Codex
Add model card metadata
61a3822
---
base_model: GSAI-ML/LLaDA-1.5
tags:
- llada
- diffusion-language-model
- lora
- checkpoints
- diffusiongrow
---
# EMNLP LLaDA-1.5 DiffusionGrow Checkpoints
This repository stores checkpoint artifacts for the LLaDA-1.5 DiffusionGrow extension runs.
Backbone: `GSAI-ML/LLaDA-1.5`
Included domains:
- `coding`: CodeParrot clean domain adaptation runs.
- `curated_math`: OpenWebMath-derived curated math runs.
Included methods:
- `lora`
- `dg_lora`
Included phases:
- `smoke`: 64 steps
- `pilot`: 1,000 steps
- `full`: 5,000 steps
Checkpoint paths mirror the local experiment tree under:
`aclw-2026/llada15_extension/**/checkpoints/*.pt`
See `CHECKPOINT_MANIFEST.txt` for the full file list and byte sizes.