Skip to content

JunnYu/RoFormer_pytorch

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

PyTorch RoFormer & RoFormer-V2

RoFormer模型和RoFormer-V2模型

更新

  • 2022/05/18

添加paddle版本RoFormerV2在分类任务上的训练结果。

  • 2022/05/11

感谢苏神提醒,添加了一个注释,其中RoFormerV2*表示未经多任务学习的RoFormerV2模型。

  • 2022/05/01

添加clue分类任务的代码和dev集结果,代码在examples/clue文件夹,缺少啥依赖安装啥,比如需要这个pip install -U accelerate

  • 2022/04/30

有个细节需要注意一下,苏神在微调时无论输入是text还是text pair类型时,token_type_id都置为了0。

如果想要使用与苏神保持一致,那么可以在tokenizer时候设置return_token_type_ids=False,这样模型会在内部处理。

否则对于text pair类型时,会返回与0,1两种类型的token_type_id

  • 2022/04/02

(1)修改RoFormerForCausalLM,支持roformer-sim并提供相关的例子,请见examples/test_sim.py

(2)修改apply_rotary实现方式,看起来更简单。

def apply_rotary(x, sinusoidal_pos=None):
    if sinusoidal_pos is None:
        return x
    sin, cos = sinusoidal_pos
    # x.shape [batch, seq_len, 2]
    x1, x2 = x[..., 0::2], x[..., 1::2]
    # [cos_nθ, -sin_nθ] [x1]
    # [sin_nθ,  cos_nθ] [x2]
    # => [x1 * cos_nθ - x2 * sin_nθ, x1 * sin_nθ + x2 * cos_nθ]
    # 苏神的rotary,使用了下面的计算方法。
    # return torch.stack([x1 * cos - x2 * sin, x1 * sin + x2 * cos], dim=-1).flatten(-2, -1)
    # 考虑到矩阵乘法torch.einsum("bhmd,bhnd->bhmn", q, k),因此可以直接在最后一个维度拼接(无需奇偶交错)
    return torch.cat([x1 * cos - x2 * sin, x1 * sin + x2 * cos], dim=-1)
  • 2022/03/21 添加roformer-v2的权重, 注:必须使用本仓库的代码,不能使用transformers仓库的代码!!!

安装

# v2版本
pip install roformer>=0.4.3
# v1版本(代码已经加入到huggingface仓库,请使用新版本的transformers)
pip install -U transformers

评测对比

CLUE-dev榜单分类任务结果,base+large版本。

iflytek tnews afqmc cmnli ocnli wsc csl avg
BERT 60.06 56.80 72.41 79.56 73.93 78.62 83.93 72.19
RoBERTa 60.64 58.06 74.05 81.24 76.00 87.50 84.50 74.57
RoFormer 60.91 57.54 73.52 80.92 76.07 86.84 84.63 74.35
RoFormerV2* 60.87 56.54 72.75 80.34 75.36 80.92 84.67 73.06
GAU-α 61.41 57.76 74.17 81.82 75.86 79.93 85.67 73.8
RoFormer-pytorch(本仓库代码) 60.60 57.51 74.44 80.79 75.67 86.84 84.77 74.37
RoFormerV2-pytorch(本仓库代码) 62.87 59.03 76.20 80.85 79.73 87.82 91.87 76.91
GAU-α-pytorch(Adafactor) 61.18 57.52 73.42 80.91 75.69 80.59 85.5 73.54
GAU-α-pytorch(AdamW wd0.01 warmup0.1) 60.68 57.95 73.08 81.02 75.36 81.25 83.93 73.32
RoFormerV2-large-pytorch(本仓库代码) 61.75 59.21 76.14 82.35 81.73 91.45 91.5 77.73
Chinesebert-large-pytorch 61.25 58.67 74.70 82.65 79.63 87.83 84.97 75.67
RoFormerV2-base-paddle 63.76 59.53 77.06 81.58 81.56 87.83 86.73 76.87
RoFormerV2-large-paddle 64.02 60.08 77.92 82.87 83.9 92.43 86.87 78.30

