andrecornman
commited on
Commit
•
77bed56
1
Parent(s):
4e5f31b
Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,85 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
datasets:
|
3 |
+
- tattabio/OMG
|
4 |
+
license: apache-2.0
|
5 |
+
---
|
6 |
+
|
7 |
+
# gLM2_150M
|
8 |
+
|
9 |
+
gLM2 is a mixed-modality genomic language model, trained on the [`OMG Dataset`](https://huggingface.co/datasets/tattabio/OMG).
|
10 |
+
The model encodes a genomic scaffold with both both amino-acid and DNA tokens.
|
11 |
+
|
12 |
+
gLM2 is trained at two scales: 150M and 650M parameters (available at [`tattabio/gLM2_650M`](https://huggingface.co/tattabio/gLM2_650M)).
|
13 |
+
|
14 |
+
See [https://github.com/TattaBio/gLM2](https://github.com/TattaBio/gLM2) for inference scripts.
|
15 |
+
|
16 |
+
### Model Description
|
17 |
+
|
18 |
+
gLM2 is a transformer encoder trained with the masked language modeling objective.
|
19 |
+
It encodes a genomic contig as a sequence of protein coding sequences (CDS) and DNA inter-genic sequences (IGS).
|
20 |
+
CDS elements are tokenized using per-amino acid tokens, and IGS elements are tokenized using per-nucleotide tokens.
|
21 |
+
|
22 |
+
|
23 |
+
- To encode the genomic strand, we prepended each genomic element with a special token, either `<+>` or `<->` to indicate the positive and negative strands.
|
24 |
+
- To avoid collision between amino acid and nucleotide tokens, the tokenizer expects all amino acids to be uppercase, and all nucleotides to be lowercase.
|
25 |
+
|
26 |
+
UPDATE(09/2024): We updated the model with longer context length (4096 tokens vs. 2048 tokens) and per-nucleotide IGS tokenization instead of BPE.
|
27 |
+
|
28 |
+
## Getting Started
|
29 |
+
|
30 |
+
|
31 |
+
```python
|
32 |
+
import torch
|
33 |
+
from transformers import AutoModel, AutoTokenizer
|
34 |
+
|
35 |
+
model = AutoModel.from_pretrained('tattabio/gLM2_150M', torch_dtype=torch.bfloat16, trust_remote_code=True).cuda()
|
36 |
+
tokenizer = AutoTokenizer.from_pretrained('tattabio/gLM2_150M', trust_remote_code=True)
|
37 |
+
|
38 |
+
# A contig with two proteins and an inter-genic sequence.
|
39 |
+
# NOTE: Nucleotides should always be lowercase, and prepended with `<+>`.
|
40 |
+
sequence = "<+>MALTKVEKRNRIKRRVRGKISGTQASPRLSVYKSNK<+>aatttaaggaa<->MLGIDNIERVKPGGLELVDRLVAVNRVTKVTKGGRAFGFSAIVVVGNED"
|
41 |
+
|
42 |
+
# Tokenize the sequence.
|
43 |
+
encodings = tokenizer([sequence], return_tensors='pt')
|
44 |
+
|
45 |
+
# Extract embeddings.
|
46 |
+
with torch.no_grad():
|
47 |
+
embeddings = model(encodings.input_ids.cuda(), output_hidden_states=True).last_hidden_state
|
48 |
+
|
49 |
+
```
|
50 |
+
|
51 |
+
### Training Data
|
52 |
+
|
53 |
+
gLM2 is trained on the [`OMG`](https://huggingface.co/datasets/tattabio/OMG) dataset.
|
54 |
+
To improve the dataset balance and remove near-duplicate examples, the data is tokenized and pruned by applying Semantic Deduplication [SemDedup](https://arxiv.org/abs/2303.09540).
|
55 |
+
We use an embedding distance threshold of 2e-3, resulting in 49% of the dataset being pruned.
|
56 |
+
|
57 |
+
## Training Details
|
58 |
+
|
59 |
+
- Pretraining tokens: 315B
|
60 |
+
- Context length: 4096
|
61 |
+
- Masking rate: 30%
|
62 |
+
- Learning rate: 1e-3
|
63 |
+
- Optimizer: AdamW (betas = (0.9, 0.95))
|
64 |
+
- Mixed precision training: bfloat16
|
65 |
+
- Weight decay: 0.1
|
66 |
+
|
67 |
+
|
68 |
+
## Citation
|
69 |
+
|
70 |
+
**BioRxiv:**
|
71 |
+
[https://www.biorxiv.org/content/10.1101/2024.08.14.607850](https://www.biorxiv.org/content/10.1101/2024.08.14.607850)
|
72 |
+
|
73 |
+
**BibTeX:**
|
74 |
+
|
75 |
+
```@article {Cornman2024.08.14.607850,
|
76 |
+
author = {Cornman, Andre and West-Roberts, Jacob and Camargo, Antonio Pedro and Roux, Simon and Beracochea, Martin and Mirdita, Milot and Ovchinnikov, Sergey and Hwang, Yunha},
|
77 |
+
title = {The OMG dataset: An Open MetaGenomic corpus for mixed-modality genomic language modeling},
|
78 |
+
elocation-id = {2024.08.14.607850},
|
79 |
+
year = {2024},
|
80 |
+
doi = {10.1101/2024.08.14.607850},
|
81 |
+
publisher = {Cold Spring Harbor Laboratory},
|
82 |
+
URL = {https://www.biorxiv.org/content/early/2024/08/17/2024.08.14.607850},
|
83 |
+
eprint = {https://www.biorxiv.org/content/early/2024/08/17/2024.08.14.607850.full.pdf},
|
84 |
+
journal = {bioRxiv}
|
85 |
+
}
|