Esmplusplus Small

🚀 ESM++

ESM++是ESMC（许可证）的忠实实现，它支持批量处理，并且无需ESM Python包即可与标准的Huggingface兼容。小版本对应于ESMC的3亿参数版本。该项目解决了在不依赖ESM Python包的情况下，实现与Huggingface的标准兼容以及批量处理的问题，为相关研究和开发提供了便利。

🚀 快速开始

之前Huggingface权重绑定存在一个bug，导致ESM++的对数几率与ESMC不同。该bug现已修复。

✨ 主要特性

• 忠实实现ESMC，支持批量处理。
• 无需ESM Python包，与标准Huggingface兼容。
• 支持序列和标记级别的分类任务。
• 提供不同浮点精度的加载选项。
• 可返回注意力图。
• 相比ESMC和其他模型，推理速度更快。

📦 安装指南

文档未提及安装步骤，暂不展示。

💻 使用示例

基础用法

from transformers import AutoModelForMaskedLM
model = AutoModelForMaskedLM.from_pretrained('Synthyra/ESMplusplus_small', trust_remote_code=True)
tokenizer = model.tokenizer

sequences = ['MPRTEIN', 'MSEQWENCE']
tokenized = tokenizer(sequences, padding=True, return_tensors='pt')

# tokenized['labels'] = tokenized['input_ids'].clone() # correctly mask input_ids and set unmasked instances of labels to -100 for MLM training

output = model(**tokenized) # get all hidden states with output_hidden_states=True
print(output.logits.shape) # language modeling logits, (batch_size, seq_len, vocab_size), (2, 11, 64)
print(output.last_hidden_state.shape) # last hidden state of the model, (batch_size, seq_len, hidden_size), (2, 11, 960)
print(output.loss) # language modeling loss if you passed labels
#print(output.hidden_states) # all hidden states if you passed output_hidden_states=True (in tuple)

高级用法

支持序列和标记级别的分类任务

from transformers import AutoModelForSequenceClassification, AutoModelForTokenClassification

model = AutoModelForSequenceClassification.from_pretrained('Synthyra/ESMplusplus_small', num_labels=2, trust_remote_code=True)
logits = model(**tokenized).logits
print(logits.shape) # (batch_size, num_labels), (2, 2)

以不同浮点精度加载权重

import torch
model = AutoModelForMaskedLM.from_pretrained('Synthyra/ESMplusplus_small', trust_remote_code=True, torch_dtype=torch.float16) # or torch.bfloat16

嵌入整个数据集

embedding_dict = model.embed_dataset(
    sequences=[
        'MALWMRLLPLLALLALWGPDPAAA', ... # list of protein sequences
    ],
    tokenizer=model.tokenizer,
    batch_size=2, # adjust for your GPU memory
    max_len=512, # adjust for your needs
    full_embeddings=False, # if True, no pooling is performed
    embed_dtype=torch.float32, # cast to what dtype you want
    pooling_types=['mean', 'cls'], # more than one pooling type will be concatenated together
    num_workers=0, # if you have many cpu cores, we find that num_workers = 4 is fast for large datasets
    sql=False, # if True, embeddings will be stored in SQLite database
    sql_db_path='embeddings.db',
    save=True, # if True, embeddings will be saved as a .pth file
    save_path='embeddings.pth',
)
# embedding_dict is a dictionary mapping sequences to their embeddings as tensors for .pth or numpy arrays for sql

model.embed_dataset()
Args:
    sequences: List of protein sequences
    batch_size: Batch size for processing
    max_len: Maximum sequence length
    full_embeddings: Whether to return full residue-wise (True) embeddings or pooled (False)
    pooling_type: Type of pooling ('mean' or 'cls')
    num_workers: Number of workers for data loading, 0 for the main process
    sql: Whether to store embeddings in SQLite database - will be stored in float32
    sql_db_path: Path to SQLite database
    
Returns:
    Dictionary mapping sequences to embeddings, or None if sql=True

Note:
    - If sql=True, embeddings can only be stored in float32
    - sql is ideal if you need to stream a very large dataset for training in real-time
    - save=True is ideal if you can store the entire embedding dictionary in RAM
    - sql will be used if it is True and save is True or False
    - If your sql database or .pth file is already present, they will be scanned first for already embedded sequences
    - Sequences will be truncated to max_len and sorted by length in descending order for faster processing

使用🤗 peft进行微调

model = AutoModelForSequenceClassification.from_pretrained('Synthyra/ESMplusplus_small', num_labels=2, trust_remote_code=True)
# these modules handle ESM++ and ESM2 attention layers
target_modules = ["layernorm_qkv.1", "out_proj", "query", "key", "value", "dense"]

lora_config = LoraConfig(
    r=8, # choose lora parameters to your liking
    lora_alpha=16,
    lora_dropout=0.01,
    bias="none",
    target_modules=target_modules,
)

# Apply LoRA to the model
model = get_peft_model(model, lora_config)

# Unfreeze the classifier head
for param in model.classifier.parameters():
    param.requires_grad = True

如需更全面的微调示例，请查看我们的示例脚本此处。

返回注意力图

output = model(**tokenized, output_attentions=True)
att = output.attentions
len(att) # 30, one for each layer, size (batch_size, num_heads, seq_len, seq_len) each

📚 详细文档

不同浮点精度和实现方式的比较

我们测量了fp32权重与fp16或bf16的最后隐藏状态的差异。发现fp16更接近fp32的输出，因此建议以fp16加载。请注意，ESM包也以fp32加载ESMC，但默认转换为bf16，这在推理/训练中各有优缺点，因此可以根据需要选择半精度加载方式。

• FP32与FP16的平均均方误差（MSE）：0.00000003
• FP32与BF16的平均均方误差（MSE）：0.00000140

我们还测量了ESM++与ESMC（均为bfloat16）在1000个随机序列上的输出差异，以确保与ESM包兼容。

• 最后隐藏状态的平均均方误差（MSE）：7.74e - 10

你可以通过将.from_pretrained(...)替换为.from_pretrained_esm('esmc_300m')，从ESM包而不是transformers加载权重。

模型探针

我们在各种PLM和标准数据集上采用线性探测技术，类似于我们之前的论文，以评估池化隐藏状态与有价值属性之间的内在相关性。ESMC（以及ESM++）表现出色。

下图展示了在负控制（随机向量嵌入）和最佳表现者之间进行归一化后的性能。分类任务得分在MCC和F1（或多标签的F1max）之间平均，回归任务在Spearman rho和R2之间平均。

推理速度

我们研究了各种ESM模型在H100上的吞吐量。在ESMC和ESM++之间添加高效的批量处理显著提高了吞吐量，即使在批量大小为1的情况下，ESM++也比ESMC更快。ESM++小版本在处理长序列时甚至比ESM2 - 35M更快！在Linux机器上使用PyTorch > 2.5时，提升最为明显。

🔧 技术细节

文档未提及技术实现细节，暂不展示。

📄 许可证

如果你使用此实现或相关工作，请引用它（以及ESMC预印本）。

@misc {ESMPlusPlus,
	author       = { Hallee, L. and Bichara, D. and Gleghorn, J, P. },
	title        = { ESMPlusPlus },
	year         = 2024,
	url          = { https://huggingface.co/Synthyra/ESMplusplus_small },
	doi          = { 10.57967/hf/3725 },
	publisher    = { Hugging Face }
}