Open-weight model
radiochat-laya-multilingual-onnx
by Shiva Prakash Foreverstorm/radiochat-laya-multilingual-onnx
radiochat-laya-multilingual-onnx is an open-weight model from Shiva Prakash, released under Apache License 2.0. Its published files total 312.0 MB.
A browser-ready export of convaiinnovations/laya-multilingual (revision e4e9ddf21a7b1903b7acffd8814ad4307bf63a67), the multilingual checkpoint of Laya by Convai Innovations — a non-autoregressive decision model that answers typed questions (choice, score…
Model Card
By Shiva Prakash, published under apache-2.0, revision 2151c98b1c63.
A browser-ready export of convaiinnovations/laya-multilingual (revision e4e9ddf21a7b1903b7acffd8814ad4307bf63a67), the multilingual checkpoint of Laya by Convai Innovations — a non-autoregressive decision model that answers typed questions (choice, score, yes/no) in one forward pass. All credit for the model goes to its authors; see the Laya repository. Licensed Apache-2.0, like the original. Used by RadioChat to moderate chat messages on the user's own device. The split layout that laya-ts loads: 1. laya-ts/scripts/exportonnx.py --repo convaiinnovations/laya-multilingual (laya commit 9d95567), which checks the ONNX outputs against PyTorch (max logit difference ~2e-6). 2. ONNX Runtime 1.30…
Read Shiva Prakash's full model card
Laya multilingual — ONNX, 8-bit (for in-browser moderation)
A browser-ready export of convaiinnovations/laya-multilingual
(revision e4e9ddf21a7b1903b7acffd8814ad4307bf63a67), the multilingual
checkpoint of Laya by Convai Innovations — a non-autoregressive decision
model that answers typed questions (choice, score, yes/no) in one forward
pass. All credit for the model goes to its authors; see the
Laya repository. Licensed Apache-2.0,
like the original.
Used by RadioChat to moderate chat messages on the user's own device.
Files
The split layout that laya-ts loads:
| file | size | contents |
|---|---|---|
encoder.onnx |
250 MB | mmBERT-base encoder — 8-bit weight-only layers (MatMulNBits, block 32), 4-bit vocabulary table (GatherBlockQuantized, block 32) |
head.onnx |
28 MB | decision head — 8-bit weight-only |
tokenizer.json, rl_agent_config.json |
unchanged from the checkpoint |
How it was made
laya-ts/scripts/export_onnx.py --repo convaiinnovations/laya-multilingual(laya commit9d95567), which checks the ONNX outputs against PyTorch (max logit difference ~2e-6).- ONNX Runtime 1.30
MatMulNBitsQuantizer(symmetric, block 32): MatMul weights 8-bit, the token-embedding Gather 4-bit; the head's MatMuls 8-bit. Each file is saved without external data (the browser loader fetches one file per graph).
Accuracy vs the full-precision model
Measured with laya-ts on onnxruntime-web (WASM) over 80 hand-labelled chat
messages × 2 moderation decisions (offensive?, spam?):
| variant | encoder | agrees with full model | correct vs labels |
|---|---|---|---|
| full precision | 1.17 GB | 160/160 | 152/160 |
| this (8-bit layers, 4-bit vocabulary, 8-bit head) | 250 MB | 158/160 | 150/160 |
| 4-bit layers | 195 MB | 153/160 | 146/160 |
4-bit layers cost clearly more accuracy; ONNX Runtime can't run 3-bit weights.
Identity and Version
- Repository
- Foreverstorm/radiochat-laya-multilingual-onnx
- Publisher
- Shiva Prakash
- Task
- Not stated by the source
- Modality
- Other
- Library
- onnx
- Parameters
- Not stated by the source
- Languages
- Not stated by the source
- Revision
- 2151c98b1c63b500b5efd1f498177820827fb118
- First published
- 2026-09-28
- Last updated
- 2026-09-28
Files and Weights
7 files, 312.0 MB in total. The weights are 2 files totalling 277.6 MB in onnx.
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| encoder.onnx | Weights | 249.7 MB | 33a6b6c75e9b |
| head.onnx | Weights | 27.9 MB | 5439bf827ac4 |
| rl_agent_config.json | Configuration | 472 B | — |
| LICENSE | Documentation | 10.3 KB | — |
| README.md | Documentation | 2.3 KB | — |
| .gitattributes | Repository | 1.6 KB | — |
| tokenizer.json | Tokenizer | 34.4 MB | 609d8f4c067c |
License and Download
- License
- apache-2.0
- Access
- Open weights, no gate
- Download size
- 277.6 MB
Released by Shiva Prakash through its official repository on Hugging Face. Read the license.
Built From
- Derived from convaiinnovations/laya-multilingual
- Quantized from convaiinnovations/laya-multilingual
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 277.6 MB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.
Questions About radiochat-laya-multilingual-onnx
Can I use radiochat-laya-multilingual-onnx commercially?
Yes. radiochat-laya-multilingual-onnx is released under Apache License 2.0. The Apache License 2.0 is a permissive open-source license. It permits commercial use, modification and redistribution. It requires keeping the license and copyright notices and any NOTICE file, stating significant changes, and it includes an express patent grant from contributors.