CLUE-1.0-test榜单分类任务结果,base+large版本。

iflytek tnews afqmc cmnli ocnli wsc csl avg
RoFormer-pytorch(本仓库代码) 59.54 57.34 74.46 80.23 73.67 80.69 84.57 72.93
RoFormerV2-pytorch(本仓库代码) 63.15 58.24 75.42 80.59 74.17 83.79 83.73 74.16
GAU-α-pytorch(Adafactor) 61.38 57.08 74.05 80.37 73.53 74.83 85.6 72.41
GAU-α-pytorch(AdamW wd0.01 warmup0.1) 60.54 57.67 72.44 80.32 72.97 76.55 84.13 72.09
RoFormerV2-large-pytorch(本仓库代码) 61.85 59.13 76.38 80.97 76.23 85.86 84.33 74.96
Chinesebert-large-pytorch 61.54 58.57 74.8 81.94 76.93 79.66 85.1 74.08
RoFormerV2-large-paddle 64.23 59.99 76.85 81.97 76.57 84.48 83.37 75.35

注:

  • 其中RoFormerV2*表示的是未进行多任务学习的RoFormerV2模型,该模型苏神并未开源,感谢苏神的提醒。
  • 其中不带有pytorch后缀结果都是从GAU-alpha仓库复制过来的。
  • 其中带有pytorch后缀的结果都是自己训练得出的。
  • 苏神代码中拿了cls标签后直接进行了分类,而本仓库使用了如下的分类头,多了2个dropout,1个dense,1个relu激活。
  • paddle版本的代码进行了grid search!
class RoFormerClassificationHead(nn.Module):
    def __init__(self, config):
        super().__init__()
        self.dense = nn.Linear(config.hidden_size, config.hidden_size)
        self.dropout = nn.Dropout(config.hidden_dropout_prob)
        self.out_proj = nn.Linear(config.hidden_size, config.num_labels)

        self.config = config

    def forward(self, features, **kwargs):
        x = features[:, 0, :]  # take <s> token (equiv. to [CLS])
        x = self.dropout(x)
        x = self.dense(x)
        x = ACT2FN[self.config.hidden_act](x) # 这里是relu
        x = self.dropout(x)
        x = self.out_proj(x)
        return x

Tips:

  • 实验环境RTX 3090

Leadborad截图

Roformer-sim测试例子

import torch
import numpy as np
from roformer import RoFormerForCausalLM, RoFormerConfig
from transformers import BertTokenizer

device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')
# 可选以下几个。
# junnyu/roformer_chinese_sim_char_small, junnyu/roformer_chinese_sim_char_base
# junnyu/roformer_chinese_sim_char_ft_small, roformer_chinese_sim_char_ft_base
pretrained_model = "junnyu/roformer_chinese_sim_char_base"
tokenizer = BertTokenizer.from_pretrained(pretrained_model)
config = RoFormerConfig.from_pretrained(pretrained_model)
config.is_decoder = True
config.eos_token_id = tokenizer.sep_token_id
config.pooler_activation = "linear"
model = RoFormerForCausalLM.from_pretrained(pretrained_model, config=config)
model.to(device)
model.eval()

def gen_synonyms(text, n=100, k=20):
    ''''含义: 产生sent的n个相似句,然后返回最相似的k个。
    做法:用seq2seq生成,并用encoder算相似度并排序。
    '''
    # 寻找所有相似的句子
    r = []
    inputs1 = tokenizer(text, return_tensors="pt")
    for _ in range(n):
        inputs1.to(device)
        output = tokenizer.batch_decode(model.generate(**inputs1, top_p=0.95, do_sample=True, max_length=128), skip_special_tokens=True)[0].replace(" ","").replace(text, "") # 去除空格,去除原始text文本。
        r.append(output)
    
    # 对相似的句子进行排序
    r = [i for i in set(r) if i != text and len(i) > 0]
    r = [text] + r
    inputs2 = tokenizer(r, padding=True, return_tensors="pt")
    with torch.no_grad():
        inputs2.to(device)
        outputs = model(**inputs2)
        Z = outputs.pooler_output.cpu().numpy()
    Z /= (Z**2).sum(axis=1, keepdims=True)**0.5
    argsort = np.dot(Z[1:], -Z[0]).argsort()
    
    return [r[i + 1] for i in argsort[:k]]

out = gen_synonyms("广州和深圳哪个好?")
print(out)
# ['深圳和广州哪个好?',
#  '广州和深圳哪个好',
#  '深圳和广州哪个好',
#  '深圳和广州哪个比较好。',
#  '深圳和广州哪个最好?',
#  '深圳和广州哪个比较好',
#  '广州和深圳那个比较好',
#  '深圳和广州哪个更好?',
#  '深圳与广州哪个好',
#  '深圳和广州,哪个比较好',
#  '广州与深圳比较哪个好',
#  '深圳和广州哪里比较好',
#  '深圳还是广州比较好?',
#  '广州和深圳哪个地方好一些?',
#  '广州好还是深圳好?',
#  '广州好还是深圳好呢?',
#  '广州与深圳哪个地方好点?',
#  '深圳好还是广州好',
#  '广州好还是深圳好',
#  '广州和深圳哪个城市好?']

