Merge details
Table | |
|---|
| Merge method | Linear |
| Models merged | Qwen/Qwen3-4B-Instruct-2507 (weight 0.5), a continually-pretrained Icelandic Qwen3-4B-Base checkpoint (weight 0.5, not publicly released) |
| CPT training data | Icelandic Gigaword Corpus (IGC) |
| dtype | bfloat16 |
| Tokenizer | Qwen/Qwen3-4B-Instruct-2507 |
models:
- model: Qwen/Qwen3-4B-Instruct-2507
parameters:
weight: 0.5
- model: <local CPT Icelandic Qwen3-4B-Base checkpoint>
parameters:
weight: 0.5
merge_method: linear
dtype: bfloat16
tokenizer_source: "Qwen/Qwen3-4B-Instruct-2507"
Intended use
Use this checkpoint together with the
LoRA adapter
for English→Icelandic machine translation, e.g. served with
vLLM:
vllm serve arnastofnun/Qwen3-4B-wmt26-AMI-en-is \
--enable-lora \
--lora-modules translate-lora=arnastofnun/Qwen3-4B-wmt26-AMI-en-is-lora
Citation
If you use this model, please cite:
@inproceedings{steingrimsson2026drag,
title = {What a DRAG (It Is Being Small) - The AMI Submission to the WMT 2026 General Translation Shared Task},
author = {Steingr{\'i}msson, Stein{\th}{\'o}r and {\TH}{\'o}r{\dh}arson, Sveinbj{\"o}rn and Da{\dh}ason, J{\'o}n Fri{\dh}rik},
booktitle = {Proceedings of the Eleventh Conference on Machine Translation},
month = oct,
year = {2026},
address = {Budapest, Hungary},
publisher = {Association for Computational Linguistics}
}