From 0b66095e0b366296f12567c18364b7f922141b99 Mon Sep 17 00:00:00 2001 From: Dylan Uys Date: Tue, 6 Oct 2026 12:39:52 -0500 Subject: [PATCH 1/3] Increase full benchmark budgets for public datasets (#183) --- pyproject.toml | 2 +- src/gasbench/dataset/config.py | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/pyproject.toml b/pyproject.toml index a4243d8..7396950 100644 --- a/pyproject.toml +++ b/pyproject.toml @@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta" [project] name = "gasbench" -version = "0.11.1" +version = "0.11.2" description = "GASBench - ML model benchmark evaluation package" readme = "README.md" requires-python = ">=3.10" diff --git a/src/gasbench/dataset/config.py b/src/gasbench/dataset/config.py index b7d1e07..a5e2e59 100644 --- a/src/gasbench/dataset/config.py +++ b/src/gasbench/dataset/config.py @@ -25,7 +25,7 @@ BENCHMARK_TOTAL_OVERRIDES = { "debug": {"image": 100, "video": 50, "audio": 50}, "small": {"image": 5000, "video": 5000, "audio": 5000}, - "full": {"image": 72500, "video": 34000, "audio": 47500}, + "full": {"image": 75000, "video": 35000, "audio": 50000}, } # Per-dataset download limits (only applied in debug/small modes for faster testing) From a95e5aa0c4acc2a8b7232eabe57522a5cc6a756b Mon Sep 17 00:00:00 2001 From: dylan Date: Tue, 6 Oct 2026 15:21:32 -0500 Subject: [PATCH 2/3] Add pinned 34data audio and video datasets --- .../dataset/configs/synthetic_audio.yaml | 57 +++++++++++++++++++ .../dataset/configs/synthetic_videos.yaml | 25 ++++++++ 2 files changed, 82 insertions(+) diff --git a/src/gasbench/dataset/configs/synthetic_audio.yaml b/src/gasbench/dataset/configs/synthetic_audio.yaml index d76cca2..2f89fb0 100644 --- a/src/gasbench/dataset/configs/synthetic_audio.yaml +++ b/src/gasbench/dataset/configs/synthetic_audio.yaml @@ -986,3 +986,60 @@ datasets: - audio modality: audio media_type: synthetic + + - name: uncovai-dia2-2b + path: 34data/uncovai-dia2-2b + hf_revision: 22920d224a354fd23d1356ce5ba11be3cdab7261 + source_format: zip + media_per_archive: -1 + archives_per_dataset: -1 + modality: audio + media_type: synthetic + - name: uncovai-maya1 + path: 34data/uncovai-maya1 + hf_revision: 801bd729c2d932380640ab451605078da66b5489 + source_format: zip + media_per_archive: -1 + archives_per_dataset: -1 + modality: audio + media_type: synthetic + - name: voxenes-tts-chatterbox + path: 34data/voxenes-tts-chatterbox + hf_revision: 47df4526bb834f8871d526b88800e4483f9b1101 + source_format: zip + media_per_archive: -1 + archives_per_dataset: -1 + modality: audio + media_type: synthetic + - name: voxenes-tts-chroma + path: 34data/voxenes-tts-chroma + hf_revision: 81cc3deaa1ace06dc77dfa0620ed3e8aedf10268 + source_format: zip + media_per_archive: -1 + archives_per_dataset: -1 + modality: audio + media_type: synthetic + - name: voxenes-tts-glm-tts + path: 34data/voxenes-tts-glm-tts + hf_revision: 628746214963bc7789722f7911cf21990d03dd5c + source_format: zip + media_per_archive: -1 + archives_per_dataset: -1 + modality: audio + media_type: synthetic + - name: voxenes-tts-vibevoice + path: 34data/voxenes-tts-vibevoice + hf_revision: 30e8165129a4af7fc476dfc9cae264a7fc87e1ff + source_format: zip + media_per_archive: -1 + archives_per_dataset: -1 + modality: audio + media_type: synthetic + - name: voxenes-tts-voxcpm + path: 34data/voxenes-tts-voxcpm + hf_revision: f8b2da4246ca4530aa65b0454ffa69fdb70746da + source_format: zip + media_per_archive: -1 + archives_per_dataset: -1 + modality: audio + media_type: synthetic diff --git a/src/gasbench/dataset/configs/synthetic_videos.yaml b/src/gasbench/dataset/configs/synthetic_videos.yaml index 91f731e..c5a3e42 100644 --- a/src/gasbench/dataset/configs/synthetic_videos.yaml +++ b/src/gasbench/dataset/configs/synthetic_videos.yaml @@ -1336,3 +1336,28 @@ datasets: source_format: mp4 modality: video media_type: synthetic + + - name: ueval-t2v-per-category-gen3-202407 + path: 34data/ueval-t2v-gen3-202407 + hf_revision: c96859435b382c407a00d514ba29e1d97e7a5016 + source_format: zip + media_per_archive: -1 + archives_per_dataset: -1 + modality: video + media_type: synthetic + - name: javedit-100k + path: 34data/javedit-100k-speech-editing + hf_revision: af5bc6bc9c857ad8f800d0769fcfb0a00bfac125 + source_format: zip + media_per_archive: -1 + archives_per_dataset: -1 + modality: video + media_type: semisynthetic + - name: genvideo-50k-shard + path: 34data/genvideo-50k-shard-fake + hf_revision: 9aed79ccdfef8179d611b2c61f8c72fcebbc05f0 + source_format: zip + media_per_archive: -1 + archives_per_dataset: -1 + modality: video + media_type: synthetic From addb9f239c95da27f1d024f3811f191c43cddbdc Mon Sep 17 00:00:00 2001 From: dylan Date: Tue, 6 Oct 2026 15:26:17 -0500 Subject: [PATCH 3/3] Bump GasBench version to 0.11.3 --- pyproject.toml | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/pyproject.toml b/pyproject.toml index 7396950..efc9393 100644 --- a/pyproject.toml +++ b/pyproject.toml @@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta" [project] name = "gasbench" -version = "0.11.2" +version = "0.11.3" description = "GASBench - ML model benchmark evaluation package" readme = "README.md" requires-python = ">=3.10"