
2 |
+
license: apache-2.0
|
3 |
+
---
|
4 |
+
|
5 |
+
# Preparing OpenMLPerf dataset
|
6 |
+
|
7 |
+
To process the semi-raw MLPerf data into the OpenMLPerf dataset, run the following command:
|
8 |
+
|
9 |
+
```bash
|
10 |
+
# Untar raw files
|
11 |
+
|
12 |
+
bzip2 -d semi-raw-mlperf-data.tar.bz2
|
13 |
+
tar xvf semi-raw-mlperf-data.tar
|
14 |
+
|
15 |
+
# Create a virtual environment
|
16 |
+
python -m venv .venv
|
17 |
+
|
18 |
+
# Activate the virtual environment
|
19 |
+
source .venv/bin/activate
|
20 |
+
|
21 |
+
# Install the required packages
|
22 |
+
pip install -r requirements.txt
|
23 |
+
|
24 |
+
# Run the processing script
|
25 |
+
python process.py
|
26 |
+
```
|
27 |
+
|
28 |
+
The processed dataset will be saved both as `data.json` and `data.parquet` in the `OpenMLPerf-dataset` directory.
|
29 |
+
The `data.json` file is a JSON file containing the processed data, while the `data.parquet` file is a Parquet file containing the same data in a more efficient format for storage and processing.
|
30 |
+
|
31 |
+
# Preprocessing raw MLPerf results using MLCommons CMX
|
32 |
+
|
33 |
+
We preprocess official raw MLPerf data, such as [inference v5.0](https://github.com/mlcommons/inference_results_v5.0),
|
34 |
+
into semi-raw format compatible with the `process.py` script, using the [MLCommons CM/CMX automation framework](https://arxiv.org/abs/2406.16791).
|
35 |
+
This is done using through the ["import mlperf results"](https://github.com/mlcommons/ck/tree/master/cmx4mlops/repo/flex.task/import-mlperf-results)
|
36 |
+
automation action, which we plan to document in more detail soon.
|
37 |
+
|
38 |
+
# License and Copyright
|
39 |
+
|
40 |
+
This project is licensed under the [Apache License 2.0](LICENSE.md).
|
41 |
+
|
42 |
+
© 2025 FlexAI
|
43 |
+
|
44 |
+
Portions of the data were adapted from the following MLCommons repositories,
|
45 |
+
which are also licensed under the Apache 2.0 license:
|
46 |
+
|
47 |
+
* [mlcommons@inference_results_v5.0](https://github.com/mlcommons/inference_results_v5.0)
|
48 |
+
* [mlcommons@inference_results_v4.1](https://github.com/mlcommons/inference_results_v4.1)
|
49 |
+
* [mlcommons@inference_results_v4.0](https://github.com/mlcommons/inference_results_v4.0)
|
50 |
+
* [mlcommons@inference_results_v3.1](https://github.com/mlcommons/inference_results_v3.1)
|
51 |
+
|
52 |
+
# Authors and maintaners
|
53 |
+
|
54 |
+
[Daniel Altunay](https://www.linkedin.com/in/daltunay) and [Grigori Fursin](https://cKnowledge.org/gfursin) (FCS Labs)
|
data.json
CHANGED
The diff for this file is too large to render.
See raw diff
|
|
data.parquet
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:a2d83fecee4037e9a4e10db98d79096c383319739d9a86b7d5e38b29e0fa054b
|
3 |
+
size 44347
|
processor.py → process.py
RENAMED
@@ -1,4 +1,6 @@
|
|
1 |
-
"""
|
|
|
|
|
2 |
|
3 |
import glob
|
4 |
import json
|
@@ -6,6 +8,7 @@ import logging
|
|
6 |
import os
|
7 |
import re
|
8 |
from collections import defaultdict
|
|
|
9 |
|
10 |
import polars as pl
|
11 |
from datasets import Dataset
|
@@ -135,6 +138,10 @@ def load_raw_data(base_path: str = "semi-raw-mlperf-data") -> pl.DataFrame:
|
|
135 |
"system.host_processor_frequency": "system.cpu.frequency",
|
136 |
"system.host_processor_caches": "system.cpu.caches",
|
137 |
"system.host_processor_vcpu_count": "system.cpu.vcpu_count",
|
|
|
|
|
|
|
|
|
138 |
}
|
139 |
|
140 |
for old_name, new_name in rename_map.items():
|
@@ -183,6 +190,39 @@ def find_similar_configurations(
|
|
183 |
return df.filter(mask)
|
184 |
|
185 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
186 |
def convert_memory_to_gb(value: str) -> float | None:
|
187 |
"""Convert memory string to GB."""
