NanoTranslator-M2

English | 简体中文

Introduction

这是 NanoTranslator 的 Medium-2 型号，目前仅支持英译中。仓库中同时提供了 ONNX 版本的模型。

	P.	Arch.	Act.	V.	H.	I.	L.	A.H.	K.H.	Tie
XXL2	102	LLaMA	SwiGLU	16K	1120	3072	6	16	8	True
XXL	100	LLaMA	SwiGLU	16K	768	4096	8	24	8	True
XL	78	LLaMA	GeGLU	16K	768	4096	6	24	8	True
L	49	LLaMA	GeGLU	16K	512	2816	8	16	8	True
M2	22	Qwen2	GeGLU	4K	432	2304	6	24	8	True
M	22	LLaMA	SwiGLU	8K	256	1408	16	16	4	True
S	9	LLaMA	SwiGLU	4K	168	896	16	12	4	True
XS	2	LLaMA	SwiGLU	2K	96	512	12	12	4	True

P. - Parameters (in million)
V. - vocab size
H. - hidden size
I. - intermediate size
L. - num layers
A.H. - num attention heads
K.H. - num kv heads
Tie - tie word embeddings

How to use

Prompt 格式如下：

<|im_start|> {English Text} <|endoftext|>

Directly using transformers

import torch
from transformers import AutoTokenizer, AutoModelForCausalLM

model_path = 'Mxode/NanoTranslator-M2'

tokenizer = AutoTokenizer.from_pretrained(model_path)
model = AutoModelForCausalLM.from_pretrained(model_path)

def translate(text: str, model, **kwargs):
    generation_args = dict(
        max_new_tokens = kwargs.pop("max_new_tokens", 512),
        do_sample = kwargs.pop("do_sample", True),
        temperature = kwargs.pop("temperature", 0.55),
        top_p = kwargs.pop("top_p", 0.8),
        top_k = kwargs.pop("top_k", 40),
        eos_token_id = kwargs.pop("eos_token_id", tokenizer.eos_token_id),
        **kwargs
    )

    prompt = "<|im_start|>" + text + "<|endoftext|>"
    model_inputs = tokenizer([prompt], return_tensors="pt").to(model.device)

    generated_ids = model.generate(model_inputs.input_ids, **generation_args)
    generated_ids = [
        output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
    ]

    response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
    return response

text = "I love to watch my favorite TV series."

response = translate(text, model, max_new_tokens=64, do_sample=False)
print(response)

ONNX

根据实际测试，使用 ONNX 模型推理会比直接使用 transformers 推理要快 2～10 倍。

如果希望使用 ONNX 模型，那么你需要手动切换到 onnx 分支并从本地加载。

参考文档：

Using ORTModelForCausalLM

from optimum.onnxruntime import ORTModelForCausalLM
from transformers import AutoTokenizer

model_path = "your/folder/to/onnx_model"

ort_model = ORTModelForCausalLM.from_pretrained(model_path)
tokenizer = AutoTokenizer.from_pretrained(model_path)

text = "I love to watch my favorite TV series."

response = translate(text, ort_model, max_new_tokens=64, do_sample=False, eos_token_id=tokenizer.eos_token_id)
print(response)

Using pipeline

from optimum.pipelines import pipeline

model_path = "your/folder/to/onnx_model"
pipe = pipeline("text-generation", model=model_path, accelerator="ort")

text = "I love to watch my favorite TV series."

response = pipe(text, max_new_tokens=64, do_sample=False, eos_token_id=2)
response