Evolutionary Scale Modeling (esm): Pretrained language models for proteins

Last update: Jan 09, 2023

Related tags

Overview

Evolutionary Scale Modeling

This repository contains code and pre-trained weights for Transformer protein language models from Facebook AI Research, including our state-of-the-art ESM-1b and MSA Transformer. Transformer protein language models were introduced in our paper, "Biological structure and function emerge from scaling unsupervised learning to 250 million protein sequences" (Rives et al., 2019).

ESM-1b outperforms all tested single-sequence protein language models across a range of structure prediction tasks. The MSA Transformer (ESM-MSA-1) can improve performance further by leveraging MSA information.

Citation

@article{rives2019biological,
  author={Rives, Alexander and Meier, Joshua and Sercu, Tom and Goyal, Siddharth and Lin, Zeming and Liu, Jason and Guo, Demi and Ott, Myle and Zitnick, C. Lawrence and Ma, Jerry and Fergus, Rob},
  title={Biological Structure and Function Emerge from Scaling Unsupervised Learning to 250 Million Protein Sequences},
  year={2019},
  doi={10.1101/622803},
  url={https://www.biorxiv.org/content/10.1101/622803v4},
  journal={bioRxiv}
}

Table of contents

Main models you should use
Comparison to related works
Usage
Available Models and Datasets
Citations
License

What's New

August 2021: Added flexibility to tokenizer to allow for spaces and special tokens (like <mask>) in sequence.
July 2021: New pre-trained model ESM-1v released, trained on UniRef90 (see Meier et al. 2021).
July 2021: New MSA Transformer released, with a minor fix in the row positional embeddings (ESM-MSA-1b).
Feb 2021: MSA Transformer added (see Rao et al. 2021). Example usage in notebook.
Dec 2020: Self-Attention Contacts for all pre-trained models (see Rao et al. 2020)
Dec 2020: Added new pre-trained model ESM-1b (see Rives et al. 2019 Appendix B)
Dec 2020: ESM Structural Split Dataset (see Rives et al. 2019 Appendix A.10)

Main models you should use

Shorthand	`esm.pretrained.`	Dataset	Description
ESM-1b	`esm1b_t33_650M_UR50S()`	UR50	SOTA general-purpose protein language model. Can be used to predict structure, function and other protein properties directly from individual sequences. Released with Rives et al. 2019 (Dec 2020 update).
ESM-MSA-1b	`esm_msa1b_t12_100M_UR50S()`	UR50 + MSA	MSA Transformer language model. Can be used to extract embeddings from an MSA. Enables SOTA inference of structure. Released with Rao et al. 2021 (ICML'21 version, June 2021).
ESM-1v	`esm1v_t33_650M_UR90S_1()` ... `esm1v_t33_650M_UR90S_5()`	UR90	Language model specialized for prediction of variant effects. Enables SOTA zero-shot prediction of the functional effects of sequence variations. Same architecture as ESM-1b, but trained on UniRef90. Released with Meier et al. 2021.

For a complete list of available models, with details and release notes, see Pre-trained Models.

Comparison to related works

Task	Unsupervised contact prediction			Supervised contact prediction		SSP
Test set	Large valid	CASP13-FM	CAMEO	CASP13-FM	CAMEO	CB513
Gremlin (Potts)	39.3	16.9	24.0	40.1	47.3
UniRep				11.2	17.8	58.4
SeqVec				13.8	22.5	62.1
TAPE	11.2	5.5	6.8	12.3	15.9	58.0
ProtBert-BFD	34.1	13.5	23.9	24.7	37.0	70.0
Prot-T5-XL-BFD	35.6	16.5	25.9	25.0	40.8	71.4 ± 0.3
ESM-1	33.7	13.6	21.4	(todo)	(todo)	69.2
ESM-1b	41.1	17.0	30.9	28.2	44.4	71.6 ± 0.1
ESM-1v	35.3	14.2	24.4
ESM-MSA-1b	57.4	44.8	43.5	54.6	55.8	73.4 ± 0.3

Comparison to related protein language models on structure prediction tasks.

All contact numbers are the top-L,LR precision metric, where long range means sequence separation of at least 24 residues
For unsupervised contact prediction, a sparse linear combination of the attention heads is used to directly predict protein contacts, fitted with logistic regression on 20 structures. For more details on the method, see Rao et al. 2020.
Supervised contact prediction all uses the same resnet (32 layers) and trRosetta training data, cf Rao et al. 2021.
(SSP) Secondary structure Q8 accuracy on CB513, transformer finetuned with convolution + LSTM head.
Direct coupling analysis methods (Gremlin, mfDCA, Psicov) and ESM-MSA-1 use the trRosetta MSAs, while other methods predict from single sequence.

