CTRL-2B — ONNX (q4f16, WebGPU-ready)

CTRL-2B (QLoRA adapter on MiniCPM5-2B, merged) exported to ONNX fp16 and quantized to q4f16 (MatMulNBits, block 128, symmetric, accuracy_level 4) for in-browser WebGPU inference via transformers.js.

Run it live: CTRL-2B demo Space

Files: onnx/model_q4f16.onnx (+ external data), Inter/MiniCPM5 tokenizer with the CTRL-2B chat template embedded in tokenizer_config.json.

Downloads last month: 15

Model tree for expedient-ai/CTRL-2B-ONNX

Base model

openbmb/MiniCPM5-2B

Finetuned

expedient-ai/CTRL-2B

Quantized

(1)

this model

Collection including expedient-ai/CTRL-2B-ONNX

Expedient AI CTRL models

Collection

Retrieval-grounded models for the Expedient knowledge base • 5 items • Updated 2 days ago