---
license: apache-2.0
base_model: Qwen/Qwen3-1.7B-Base
language:
- zh
- en
pipeline_tag: translation
library_name: transformers
tags:
- translation
- qwen3
- zh-en
- en-zh
- sft
- cpo
- grpo
datasets:
- haoranxu/ALMA-Human-Parallel
- haoranxu/X-ALMA-Parallel-Data
---

# Interpreter-Qwen3-1.7B

A 1.7B Chinese↔English translation model, trained **SFT → CPO → GRPO**.

This model is one half of an A/B pair. The two differ in **exactly two things** — the base checkpoint and the tokenizer — and share the same data, hyperparameters, losses and rewards. The other half is [`Ismantic/Interpreter-Qwen3-1.7B-ReTok`](https://huggingface.co/Ismantic/Interpreter-Qwen3-1.7B-ReTok).

| | this model |
|---|---|
| Tokenizer | Qwen3 native BBPE (151643) |
| Chat format | ChatML |
| Pipeline | SFT → CPO → GRPO |

## Results (WMT23)

| direction | BLEU / COMET |
|---|---|
| zh→en | 20.31 / 0.8053 |
| en→zh | 33.69 / 0.8540 |

### How to read these numbers

WMT23 is the primary test set — **WMT22 is excluded**, roughly 17.5% of it leaked
into the ALMA SFT data. COMET is `Unbabel/wmt22-comet-da`, which is *also* the
GRPO reward, so gains were cross-checked on WMT23/24 + Flores-200 with BLEU and
chrF to rule out reward hacking.

vLLM greedy decoding is not bit-reproducible; BLEU varies by ~0.1 between runs.
Differences at that scale are noise.

## Usage

Standard ChatML with a fixed translation instruction; `<|im_end|>` is the stop
token. Greedy decoding recommended.

```python
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

tok = AutoTokenizer.from_pretrained("Ismantic/Interpreter-Qwen3-1.7B")
model = AutoModelForCausalLM.from_pretrained("Ismantic/Interpreter-Qwen3-1.7B", torch_dtype=torch.bfloat16).cuda().eval()

instr = ("Translate the following text from Chinese to English.\n"
         "Chinese: 人工智能正在深刻改变我们的生活方式。\nEnglish:")
prompt = f"<|im_start|>user\n{instr}<|im_end|>\n<|im_start|>assistant\n"
ids = tok(prompt, return_tensors="pt").to(model.device)
out = model.generate(**ids, max_new_tokens=256, do_sample=False,
                     eos_token_id=tok.convert_tokens_to_ids("<|im_end|>"))
print(tok.decode(out[0][ids.input_ids.shape[1]:], skip_special_tokens=True).strip())
```

vLLM works too — same ChatML prompt, `stop=["<|im_end|>"]`.

For a quick interactive demo, grab `translate.py` from this repo and run
`python translate.py` (needs `vllm`) — type sentences, zh↔en auto-detected.

## Training

Three stages, all on a single RTX 4090:

- **SFT** — full fine-tune on ~36.8K zh↔en pairs (ALMA + X-ALMA human parallel,
  WMT22/23 leakage removed), loss on the response only.
- **CPO** — LoRA preference training, `-log σ(β·(logπ_w − logπ_l)) + λ·NLL(y_w)`,
  on ~44K **self-generated** preference pairs (5 candidates sampled from the SFT
  model, best/worst picked by COMET). LoRA only — full-parameter CPO collapses
  the model.
- **GRPO** — full-parameter RL. Reward = reference-based `wmt22-comet-da` COMET
  (1.0) + a 4-gram repetition penalty (0.3), over WMT17–21 source prompts.

Training code, data provenance and the negative results (what was tried and
rejected, with numbers) are in the
[Interpreter repo](https://github.com/Ismantic/Interpreter).

## License & attribution

Apache-2.0, following the base model. Training data derives from the ALMA /
X-ALMA parallel corpora and WMT news test sets. Please respect the licenses of
those upstream models and datasets.