source languages: nl; target languages: en; OPUS readme: nl-en; dataset: opus; model: transformer-align; pre-processing: normalization + SentencePiece.
qwen3-4b-en2zh-orpo is an open-weight model for translation from Tatasauce, released under Apache License 2.0. Its published files total 275.8 MB.
模型類型: LoRA Adapter (SFT + ORPO 累積更新) - 基礎模型: unsloth/Qwen3-4B-Base-unsloth-bnb-4bit - SFT 模型來源: tatasauce4life/qwen3-4b-en2zh-lora - 訓練方法: ORPO (Odds Ratio Preference Optimization) 1. SFT 階段: 使用英中平行小說語料做 LoRA SFT 2. ORPO 階段: 使用人工譯文 (chosen) vs.
Parameters—
Context—
Weights264.3 MB
Licenseapache-2.0
AccessOpen weights
Monthly Downloads—
Model Card
By Tatasauce, published under apache-2.0, revision c0e12dd4c5a1.
模型類型: LoRA Adapter (SFT + ORPO 累積更新) - 基礎模型: unsloth/Qwen3-4B-Base-unsloth-bnb-4bit - SFT 模型來源: tatasauce4life/qwen3-4b-en2zh-lora - 訓練方法: ORPO (Odds Ratio Preference Optimization) 1. SFT 階段: 使用英中平行小說語料做 LoRA SFT 2. ORPO 階段: 使用人工譯文 (chosen) vs. 機器翻譯 (rejected) 做偏好優化 此模型是 LoRA adapter，包含 SFT + ORPO 的累積更新。
Read Tatasauce's full model card
Qwen3-4B 英翻中小說翻譯模型 (ORPO)
概述
- 模型類型: LoRA Adapter (SFT + ORPO 累積更新)
- 基礎模型:
unsloth/Qwen3-4B-Base-unsloth-bnb-4bit - SFT 模型來源:
tatasauce4life/qwen3-4b-en2zh-lora - 訓練方法: ORPO (Odds Ratio Preference Optimization)
- 任務: 英文到中文小說翻譯
訓練流程
- SFT 階段: 使用英中平行小說語料做 LoRA SFT
- ORPO 階段: 使用人工譯文 (chosen) vs. 機器翻譯 (rejected) 做偏好優化
ORPO 資料格式
{
"prompt": "英文原文",
"chosen": "人工譯文（高品質）",
"rejected": "機器翻譯（低品質）"
}
翻譯模板
### English
{english_text}
### Chinese
{chinese_translation}
主要訓練參數
| 參數 | 值 |
|---|---|
| ORPO beta | 0.1 |
| Learning Rate | 5e-06 |
| Epochs | 3 |
| Batch Size | 1 |
| Gradient Accumulation | 16 |
| Max Sequence Length | 1024 |
| LoRA r | 16 |
| LoRA alpha | 32 |
如何載入模型
此模型是 LoRA adapter，包含 SFT + ORPO 的累積更新。
使用 Unsloth（推薦）
from unsloth import FastLanguageModel
model, tokenizer = FastLanguageModel.from_pretrained(
model_name="tatasauce4life/qwen3-4b-en2zh-orpo",
max_seq_length=1024,
load_in_4bit=True,
)
FastLanguageModel.for_inference(model)
使用 PEFT
from peft import AutoPeftModelForCausalLM
from transformers import AutoTokenizer
model = AutoPeftModelForCausalLM.from_pretrained(
"tatasauce4life/qwen3-4b-en2zh-orpo",
load_in_4bit=True,
)
tokenizer = AutoTokenizer.from_pretrained("tatasauce4life/qwen3-4b-en2zh-orpo")
注意: 此 adapter 的 base model 為
unsloth/Qwen3-4B-Base-unsloth-bnb-4bit。 adapter 包含 SFT 階段與 ORPO 階段的累積權重更新。
使用範例
prompt = "### English\n{your_english_text}\n### Chinese\n"
inputs = tokenizer(prompt, return_tensors="pt", add_special_tokens=False).to("cuda")
outputs = model.generate(**inputs, max_new_tokens=512, do_sample=False)
result = tokenizer.decode(outputs[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True)
print(result)
Identity and Version
- Repository
- tatasauce4life/qwen3-4b-en2zh-orpo
- Publisher
- Tatasauce
- Task
- Translation
- Modality
- Text
- Library
- Not stated by the source
- Parameters
- Not stated by the source
- Languages
- en, zh
- Revision
- c0e12dd4c5a1423bd0409284986c9668692759b4
- First published
- 2026-09-19
- Last updated
- 2026-09-19
Files and Weights
8 files, 275.8 MB in total. The weights are 1 file totalling 264.3 MB in safetensors.
