Dataset Versioning And Reproducible Splits With Fixed Random Seeds

via “dataset versioning and reproducible splits”

250GB curated code dataset for StarCoder training.

Unique: Provides versioned, reproducible splits with transparent curation metadata, enabling researchers to understand exactly which code samples were used and how they were selected. Supports ablation studies on filtering steps.

vs others: More reproducible than ad-hoc dataset creation and more transparent than proprietary datasets like Codex. Enables fair comparison across research papers and models trained on the same data.

via “dataset versioning and reproducibility”

70K commonsense reasoning questions with adversarial distractors.

Unique: Provides a fixed, versioned dataset on Hugging Face with explicit train/validation/test splits, enabling reproducible evaluation and fair comparison across models. The fixed nature ensures that improvements reflect genuine capability gains rather than dataset variance or adversarial augmentation at test time.

vs others: More reproducible than dynamically-generated benchmarks because the dataset is fixed and versioned, and more comparable than benchmarks with multiple variants because all researchers use the same evaluation set.

via “deterministic generation with seed control and reproducibility”

text-to-image model by undefined. 20,41,667 downloads.

Unique: Implements seed control at scheduler level, ensuring reproducibility across PyTorch, ONNX, and different hardware; supports seed ranges for deterministic batch variation without requiring separate model invocations

vs others: More reliable than manual random state management; comparable to other diffusion models but with explicit reproducibility guarantees and documentation

via “reproducible output generation with seed parameter”

Enhanced GPT-4 with 128K context and improved speed.

Unique: Exposes seed parameter at the API level to control the random number generator used in token sampling, enabling reproducible outputs without requiring model retraining or checkpoint management

vs others: Provides reproducibility guarantees that Anthropic Claude lacks (no seed parameter support), enabling deterministic testing workflows that are impossible with non-seeded models

via “seed-based reproducible generation”

text-to-image model by undefined. 6,21,488 downloads.

Unique: Implements seed-based reproducibility via PyTorch's generator object, enabling deterministic generation without modifying model weights or architecture. Seed controls both latent initialization and timestep sampling.

vs others: Standard approach across ML frameworks; enables reproducible research and testing comparable to proprietary services.

via “seed-based reproducible generation for deterministic outputs”

text-to-image model by undefined. 6,08,507 downloads.

Unique: Integrates seed-based reproducibility into the diffusers pipeline, enabling deterministic generation by controlling noise initialization and scheduler randomness; the same seed produces identical outputs across runs (within floating-point precision), unlike some proprietary models that do not expose seed control

vs others: More reproducible than models without seed control (e.g., some cloud-based APIs), but less reproducible than fully deterministic algorithms due to floating-point precision variations; enables testing and validation that non-reproducible models cannot support

via “reproducible generation with seed control and deterministic inference”

🔥 [ICCV 2025 Highlight] InfiniteYou: Flexible Photo Recrafting While Preserving Your Identity

Unique: Implements comprehensive seed management across the entire pipeline (PyTorch, NumPy, random) to ensure deterministic generation, critical for research and evaluation workflows.

vs others: More reliable than ad-hoc seed setting; ensures reproducibility across the entire codebase rather than just the diffusion sampler.

via “seed-based reproducible generation with deterministic sampling”

text-to-video model by undefined. 39,484 downloads.

Unique: Implements seed-based reproducibility by controlling all sources of randomness in the diffusion pipeline (noise initialization, dropout, stochastic depth) through PyTorch's global random state. This approach ensures bit-exact reproducibility within the same environment while remaining transparent to users.

vs others: Simpler and more transparent than checkpoint-based reproducibility (no need to save intermediate states), while providing stronger guarantees than probabilistic reproducibility approaches.

via “seed-based reproducible generation with deterministic randomness”

Just playing with getting VQGAN+CLIP running locally, rather than having to use colab.

Unique: Implements comprehensive seed-based reproducibility by controlling random number generation across PyTorch, NumPy, and Python's built-in random module, ensuring identical results across runs with identical seeds and hyperparameters. Extends seed control to all stochastic components including latent initialization and augmentation.

vs others: Enables true reproducibility unlike non-seeded generation, but with caveats around hardware/software dependencies; similar to other seeded generative models but with explicit control over all randomness sources.

via “reproducible generation with seed control and deterministic sampling”

VideoCrafter2: Overcoming Data Limitations for High-Quality Video Diffusion Models

Unique: Combines seed control with deterministic DDIM sampling (eta=0) to ensure reproducible generation. Enables users to generate identical videos for debugging and testing.