|
188 |
if value is None:
|
@@ -382,6 +422,17 @@ def normalize_memory_values(df: pl.DataFrame) -> pl.DataFrame:
|
|
382 |
)
|
383 |
|
384 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
385 |
def add_vendor_columns(df: pl.DataFrame) -> pl.DataFrame:
|
386 |
"""Add vendor columns based on model names."""
|
387 |
return df.with_columns(
|
@@ -575,6 +626,7 @@ def process_data(base_path: str = "semi-raw-mlperf-data") -> pl.DataFrame:
|
|
575 |
load_raw_data(base_path)
|
576 |
.pipe(clean_string_values)
|
577 |
.pipe(normalize_memory_values)
|
|
|
578 |
.pipe(cast_columns)
|
579 |
.pipe(add_vendor_columns)
|
580 |
.pipe(normalize_interconnect_values)
|
|
|
1 |
+
"""
|
2 |
+
Data processing module for MLPerf benchmark data.
|
3 |
+
"""
|
4 |
|
5 |
import glob
|
6 |
import json
|
|
|
8 |
import os
|
9 |
import re
|
10 |
from collections import defaultdict
|
11 |
+
from datetime import datetime
|
12 |
|
13 |
import polars as pl
|
14 |
from datasets import Dataset
|
|
|
138 |
"system.host_processor_frequency": "system.cpu.frequency",
|
139 |
"system.host_processor_caches": "system.cpu.caches",
|
140 |
"system.host_processor_vcpu_count": "system.cpu.vcpu_count",
|
141 |
+
"benchmark_name": "benchmark.name",
|
142 |
+
"benchmark_version": "benchmark.version",
|
143 |
+
"datetime_last_commit": "datetime",
|
144 |
+
"debug_uid": "debug_uid",
|
145 |
}
|
146 |
|
147 |
for old_name, new_name in rename_map.items():
|
|
|
190 |
return df.filter(mask)
|
191 |
|
192 |
|
193 |
+
def convert_datetime_to_iso(value: str) -> str | None:
|
194 |
+
"""Convert datetime string to ISO 8601 format."""
|
195 |
+
if not value or value in ["", "N/A", "null"]:
|
196 |
+
MISSING_VALUES["datetime_values"].add(str(value))
|
197 |
+
return None
|
198 |
+
|
199 |
+
try:
|
200 |
+
# Handle format like "2025/04/03_22:56:53"
|
201 |
+
if "/" in value and "_" in value:
|
202 |
+
# Replace / with - and _ with T for ISO format
|
203 |
+
iso_value = value.replace("/", "-").replace("_", "T")
|
204 |
+
# Validate by parsing
|
205 |
+
datetime.fromisoformat(iso_value)
|
206 |
+
return iso_value
|
207 |
+
|
208 |
+
# Try to parse other common formats and convert to ISO
|
209 |
+
# Add more format patterns as needed
|
210 |
+
for fmt in ["%Y-%m-%d %H:%M:%S", "%Y/%m/%d %H:%M:%S", "%Y-%m-%dT%H:%M:%S"]:
|
211 |
+
try:
|
212 |
+
dt = datetime.strptime(value, fmt)
|
213 |
+
return dt.isoformat()
|
214 |
+
except ValueError:
|
215 |
+
continue
|
216 |
+
|
217 |
+
# If no format matches, log as missing value
|
218 |
+
MISSING_VALUES["datetime_values"].add(str(value))
|
219 |
+
return None
|
220 |
+
|
221 |
+
except Exception as e:
|
222 |
+
MISSING_VALUES["datetime_values"].add(str(value))
|
223 |
+
return None
|
224 |
+
|
225 |
+
|
226 |
def convert_memory_to_gb(value: str) -> float | None:
|
227 |
"""Convert memory string to GB."""
|
228 |
if value is None:
|
|
|
422 |
)
|
423 |
|
424 |
|
425 |
+
def normalize_datetime_values(df: pl.DataFrame) -> pl.DataFrame:
|
426 |
+
"""Convert datetime values to ISO 8601 format."""
|
427 |
+
if "datetime" in df.columns:
|
428 |
+
return df.with_columns(
|
429 |
+
pl.col("datetime")
|
430 |
+
.map_elements(convert_datetime_to_iso, return_dtype=str)
|
431 |
+
.alias("datetime")
|
432 |
+
)
|
433 |
+
return df
|
434 |
+
|
435 |
+
|
436 |
def add_vendor_columns(df: pl.DataFrame) -> pl.DataFrame:
|
437 |
"""Add vendor columns based on model names."""
|
438 |
return df.with_columns(
|
|
|
626 |
load_raw_data(base_path)
|
627 |
.pipe(clean_string_values)
|
628 |
.pipe(normalize_memory_values)
|
629 |
+
.pipe(normalize_datetime_values)
|
630 |
.pipe(cast_columns)
|
631 |
.pipe(add_vendor_columns)
|
632 |
.pipe(normalize_interconnect_values)
|