Usage

Quick Start

As a prerequisite, you must have PyTorch 1.5 or later installed to use this repository.

You can use this one-liner for installation:

$ pip install fair-esm

We also support PyTorch Hub, which removes the need to clone and/or install this repository yourself:

import torch
model, alphabet = torch.hub.load("facebookresearch/esm:main", "esm1b_t33_650M_UR50S")

Then, you can load and use a pretrained model as follows:

import torch
import esm

# Load ESM-1b model
model, alphabet = esm.pretrained.esm1b_t33_650M_UR50S()
batch_converter = alphabet.get_batch_converter()

# Prepare data (first 2 sequences from ESMStructuralSplitDataset superfamily / 4)
data = [
    ("protein1", "MKTVRQERLKSIVRILERSKEPVSGAQLAEELSVSRQVIVQDIAYLRSLGYNIVATPRGYVLAGG"),
    ("protein2", "KALTARQQEVFDLIRDHISQTGMPPTRAEIAQRLGFRSPNAAEEHLKALARKGVIEIVSGASRGIRLLQEE"),
    ("protein2 with mask","KALTARQQEVFDLIRD<mask>ISQTGMPPTRAEIAQRLGFRSPNAAEEHLKALARKGVIEIVSGASRGIRLLQEE"),
    ("protein3",  "K A <mask> I S Q"),
]
batch_labels, batch_strs, batch_tokens = batch_converter(data)

# Extract per-residue representations (on CPU)
with torch.no_grad():
    results = model(batch_tokens, repr_layers=[33], return_contacts=True)
token_representations = results["representations"][33]

# Generate per-sequence representations via averaging
# NOTE: token 0 is always a beginning-of-sequence token, so the first residue is token 1.
sequence_representations = []
for i, (_, seq) in enumerate(data):
    sequence_representations.append(token_representations[i, 1 : len(seq) + 1].mean(0))

# Look at the unsupervised self-attention map contact predictions
import matplotlib.pyplot as plt
for (_, seq), attention_contacts in zip(data, results["contacts"]):
    plt.matshow(attention_contacts[: len(seq), : len(seq)])
    plt.title(seq)
    plt.show()

Compute embeddings in bulk from FASTA

We provide a script that efficiently extracts embeddings in bulk from a FASTA file. A cuda device is optional and will be auto-detected. The following command extracts the final-layer embedding for a FASTA file from the ESM-1b model:

$ python extract.py esm1b_t33_650M_UR50S examples/some_proteins.fasta examples/some_proteins_emb_esm1b/ \
    --repr_layers 0 32 33 --include mean per_tok

Directory examples/some_proteins_emb_esm1b/ now contains one .pt file per FASTA sequence; use torch.load() to load them. extract.py has flags that determine what's included in the .pt file:

--repr-layers (default: final only) selects which layers to include embeddings from.
--include specifies what embeddings to save. You can use the following:
- per_tok includes the full sequence, with an embedding per amino acid (seq_len x hidden_dim).
- mean includes the embeddings averaged over the full sequence, per layer.
- bos includes the embeddings from the beginning-of-sequence token. (NOTE: Don't use with the pre-trained models - we trained without bos-token supervision)

Zero-shot variant prediction

See "./variant-prediction/" for code and pre-trained weights for the ESM-1v models described in Language models enable zero-shot prediction of the effects of mutations on protein function. (Meier et al. 2021).

Notebooks

Supervised variant prediction - training a classifier on the embeddings

To help you get started with using the embeddings, this jupyter notebook tutorial shows how to train a supervised variant predictor using embeddings from ESM-1. You can adopt a similar protocol to train a model for any downstream task, even with limited data. First you can obtain the embeddings for examples/P62593.fasta either by downloading the precomputed embeddings as instructed in the notebook or by running the following:

# Obtain the embeddings
$ python extract.py esm1_t34_670M_UR50S examples/P62593.fasta examples/P62593_reprs/ \
    --repr_layers 34 --include mean

Then, follow the remaining instructions in the tutorial. You can also run the tutorial in a colab notebook.

Note this is somewhat outdated: use esm1v_t33_650M_UR90S instead, and see the newer instructions for zero-shot variant prediction, that is without any supervised training.

