library_name: transformers tags: []

Mamba模型

本仓库包含与transfromers兼容的mamba-2.8b模型。检查点文件保持原始状态，完整的config.json配置和分词器已推送至该仓库。

使用指南

在transformers=4.39.0正式发布前，需通过以下命令安装main分支版本：

pip install git+https://github.com/huggingface/transformers@main

建议同时安装优化组件：

pip install causal-conv1d>=1.2.0
pip install mamba-ssm

若未安装上述组件，系统将自动回退至"eager"基础实现。安装后可启用优化的cuda内核。

文本生成

使用标准生成API：

>>> from transformers import MambaConfig, MambaForCausalLM, AutoTokenizer
>>> import torch

>>> tokenizer = AutoTokenizer.from_pretrained("state-spaces/mamba-1.4b-hf")
>>> model = MambaForCausalLM.from_pretrained("state-spaces/mamba-1.4b-hf")
>>> input_ids = tokenizer("最近过得怎么样？", return_tensors="pt")["input_ids"]

>>> out = model.generate(input_ids, max_new_tokens=10)
>>> print(tokenizer.batch_decode(out))
["最近过得怎么样？\n\n我挺好的。\n\n我"]

PEFT微调示例

使用peft库微调时，建议保持模型为float32精度：

from datasets import load_dataset
from trl import SFTTrainer
from peft import LoraConfig
from transformers import AutoTokenizer, AutoModelForCausalLM, TrainingArguments
tokenizer = AutoTokenizer.from_pretrained("state-spaces/mamba-1.4b-hf")
model = AutoModelForCausalLM.from_pretrained("state-spaces/mamba-1.4b-hf")
dataset = load_dataset("Abirate/english_quotes", split="train")
training_args = TrainingArguments(
    output_dir="./results",
    num_train_epochs=3,
    per_device_train_batch_size=4,
    logging_dir='./logs',
    logging_steps=10,
    learning_rate=2e-3
)
lora_config =  LoraConfig(
        r=8,
        target_modules=["x_proj", "embeddings", "in_proj", "out_proj"],
        task_type="CAUSAL_LM",
        bias="none"
)
trainer = SFTTrainer(
    model=model,
    tokenizer=tokenizer,
    args=training_args,
    peft_config=lora_config,
    train_dataset=dataset,
    dataset_text_field="quote",
)
trainer.train()