Open-weight model
ebc-jepa-vjepa-lora-ed24-denoiser
by Henry h3lloworld/ebc-jepa-vjepa-lora-ed24-denoiser
ebc-jepa-vjepa-lora-ed24-denoiser is an open-weight model from Henry, released under MIT License. Its published files total 7.6 MB.
Event tokenizer + V-JEPA 2.1 ViT-B with LoRA (qkv, proj; r8) + per-event head; 4096 events per inference; trained on all ED24 with the denoising loss.
Model Card
By Henry, published under mit, revision 8271c91aceaa.
Event tokenizer + V-JEPA 2.1 ViT-B with LoRA (qkv, proj; r8) + per-event head; 4096 events per inference; trained on all ED24 with the denoising loss. Code: https://github.com/whohyf/EBC-JEPA-share/tree/denoise-lora (src/ebcjepa/loradenoise, tools/loradenoise; table protocol in docs/loradenoise/HANDOFF.md). Training data: ED24 (EDformer, all 2,100 official files); generalisation test: DND21, E-MLB. - trainable.pt - metrics.json - config.json trainable.pt holds only the trained tensors (event tokenizer projection, head, LoRA); they load on top of the public V-JEPA 2.1 ViT-B checkpoint (https://dl.fbaipublicfiles.com/vjepa2/vjepa21vitbdistvitG384.pt). Evaluate with…
Read Henry's full model card
EBC-JEPA V-JEPA-only denoiser, LoRA ViT-B (ED24)
Event tokenizer + V-JEPA 2.1 ViT-B with LoRA (qkv, proj; r8) + per-event head; 4096 events per inference; trained on all ED24 with the denoising loss.
Code: https://github.com/whohyf/EBC-JEPA-share/tree/denoise-lora (src/ebc_jepa/lora_denoise, tools/lora_denoise; table protocol in docs/lora_denoise/HANDOFF.md).
Training data: ED24 (EDformer, all 2,100 official files); generalisation test: DND21, E-MLB.
Files
trainable.ptmetrics.jsonconfig.json
Use
trainable.pt holds only the trained tensors (event tokenizer projection, head, LoRA); they load on top of the public V-JEPA 2.1 ViT-B checkpoint (https://dl.fbaipublicfiles.com/vjepa2/vjepa2_1_vitb_dist_vitG_384.pt). Evaluate with tools/lora_denoise/eval_emlb.py --runs-dir <dir holding this folder> --arms <folder name> --checkpoint vjepa2_1_vitb_dist_vitG_384.pt.
License
MIT. Builds on V-JEPA 2 / 2.1 by Meta (MIT, https://dl.fbaipublicfiles.com/vjepa2/vjepa2_1_vitb_dist_vitG_384.pt).
Identity and Version
- Repository
- h3lloworld/ebc-jepa-vjepa-lora-ed24-denoiser
- Publisher
- Henry
- Task
- Not stated by the source
- Modality
- Other
- Library
- Not stated by the source
- Parameters
- Not stated by the source
- Languages
- ebc-jepa
- Revision
- 8271c91aceaac571f27b5c62b0cad0efdc821d23
- First published
- 2026-10-01
- Last updated
- 2026-10-01
Files and Weights
5 files, 7.6 MB in total. The weights are 1 file totalling 7.6 MB in pt.
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| trainable.pt | Weights | 7.6 MB | ab1d858b432e |
| config.json | Configuration | 1.3 KB | — |
| metrics.json | Configuration | 64.9 KB | — |
| README.md | Documentation | 1.1 KB | — |
| .gitattributes | Repository | 1.5 KB | — |
License and Download
- License
- mit
- Access
- Open weights, no gate
- Download size
- 7.6 MB
Released by Henry through its official repository on Hugging Face. Read the license.
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 7.6 MB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.
Questions About ebc-jepa-vjepa-lora-ed24-denoiser
Can I use ebc-jepa-vjepa-lora-ed24-denoiser commercially?
Yes. ebc-jepa-vjepa-lora-ed24-denoiser is released under MIT License. The MIT License is a short permissive license. It permits commercial use, modification and redistribution, provided the copyright notice and permission notice are included.