Unsupervised contact prediction

This jupyter notebook tutorial demonstrates contact prediction with both the ESM-1b and MSA Transformer (ESM-MSA-1) models. Contact prediction is based on a logistic regression over the model's attention maps. This methodology is based on our ICLR 2021 paper, Transformer protein language models are unsupervised structure learners. (Rao et al. 2020) The MSA Transformer (ESM-MSA-1) takes a multiple sequence alignment (MSA) as input, and uses the tied row self-attention maps in the same way. See MSA Transformer. (Rao et al. 2021).

To get unsupervised attention-based contacts, call model.predict_contacts(tokens) or model(tokens, return_contacts=True).

ESMStructuralSplitDataset and self-attention contact prediction

And this jupyter notebook tutorial shows how to load and index the ESMStructuralSplitDataset, and computes the self-attention map unsupervised contact predictions using ESM-1b.

Available Models and Datasets

Pre-trained Models

Shorthand	`esm.pretrained.`	#layers	#params	Dataset	Embedding Dim	Model URL (automatically downloaded to `~/.cache/torch/hub/checkpoints`)
ESM-1v	`esm1v_t33_650M_UR90S_[1-5]`	33	650M	UR90/S 2020_03	1280	https://dl.fbaipublicfiles.com/fair-esm/models/esm1v_t33_650M_UR90S_1.pt
ESM-MSA-1b	`esm_msa1b_t12_100M_UR50S`	12	100M	UR50/S + MSA 2018_03	768	https://dl.fbaipublicfiles.com/fair-esm/models/esm_msa1b_t12_100M_UR50S.pt
ESM-MSA-1	`esm_msa1_t12_100M_UR50S`	12	100M	UR50/S + MSA 2018_03	768	https://dl.fbaipublicfiles.com/fair-esm/models/esm_msa1_t12_100M_UR50S.pt
ESM-1b	`esm1b_t33_650M_UR50S`	33	650M	UR50/S 2018_03	1280	https://dl.fbaipublicfiles.com/fair-esm/models/esm1b_t33_650M_UR50S.pt
ESM-1	`esm1_t34_670M_UR50S`	34	670M	UR50/S 2018_03	1280	https://dl.fbaipublicfiles.com/fair-esm/models/esm1_t34_670M_UR50S.pt
	`esm1_t34_670M_UR50D`	34	670M	UR50/D 2018_03	1280	https://dl.fbaipublicfiles.com/fair-esm/models/esm1_t34_670M_UR50D.pt
	`esm1_t34_670M_UR100`	34	670M	UR100 2018_03	1280	https://dl.fbaipublicfiles.com/fair-esm/models/esm1_t34_670M_UR100.pt
	`esm1_t12_85M_UR50S`	12	85M	UR50/S 2018_03	768	https://dl.fbaipublicfiles.com/fair-esm/models/esm1_t12_85M_UR50S.pt
	`esm1_t6_43M_UR50S`	6	43M	UR50/S 2018_03	768	https://dl.fbaipublicfiles.com/fair-esm/models/esm1_t6_43M_UR50S.pt

Here is a chronological list of the released models and the paper they were introduced in:

Shorthand	Release Notes
ESM-1	Released with Rives et al. 2019 (Aug 2020 update).
ESM-1b	Released with Rives et al. 2019 (Dec 2020 update). See Appendix B.
ESM-MSA-1	Released with Rao et al. 2021 (Preprint v1).
ESM-MSA-1b	Released with Rao et al. 2021 (ICML'21 version, June 2021).
ESM-1v	Released with Meier et al. 2021.

ESM Structural Split Dataset

This is a five-fold cross validation dataset of protein domain structures that can be used to measure generalization of representations across different levels of structural dissimilarity. The dataset implements structural holdouts at the family, superfamily, and fold level. The SCOPe database is used to classify domains. Independently for each level of structural hold-out, the domains are split into 5 equal sets, i.e. five sets of folds, superfamilies, or families. This ensures that for each of the five partitions, structures having the same classification do not appear in both the train and test sets. For a given classification level each structure appears in a test set once, so that in the cross validation experiment each of the structures will be evaluated exactly once.

The dataset provides 3d coordinates, distance maps, and secondary structure labels. For further details on the construction of the dataset see Rives et al. 2019 Appendix A.10.

This jupyter notebook tutorial shows how to load and index the ESMStructuralSplitDataset.