模型权重对照表

中文模型 roformer-v2

huggingface.co bert4keras
roformer_v2_chinese_char_small chinese_roformer-v2-char_L-6_H-384_A-6.zip (download code:ttn4)
roformer_v2_chinese_char_base chinese_roformer-v2-char_L-12_H-768_A-12.zip (download code:pfoh)
roformer_v2_chinese_char_large chinese_roformer-v2-char_L-24_H-1024_A-16.zip (download code:npfv)

中文模型 roformer-v1

huggingface.co bert4keras
roformer_chinese_base chinese_roformer_L-12_H-768_A-12.zip (download code:xy9x)
roformer_chinese_small chinese_roformer_L-6_H-384_A-6.zip (download code:gy97)
roformer_chinese_char_base chinese_roformer-char_L-12_H-768_A-12.zip (download code:bt94)
roformer_chinese_char_small chinese_roformer-char_L-6_H-384_A-6.zip (download code:a44c)
roformer_chinese_sim_char_base chinese_roformer-sim-char_L-12_H-768_A-12.zip (download code:2cgz)
roformer_chinese_sim_char_small chinese_roformer-sim-char_L-6_H-384_A-6.zip (download code:h68q)
roformer_chinese_sim_char_ft_base chinese_roformer-sim-char-ft_L-12_H-768_A-12.zip (download code:w15n)
roformer_chinese_sim_char_ft_small chinese_roformer-sim-char-ft_L-6_H-384_A-6.zip (download code:gty5)

英文模型(使用electra的训练方法在openwebtext上训练的small模型(rotary value = True))

huggingface.co
roformer_small_generator
roformer_small_discriminator

Roformer-v2 MLM测试

import torch
import tensorflow as tf
from transformers import BertTokenizer
from roformer import RoFormerForMaskedLM, TFRoFormerForMaskedLM

text = "今天[MASK]很好,我[MASK]去公园玩。"
tokenizer = BertTokenizer.from_pretrained("junnyu/roformer_v2_chinese_char_base")
pt_model = RoFormerForMaskedLM.from_pretrained("junnyu/roformer_v2_chinese_char_base")
tf_model = TFRoFormerForMaskedLM.from_pretrained(
    "junnyu/roformer_v2_chinese_char_base", from_pt=True
)
pt_inputs = tokenizer(text, return_tensors="pt")
tf_inputs = tokenizer(text, return_tensors="tf")
# pytorch
with torch.no_grad():
    pt_outputs = pt_model(**pt_inputs).logits[0]
pt_outputs_sentence = "pytorch: "
for i, id in enumerate(tokenizer.encode(text)):
    if id == tokenizer.mask_token_id:
        tokens = tokenizer.convert_ids_to_tokens(pt_outputs[i].topk(k=5)[1])
        pt_outputs_sentence += "[" + "||".join(tokens) + "]"
    else:
        pt_outputs_sentence += "".join(
            tokenizer.convert_ids_to_tokens([id], skip_special_tokens=True)
        )
print(pt_outputs_sentence)
# tf
tf_outputs = tf_model(**tf_inputs, training=False).logits[0]
tf_outputs_sentence = "tf: "
for i, id in enumerate(tokenizer.encode(text)):
    if id == tokenizer.mask_token_id:
        tokens = tokenizer.convert_ids_to_tokens(tf.math.top_k(tf_outputs[i], k=5)[1])
        tf_outputs_sentence += "[" + "||".join(tokens) + "]"
    else:
        tf_outputs_sentence += "".join(
            tokenizer.convert_ids_to_tokens([id], skip_special_tokens=True)
        )
print(tf_outputs_sentence)
# small
# pytorch: 今天[的||,||是||很||也]很好,我[要||会||是||想||在]去公园玩。
# tf: 今天[的||,||是||很||也]很好,我[要||会||是||想||在]去公园玩。
# base
# pytorch: 今天[我||天||晴||园||玩]很好,我[想||要||会||就||带]去公园玩。
# tf: 今天[我||天||晴||园||玩]很好,我[想||要||会||就||带]去公园玩。
# large
# pytorch: 今天[天||气||我||空||阳]很好,我[又||想||会||就||爱]去公园玩。
# tf: 今天[天||气||我||空||阳]很好,我[又||想||会||就||爱]去公园玩。