vs others: Seed control is standard in diffusion models; deterministic DDIM sampling enables reproducibility without sacrificing quality; enables reproducible research and testing unlike stochastic-only approaches.

via “dataset splitting and train/validation/test partitioning with stratification”

[Slack](https://camel-kwr1314.slack.com/join/shared_invite/zt-1vy8u9lbo-ZQmhIAyWSEfSwLCl2r2eKA#/shared-invite/email)

Unique: Implements stratified splitting using Arrow's compute kernels for efficient label distribution analysis, and supports temporal splitting with automatic time-based ordering. Uses deterministic hashing for reproducible random splits across different machines.

vs others: More efficient than scikit-learn's train_test_split for large datasets because it operates on Arrow-backed data without materializing in memory, and more flexible because it supports temporal and custom splitting strategies.

via “seed management and reproducibility control”

Stableboost is a Stable Diffusion WebUI that lets you quickly generate a lot of images so you can find the perfect ones.

Unique: Provides explicit seed tracking and management in the UI, making seed values first-class parameters that users can control and inspect, rather than hidden implementation details

vs others: More reproducible than manual seed tracking because seeds are automatically captured and displayed with each image, enabling users to recreate specific outputs without manual note-taking

via “dataset splitting and train/test/validation partitioning”

HuggingFace community-driven open-source library of datasets

Unique: Implements deterministic splitting with optional stratification, returning a DatasetDict for easy access to splits. The system integrates with the fingerprinting system to ensure reproducible splits across runs.

vs others: More convenient than scikit-learn's train_test_split for dataset objects; supports stratification natively; integrates with dataset pipeline unlike external splitting tools.

via “seed-based reproducible generation”

TRELLIS.2 — AI demo on HuggingFace

Unique: Exposes seed control directly in the Gradio UI rather than hiding it in API parameters, making reproducibility a first-class feature accessible to non-technical users and enabling collaborative workflows without requiring API documentation

vs others: More discoverable than API-only seed control, though less flexible than programmatic access for systematic seed sweeps

via “task-specific train/validation/test split provisioning”

Dataset by nyu-mll. 3,97,160 downloads.

Unique: Implements fixed, peer-reviewed splits across 9 tasks with documented random seeds and class balance constraints, enabling exact reproduction of published results — unlike ad-hoc dataset splits that vary across implementations. Integrates with HuggingFace Datasets' lazy-loading architecture to avoid materializing full splits in memory until needed.

vs others: Eliminates split variance that plagues custom benchmarks by providing official, immutable partitions used in 1000+ published papers, reducing experimental variance from data leakage and enabling fair cross-paper comparisons unlike task-specific datasets with inconsistent split definitions.

via “dataset versioning and reproducible snapshot loading”

Dataset by lavita. 5,55,826 downloads.

Unique: Leverages HuggingFace Hub's Git-based versioning infrastructure to provide immutable dataset snapshots with full history tracking. Enables citation-grade reproducibility through semantic versioning and automatic version pinning in code.

vs others: More reproducible than ad-hoc dataset downloads because versions are immutable and citable; better than manual versioning because Git history is automatically maintained and queryable

via “dataset-versioning-and-reproducible-snapshot-management”

Dataset by Rowan. 3,02,991 downloads.

Unique: Leverages HuggingFace Hub's Git-based versioning to provide immutable dataset snapshots with automatic caching and rollback support, without requiring separate version control infrastructure

vs others: More convenient than manual dataset versioning (Git, DVC) and simpler than data warehouse versioning, with tight integration to HuggingFace's ecosystem and automatic caching

Dataset by bigcode. 4,30,889 downloads.

Unique: Implements immutable versioned snapshots with fixed random seeds and pre-computed splits, enabling bit-for-bit reproducible dataset loading across machines and time — most datasets lack version control or use non-deterministic sampling

vs others: Enables reproducible research by eliminating randomness in data splits; simplifies citation and comparison across papers; maintains backward compatibility with older versions

via “reproducible train-test split generation”

Dataset by m-a-p. 4,59,057 downloads.

Unique: Leverages HuggingFace's dataset versioning and deterministic sampling to ensure splits are reproducible across runs, environments, and teams; integrates with the datasets library's native .train_test_split() API for seamless integration into training pipelines

vs others: More reproducible than manual splitting (which is error-prone) and more transparent than proprietary benchmark splits (which hide methodology); seed-based approach enables both reproducibility and statistical rigor via multiple independent splits

via “dataset splitting and train/validation/test set management”

Intuitive app to build your own AI models. Includes no-code synthetic data generation, fine-tuning, dataset collaboration, and more.