ESMStructuralSplitDataset, upon initializing, will download splits and pkl. We also provide msas for each of the domains. The data can be directly downloaded below.

Name	Description	URL
splits	train/valid splits	https://dl.fbaipublicfiles.com/fair-esm/structural-data/splits.tar.gz
pkl	pkl objects containing sequence, SSP labels, distance map, and 3d coordinates	https://dl.fbaipublicfiles.com/fair-esm/structural-data/pkl.tar.gz
msas	a3m files containing MSA for each domain	https://dl.fbaipublicfiles.com/fair-esm/structural-data/msas.tar.gz

Pre-training Dataset Split

The split files establishing which UniRef50 clusters were used as held-out evaluation set for pre-training in Rives et al. 2019 and Rao et al. 2021 can be found here:

UniRef50 IDs of evaluation set: 3.016 M clusters
UniRef100 IDs of evaluation set: 13.745 M proteins, expanding the same UniRef50 clusters.

These files only contain only the UniRef50 IDs and UniRef100 IDs corresponding to the UniRef database, 2018-03 release which is released by the UniProt Consortium under a Creative Commons Attribution (CC BY 4.0) License.

Citations

If you find the models useful in your research, we ask that you cite the relevant paper:

@article{rives2019biological,
  author={Rives, Alexander and Meier, Joshua and Sercu, Tom and Goyal, Siddharth and Lin, Zeming and Liu, Jason and Guo, Demi and Ott, Myle and Zitnick, C. Lawrence and Ma, Jerry and Fergus, Rob},
  title={Biological Structure and Function Emerge from Scaling Unsupervised Learning to 250 Million Protein Sequences},
  year={2019},
  doi={10.1101/622803},
  url={https://www.biorxiv.org/content/10.1101/622803v4},
  journal={bioRxiv}
}

For the self-attention contact prediction:

@article{rao2020transformer,
  author = {Rao, Roshan M and Meier, Joshua and Sercu, Tom and Ovchinnikov, Sergey and Rives, Alexander},
  title={Transformer protein language models are unsupervised structure learners},
  year={2020},
  doi={10.1101/2020.12.15.422761},
  url={https://www.biorxiv.org/content/10.1101/2020.12.15.422761v1},
  journal={bioRxiv}
}

For the MSA Transformer:

@article{rao2021msa,
  author = {Rao, Roshan and Liu, Jason and Verkuil, Robert and Meier, Joshua and Canny, John F. and Abbeel, Pieter and Sercu, Tom and Rives, Alexander},
  title={MSA Transformer},
  year={2021},
  doi={10.1101/2021.02.12.430858},
  url={https://www.biorxiv.org/content/10.1101/2021.02.12.430858v1},
  journal={bioRxiv}
}

For variant prediction using ESM-1v:

@article{meier2021language,
  author = {Meier, Joshua and Rao, Roshan and Verkuil, Robert and Liu, Jason and Sercu, Tom and Rives, Alexander},
  title = {Language models enable zero-shot prediction of the effects of mutations on protein function},
  year={2021},
  doi={10.1101/2021.07.09.450648},
  url={https://www.biorxiv.org/content/10.1101/2021.07.09.450648v1},
  journal={bioRxiv}
}

Much of this code builds on the fairseq sequence modeling framework. We use fairseq internally for our protein language modeling research. We highly recommend trying it out if you'd like to pre-train protein language models from scratch.

License

This source code is licensed under the MIT license found in the LICENSE file in the root directory of this source tree.

Comments

Provide pre-training code?

Hi there!

I'm trying to compare ESM to UniRep, the embedding from the Church lab, for variant function prediction. Eventually, there are a few proteins our lab would like to optimize, and ESM has some advantages over UniRep. I need to "evolutionarily fine tune" ESM, as the Church lab does for UniRep: refine the global model's weights by continuing training on a small neighborhood (~100k sequences) around the target protein.

Could y'all provide any of the code you used in the pre-training task? Eg, your implementations of noising / masking, your loss function, or your gradient descent function?

Thank you, I think ESM is super cool! Best, Jacob

opened by Jacoberts 14

ESMAtlas api access: Rate limit?

Hi,

I would like to automatically retrieve structures from ESMAtlas. Basically I have a list of MGinfy IDs that I want to retrieve.