Weights1 file · 264.3 MB
Configuration2 files · 2.2 KB
Tokenizer2 files · 11.4 MB
Documentation1 file · 2.4 KB
Other1 file · 9.5 KB
Repository1 file · 1.6 KB
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| adapter_model.safetensors | Weights | 264.3 MB | 42acd7bf8f64 |
| adapter_config.json | Configuration | 1.3 KB | — |
| training_config.json | Configuration | 930 B | — |
| README.md | Documentation | 2.4 KB | — |
| generation_samples.jsonl | Other | 9.5 KB | — |
| .gitattributes | Repository | 1.6 KB | — |
| tokenizer.json | Tokenizer | 11.4 MB | 476870a1f2fb |
| tokenizer_config.json | Tokenizer | 5.1 KB | — |
License and Download
- License
- apache-2.0
- Access
- Open weights, no gate
- Download size
- 264.3 MB
Download from Tatasauce
Released by Tatasauce through its official repository on Hugging Face. Read the license.
Built From
- Adapter of unsloth/Qwen3-4B-Base-unsloth-bnb-4bit
- Derived from unsloth/Qwen3-4B-Base-unsloth-bnb-4bit
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 264.3 MB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.
Questions About qwen3-4b-en2zh-orpo
Can I use qwen3-4b-en2zh-orpo commercially?
Yes. qwen3-4b-en2zh-orpo 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.
Similar Models
This is the model card of NLLB-200's distilled 600M variant. Here are the metrics for that particular checkpoint. - Information about training algorithms, parameters, fairness constraints or other applied approaches, and features. The exact training algorithm, data and the strategies to handle data imbalances for high and low resource languages that were used to train NLLB-200 is described in the paper. - Paper or other resource for more information NLLB Team et al, No Language Left Behind: Scaling Human-Centered Machine Translation, Arxiv, 2022 - Where to send questions or comments about the model: https://github.com/facebookresearch/fairseq/issues • Model performance measures: NLLB-200…
source languages: en; target languages: ru; OPUS readme: en-ru; dataset: opus; model: transformer-align; pre-processing: normalization + SentencePiece.
This model can be used for translation and text-to-text generation. CONTENT WARNING: Readers should be aware this section contains content that is disturbing, offensive, and can propagate historical and current stereotypes. Significant research has explored bias and fairness issues with language models (see, e.g., Sheng et al. (2021) and Bender et al. (2021)). Further details about the dataset for this model can be found in the OPUS readme: en-de
hfname: kor-eng - sourcelanguages: kor - targetlanguages: eng - opusreadmeurl: https://github.com/Helsinki-NLP/Tatoeba-Challenge/tree/master/models/kor-eng/README.md - originalrepo: Tatoeba-Challenge - srcconstituents: {'korHani', 'korHang', 'korLatn', 'kor'} - tgtconstituents: {'eng'} - srcmultilingual: False - tgtmultilingual: False - urlmodel: https://object.pouta.csc.fi/Tatoeba-MT-models/kor-eng/opus-2020-06-17.zip - urltestset: https://object.pouta.csc.fi/Tatoeba-MT-models/kor-eng/opus-2020-06-17.test.txt - srcalpha3: kor - tgtalpha3: eng - shortpair: ko-en - chrF2score: 0.588 - brevitypenalty: 0.9590000000000001 - reflen: 17711.0 - srcname: Korean - tgtname: English - traindate…
source languages: de; target languages: en; OPUS readme: de-en; dataset: opus; model: transformer-align; pre-processing: normalization + SentencePiece.