Ross ROSS = Recommend OSS · open-source software intelligence for agents

OpenDCAI/DataFlex

Data-centric LLM training with dynamic sample selection, domain mixture optimization, and example reweighting inside the LLaMA-Factory training loop. observed · 2026-08-28

github.com/OpenDCAI/DataFlex · homepage · Python · Apache-2.0 (permissive) observed · 2026-08-28

Health v2 · maintenance only

67/100

  • Activity 99
  • Release rhythm 48
  • Longevity 27
How is this computed?

round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10) — computed 2026-09-02. Adoption (stars, forks) is never an input.

  • gap_med: n/a
  • age_days: 389
  • days_rel: 139
  • days_push: 7
  • n_releases_24m: 1

Full methodology

Adoption not part of the score

2270 stars · 310 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

DataFlex is a data-centric training framework built on LLaMA-Factory that dynamically schedules LLM training data during optimization. It provides reproducible implementations of data selection, data mixture optimization, and example reweighting inside the training loop.

Use cases

  • dynamically select training samples for LLM fine-tuning
  • optimize domain mixture of training data
  • reweight training examples during model training
  • improve reproducibility of LLM training experiments
  • focus training on hard samples
  • train larger models with DeepSpeed ZeRO-3 gradient computation

When to choose

  • you already use LLaMA-Factory and want smarter data scheduling
  • you need reproducible data selection, mixture, and reweighting in one framework
  • you want to improve model performance by curating training data dynamically

When to avoid

  • you train models outside the LLaMA-Factory ecosystem
  • you need a general-purpose data pipeline unrelated to LLM training
  • you require a fully stable, long-proven production system

Facets

framework · maturity active

machine-learning llm-training data-science etl large-language-models machine-learning data-science deep-learning python data-selection data-mixture data-reweighting data-centric-ai llama-factory training-data-scheduling gpu linux

3 sources

Member repositories

RepositoryRoleHealth v2
OpenDCAI/DataFlexmain67

For agents

markdown · JSON · MCP: product_card(name="OpenDCAI/DataFlex")

Data as of 2026-08-30T08:39:29.467469+00:00 · Report a problem