I do that using the following:

def get_esm_pdb_file(mgnify_id, out_dir):
    url = f'https://api.esmatlas.com/fetchPredictedStructure/{mgnify_id}.pdb'
    response = requests.get(url, stream=True)
    if response.status_code == 200:
        pdb_file = os.path.join(out_dir,f'{mgnify_id}.pdb')
        with open(pdb_file, 'w') as f:
            f.write(response.text)
    else:
        print(mgnify_id, response.status_code)
        print(response.text)

To speed it up, I parallelize it over my list using a ProcessPool. For some IDs I then get the following:

MGYP000547115894 403
{"message":"Forbidden"}
MGYP000526977164 403
{"message":"Forbidden"}
MGYP003555372774 403
{"message":"Forbidden"}

Could you provide some details if/what rate limiting is in place for querying ESMAtlas, or if there is another way to retrieve structure files?

Thanks!

opened by fteufel 11

Predicted ESM2 logits depend on other elements within a batch

Thank you so much for all the work on ESM and ESM2. I ran into some surprising behaviour:

Bug description ESM2 predicts slightly different logits even when in eval mode depending on other elements within a batch.

Reproduction steps

device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')
print(f'device: {device}')

model_path = "facebookresearch/esm:main"
model_name = f"esm2_t36_3B_UR50D"
esm_model, alphabet = torch.hub.load(model_path, model_name)
    
esm_model = esm_model.eval().cuda()
batch_converter = alphabet.get_batch_converter()

# Those are arbitrary sequences, doesn't matter which ones are used
sequences = [
    'A' * 255,
    'Y' * 310
]

model_input = batch_converter([(None, seq) for seq in sequences[:2]])[2]
model_input = model_input.to(device)

# Here is the surprising part:
logits1 = esm_model(model_input[[0]])['logits']
logits2 = esm_model(model_input)['logits']

torch.linalg.norm(logits1 - logits2[0])

tensor(0.3426, device='cuda:0')

This gives roughly 0.3426 - with many values significantly different than zero. I was expecting this to be due to some kind of batch norm like functionality, but, the model is in eval mode.

opened by FedericoV 11

Error when loading esm transformer

model, alphabet = esm.pretrained.esm_msa1b_t12_100M_UR50S() results in:

Traceback (most recent call last):
  File "/home/kevyan/anaconda3/envs/pytorch_p37/lib/python3.7/site-packages/esm/pretrained.py", line 27, in load_hub_workaround
    data = torch.hub.load_state_dict_from_url(url, progress=False, map_location='cpu')
  File "/home/kevyan/anaconda3/envs/pytorch_p37/lib/python3.7/site-packages/torch/hub.py", line 504, in load_state_dict_from_url
    raise RuntimeError('Only one file(not dir) is allowed in the zipfile')
RuntimeError: Only one file(not dir) is allowed in the zipfile
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
  File "/home/kevyan/anaconda3/envs/pytorch_p37/lib/python3.7/site-packages/IPython/core/interactiveshell.py", line 3331, in run_code
    exec(code_obj, self.user_global_ns, self.user_ns)
  File "<ipython-input-15-00c0cd3f832b>", line 1, in <module>
    model, alphabet = esm.pretrained.esm_msa1b_t12_100M_UR50S()
  File "/home/kevyan/anaconda3/envs/pytorch_p37/lib/python3.7/site-packages/esm/pretrained.py", line 191, in esm_msa1b_t12_100M_UR50S
    return load_model_and_alphabet_hub("esm_msa1b_t12_100M_UR50S")
  File "/home/kevyan/anaconda3/envs/pytorch_p37/lib/python3.7/site-packages/esm/pretrained.py", line 47, in load_model_and_alphabet_hub
    model_data = load_hub_workaround(url)
  File "/home/kevyan/anaconda3/envs/pytorch_p37/lib/python3.7/site-packages/esm/pretrained.py", line 32, in load_hub_workaround
    f"{torch.hub.get_dir()}/checkpoints/{fn}",
AttributeError: module 'torch.hub' has no attribute 'get_dir'

As a workaround, I tried downloading the weights directly and loading them: model, alphabet = load_model_and_alphabet('/home/kevyan/.cache/torch/checkpoints/esm_msa1b_t12_100M_UR50S.pt')