Roformer-v1 MLM测试

import torch
import tensorflow as tf
from transformers import RoFormerForMaskedLM, RoFormerTokenizer, TFRoFormerForMaskedLM

text = "今天[MASK]很好,我[MASK]去公园玩。"
tokenizer = RoFormerTokenizer.from_pretrained("junnyu/roformer_chinese_base")
pt_model = RoFormerForMaskedLM.from_pretrained("junnyu/roformer_chinese_base")
tf_model = TFRoFormerForMaskedLM.from_pretrained(
    "junnyu/roformer_chinese_base", from_pt=True
)
pt_inputs = tokenizer(text, return_tensors="pt")
tf_inputs = tokenizer(text, return_tensors="tf")
# pytorch
with torch.no_grad():
    pt_outputs = pt_model(**pt_inputs).logits[0]
pt_outputs_sentence = "pytorch: "
for i, id in enumerate(tokenizer.encode(text)):
    if id == tokenizer.mask_token_id:
        tokens = tokenizer.convert_ids_to_tokens(pt_outputs[i].topk(k=5)[1])
        pt_outputs_sentence += "[" + "||".join(tokens) + "]"
    else:
        pt_outputs_sentence += "".join(
            tokenizer.convert_ids_to_tokens([id], skip_special_tokens=True)
        )
print(pt_outputs_sentence)
# tf
tf_outputs = tf_model(**tf_inputs, training=False).logits[0]
tf_outputs_sentence = "tf: "
for i, id in enumerate(tokenizer.encode(text)):
    if id == tokenizer.mask_token_id:
        tokens = tokenizer.convert_ids_to_tokens(tf.math.top_k(tf_outputs[i], k=5)[1])
        tf_outputs_sentence += "[" + "||".join(tokens) + "]"
    else:
        tf_outputs_sentence += "".join(
            tokenizer.convert_ids_to_tokens([id], skip_special_tokens=True)
        )
print(tf_outputs_sentence)
# pytorch: 今天[天气||天||心情||阳光||空气]很好,我[想||要||打算||准备||喜欢]去公园玩。
# tf:      今天[天气||天||心情||阳光||空气]很好,我[想||要||打算||准备||喜欢]去公园玩。

手动权重转换

python convert_roformer_original_tf_checkpoint_to_pytorch.py \
    --tf_checkpoint_path=xxxxxx/chinese_roformer_L-12_H-768_A-12/bert_model.ckpt \
    --bert_config_file=pretrained_models/chinese_roformer_base/config.json \
    --pytorch_dump_path=pretrained_models/chinese_roformer_base/pytorch_model.bin

tf与pytorch精度对齐

small版本
bert4keras vs pytorch
mean diff : tensor(5.9108e-07)
max diff : tensor(5.7220e-06)
bert4keras vs tf2.0
mean diff : tensor(4.5976e-07)
max diff : tensor(3.5763e-06)

base版本
python compare_model.py
bert4keras vs pytorch
mean diff : tensor(4.3340e-07)
max diff : tensor(5.7220e-06)
bert4keras vs tf2.0
mean diff : tensor(3.4319e-07)
max diff : tensor(5.2452e-06)

参考

https://github.com/pengming617/bert_classification

https://github.com/bojone/bert4keras

https://github.com/ZhuiyiTechnology/roformer

https://github.com/lonePatient/NeZha_Chinese_PyTorch

https://github.com/lonePatient/TorchBlocks

https://github.com/huggingface/transformers

Citation

Bibtex:

@misc{su2021roformer,
      title={RoFormer: Enhanced Transformer with Rotary Position Embedding}, 
      author={Jianlin Su and Yu Lu and Shengfeng Pan and Bo Wen and Yunfeng Liu},
      year={2021},
      eprint={2104.09864},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}
@techreport{roformerv2,
  title={RoFormerV2: A Faster and Better RoFormer - ZhuiyiAI},
  author={Jianlin Su, Shengfeng Pan, Bo Wen, Yunfeng Liu},
  year={2022},
  url="https://github.com/ZhuiyiTechnology/roformer-v2",
}