Traceback (most recent call last):
  File "/home/kevyan/anaconda3/envs/pytorch_p37/lib/python3.7/site-packages/IPython/core/interactiveshell.py", line 3331, in run_code
    exec(code_obj, self.user_global_ns, self.user_ns)
  File "<ipython-input-16-75482e322f76>", line 1, in <module>
    model, alphabet = load_model_and_alphabet('/home/kevyan/.cache/torch/checkpoints/esm_msa1b_t12_100M_UR50S.pt')
  File "/home/kevyan/anaconda3/envs/pytorch_p37/lib/python3.7/site-packages/esm/pretrained.py", line 21, in load_model_and_alphabet
    return load_model_and_alphabet_local(model_name)
  File "/home/kevyan/anaconda3/envs/pytorch_p37/lib/python3.7/site-packages/esm/pretrained.py", line 57, in load_model_and_alphabet_local
    if _has_regression_weights(model_name):
NameError: name 'model_name' is not defined

opened by yangkky 11

the 41 deep mutational scanning datasets

Hi, @joshim5 Thanks for your work. I was reading your paper 'Language models enable zero-shot prediction of the effects of mutations on protein function' for understanding the model and also want to apply the datasets into my modified model. I searched on the github but couldn't find where the 41 datasets are located (as in Figure 3)? Can you guide me where it is or any links that will get me access to them? Thanks a lot.

I can

opened by lzhangUT 9
Provide MGnify sequences available in ESM Metagenomic Atlas

It would be useful if you could offer a download of the 500 million sequences of the atlas as a fasta file. I want this because it allows me to do fast local searches. The online search capability provided by the atlas takes many minutes to do a single search and we have much faster (but less sensitive) search methods. I'd like to allow users of our ChimeraX visualization software to quickly search the atlas.

I'm aware that I can get the sequences from the EBI MGnify database (2.5 billion sequences), then filter them using the stats.parquet file for the atlas to just the ones used in the atlas. I am pursing that currently, but there appear to be no mirrors of the MGnify database and downloading it from EBI to the United States will take about 10 days to transfer this 250 Gbytes apparently bottlenecked by EBI providing only ~1 Mbit/sec. A download directly from Meta would be 1/5 the size, and if you use better compression than what EBI is using (e.g. bzip2 or xzip instead of gzip) it could be 1/10th the size ~25 Gbytes and on a much faster connection could be downloaded in hours instead of 10 days.

I realize I could scrape the sequences from the 15 Tbytes of structure prediction files that the Atlas provides -- but I'd prefer to not download 500 times more data than I actually need.

Thanks!

Tom Goddard University of California, San Francisco ChimeraX molecular visualization software developer

opened by tomgoddard 8

unable to import esm

I install by pip install fair-esm. Then try the example code:


import torch
import esm

# Load ESM-2 model
model, alphabet = esm.pretrained.esm2_t48_15B_UR50D()
batch_converter = alphabet.get_batch_converter()
model.eval()  # disables dropout for deterministic results

The following error is raised:

ModuleNotFoundError                       Traceback (most recent call last)
Untitled-1.ipynb Cell 1 in <cell line: 2>()
      [1](vscode-notebook-cell:Untitled-1.ipynb?jupyter-notebook#W0sdW50aXRsZWQ%3D?line=0) import torch
----> [2](vscode-notebook-cell:Untitled-1.ipynb?jupyter-notebook#W0sdW50aXRsZWQ%3D?line=1) import esm
      [4](vscode-notebook-cell:Untitled-1.ipynb?jupyter-notebook#W0sdW50aXRsZWQ%3D?line=3) # Load ESM-2 model
      [5](vscode-notebook-cell:Untitled-1.ipynb?jupyter-notebook#W0sdW50aXRsZWQ%3D?line=4) model, alphabet = esm.pretrained.esm2_t48_15B_UR50D()

File ~/anaconda3/envs/torch/lib/python3.8/site-packages/esm/__init__.py:9, in <module>
      6 from .version import version as __version__  # noqa
      8 from .data import Alphabet, BatchConverter, FastaBatchedDataset  # noqa
----> 9 from .model.esm1 import ProteinBertModel  # noqa
     10 from .model.esm2 import ESM2  # noqa
     11 from .model.msa_transformer import MSATransformer  #noqa

ModuleNotFoundError: No module named 'esm.model.esm1'; 'esm.model' is not a package

opened by pengzhangzhi 8

What model is needed for regression_data in esm.pretrained.load_model_and_alphabet_local？
Bug description Use esm.pretrained.load_model_and_alphabet_local( 'esm1_t12_85M_UR50S.pt'), i need to load esm1_t12_85M_UR50S-contact-regression.pt, but in fact this is not provided.

Logs

Output goes here

Additional context If you want to load the local esm model, besides the esm1_t12_85M_UR50S.pt you provided, do you need any regression model? What is the meaning of regression_data in your code esm.pretrained.load_model_and_alphabet_local .
opened by xmhh 8
Broken pip install

Hello,

thank you for the amazing work!

I just would like to let you know, that the latest modification of README has broken pip install, as there is a README.rst among data_files in the setup.py, but now the repo has README.md instead.

Best regards, Raman

opened by SamusRam 8

ModuleNotFoundError - "model" directory is missed in setup.py

Bug description "model" directory is missed in "packages" argument in setup.py script. That issue raised ModuleNotFoundError: No module named 'esm.model' exception.

Reproduction steps Try to install esm package from PyPI or from the GitHub repository.

Expected behavior I was expected successful installation.

Logs

---------------------------------------------------------------------------
ModuleNotFoundError                       Traceback (most recent call last)
Input In [1], in <cell line: 2>()
      1 import torch
----> 2 import esm

File ~/protein_embeddings/env/lib/python3.8/site-packages/esm/__init__.py:9, in <module>
      6 from .version import version as __version__  # noqa
      8 from .data import Alphabet, BatchConverter, FastaBatchedDataset  # noqa
----> 9 from .model.esm1 import ProteinBertModel  # noqa
     10 from .model.esm2 import ESM2  # noqa
     11 from .model.msa_transformer import MSATransformer  #noqa

ModuleNotFoundError: No module named 'esm.model'

opened by ptynecki 7

download the model weights to local

Hi, @rmrao, I am interested in using the esm-iv models (1-5), it takes a long time to download the model/model weights every time when I run one sequence. I wonder if I can download the model weights (or model) into my local workspace, as in Azure Databricks? so It would read the model very quickly when I need to run a lot of sequences. if so, what would be the code to do that? THank you.

opened by lzhangUT 7
Only CPU Malfunction -- HF Colab Notebook

Hello, I am trying to utilize ESM as made available in the HuggingFace Colab Notebook. I do not have a local GPU to use, so have been attempting the only CPU mode. The first issue is in preparing the model and tokenizer where it states " If using GPU, use model.cuda() to transfer the model to GPU." and the code below is:

from transformers import AutoTokenizer, EsmForProteinFolding

tokenizer = AutoTokenizer.from_pretrained("facebook/esmfold_v1") model = EsmForProteinFolding.from_pretrained("facebook/esmfold_v1", low_cpu_mem_usage=True)

model = model.cuda()**

To get around this, I simply removed the "model = model.cuda()" and the code ran fine. Next, down the line, I got to the Tokenization step which requires a GPU:

tokenized_input = tokenized_input.cuda()

Obviously when ran the error of "NO GPU" was displayed. So I went ahead and tried to run actual prediction lines:

import torch

with torch.no_grad(): output = model(tokenized_input)**

This displayed the following error:

RuntimeError: "LayerNormKernelImpl" not implemented for 'Half'

So how are you supposed to run this sheet without a GPU? Is there something in the tokenization stage I should alter to provide good inputs for the prediction model?

opened by coopersvajda 0
download model from huggingface

Hi dear support team

I want to use your awesome transformer, but our HPC cluster does not have internet access due to firewall restriction. I download model and config, ... files from huggingface (https://huggingface.co/facebook/esm2_t33_650M_UR50D/tree/main), i try to load model with: path = Path(file).resolve().parent.joinpath('esm') model, alphabet = esm.pretrained(path) This trick dont work! and this error occur: TypeError: 'module' object is not callable

How i can to load model?

Thanks

opened by saeid976 1
Adjust header in output file
From some research, it looked to me that the pertinent part of the name in the header was the "Seita.9G099800" part, so I wanted to split that from the rest. First, see if the header is longer than one element and then isolate the "Seita.9G099800" portion:

if len(header.split()) > 1: first_elem = header.split()[0] name = '.'.join(first_elem.split('.')[:2])

If the header is just one word, simply assign the header to the name and put that in the output_file:

else: name = header output_file = args.pdb / f"{name}.pdb"

Please let me know if this is satisfactory, or if you had something else in mind.
CLA Signed
opened by dawsonhunt 3
EMSIF1-Encoder output as structure representation

rep = esm.inverse_folding.multichain_util.get_encoder_output_for_complex( model, alphabet, coords, target_chain_id). I found that target_chain_id can only be a single chain. Can the input (target_chain_id) improve the code to get multi-chain structure embeddings?

opened by LDAIprotein 0
can't download the contact-regression.pt for those ESM models

Hi，thank you for your great jobs. Though the big language model e.g. <esm2_t33_650M_UR50D.pt> can be download directly from the repo, when I init the model, it seems need the <esm2_t33_650M_UR50D-contact-regression.pt> which can't download directly from the repo. Could you please tell me how to download those <contact-regression.pt> files directly in the repo?

opened by imSeaton 0

Releases(v1.0.3)

v1.0.3(Oct 18, 2022)
Recent fixes

Source code(tar.gz)
Source code(zip)
v1.0.2(Aug 23, 2022)
Restrict fairscale usage only to FSDP-specific inference example

Fixed setup.py to include esm.model

README updates

Source code(tar.gz)
Source code(zip)
v1.0.0(Aug 22, 2022)

Source code and model weights release of ESM-2 Protein Language Model from the paper Language models of protein sequences at the scale of evolution enable accurate structure prediction (Z. Lin, H. Akin, R. Rao, B. Hie, Z. Zhu, W. Lu, A.S. Costa, M. Fazel-Zarandi, T. Sercu, S. Candido, A. Rives, 2022)
Source code(tar.gz)
Source code(zip)

Owner

Meta Research

GitHub Repository

Code for CPM-2 Pre-Train

CPM-2 Pre-Train Pre-train CPM-2 此分支为110亿非 MoE 模型的预训练代码，MoE 模型的预训练代码请切换到 moe 分支 CPM-2技术报告请参考link。 0 模型下载请在智源资源下载页面进行申请，文件介绍如下：文件名描述参数大小 100000.tar

136 Dec 28, 2022

Yet another video caption

5 May 26, 2022

ML-Ensemble – high performance ensemble learning

A Python library for high performance ensemble learning ML-Ensemble combines a Scikit-learn high-level API with a low-level computational graph framew

764 Dec 31, 2022

Code for the paper One Thing One Click: A Self-Training Approach for Weakly Supervised 3D Semantic Segmentation, CVPR 2021.

One Thing One Click One Thing One Click: A Self-Training Approach for Weakly Supervised 3D Semantic Segmentation (CVPR2021) Code for the paper One Thi

44 Dec 12, 2022

Accelerating BERT Inference for Sequence Labeling via Early-Exit

Sequence-Labeling-Early-Exit Code for ACL 2021 paper: Accelerating BERT Inference for Sequence Labeling via Early-Exit Requirement: Please refer to re

23 Oct 14, 2022

This repository will be a summary and outlook on all our open, medical, AI advancements.

medical by LAION This repository will be a summary and outlook on all our open, medical, AI advancements. See the medical-general channel in the medic

18 Dec 30, 2022

Jax/Flax implementation of Variational-DiffWave.

jax-variational-diffwave Jax/Flax implementation of Variational-DiffWave. (Zhifeng Kong et al., 2020, Diederik P. Kingma et al., 2021.) DiffWave with

37 Dec 16, 2022

Boost learning for GNNs from the graph structure under challenging heterophily settings. (NeurIPS'20)

Beyond Homophily in Graph Neural Networks: Current Limitations and Effective Designs Jiong Zhu, Yujun Yan, Lingxiao Zhao, Mark Heimann, Leman Akoglu,

70 Dec 18, 2022

Code for Overinterpretation paper Overinterpretation reveals image classification model pathologies

Overinterpretation This repository contains the code for the paper: Overinterpretation reveals image classification model pathologies Authors: Brandon

17 Dec 10, 2022

Companion code for the paper "An Infinite-Feature Extension for Bayesian ReLU Nets That Fixes Their Asymptotic Overconfidence" (NeurIPS 2021)

ReLU-GP Residual (RGPR) This repository contains code for reproducing the following NeurIPS 2021 paper: @inproceedings{kristiadi2021infinite, title=

4 Dec 26, 2021

A repository that shares tuning results of trained models generated by TensorFlow / Keras. Post-training quantization (Weight Quantization, Integer Quantization, Full Integer Quantization, Float16 Quantization), Quantization-aware training. TensorFlow Lite. OpenVINO. CoreML. TensorFlow.js. TF-TRT. MediaPipe. ONNX. [.tflite,.h5,.pb,saved_model,tfjs,tftrt,mlmodel,.xml/.bin, .onnx]

PINTO_model_zoo Please read the contents of the LICENSE file located directly under each folder before using the model. My model conversion scripts ar

2.4k Jan 05, 2023