Update README.md
Browse files
README.md
CHANGED
@@ -15,35 +15,44 @@ pip install git+https://github.com/etowahadams/interprot.git
|
|
15 |
|
16 |
## Usage
|
17 |
|
18 |
-
|
19 |
```python
|
|
|
|
|
20 |
from safetensors.torch import load_file
|
21 |
from interprot.sae_model import SparseAutoencoder
|
|
|
22 |
|
23 |
-
|
24 |
-
|
25 |
-
|
26 |
-
```
|
27 |
|
28 |
-
Load ESM
|
29 |
-
```
|
30 |
-
import torch
|
31 |
-
from transformers import AutoTokenizer, EsmModel
|
32 |
-
|
33 |
-
# Load ESM model and tokenizer
|
34 |
tokenizer = AutoTokenizer.from_pretrained("facebook/esm2_t33_650M_UR50D")
|
35 |
esm_model = EsmModel.from_pretrained("facebook/esm2_t33_650M_UR50D")
|
36 |
|
37 |
-
#
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
38 |
seq = "MKTVRQERLKSIVRILERSKEPVSGAQLAEELSVSRQVVAAIVQDIAYLRSLGYNIVATPRGYVLAGG"
|
39 |
-
esm_layer = 24
|
40 |
|
41 |
-
|
|
|
42 |
with torch.no_grad():
|
43 |
outputs = esm_model(**inputs, output_hidden_states=True)
|
44 |
-
esm_layer_acts = outputs.hidden_states[esm_layer] # (1, sequence length + 2, 1280)
|
45 |
|
46 |
-
#
|
47 |
-
|
48 |
-
|
|
|
|
|
|
|
49 |
```
|
|
|
15 |
|
16 |
## Usage
|
17 |
|
18 |
+
Install InterProt, load ESM and SAE
|
19 |
```python
|
20 |
+
import torch
|
21 |
+
from transformers import AutoTokenizer, EsmModel
|
22 |
from safetensors.torch import load_file
|
23 |
from interprot.sae_model import SparseAutoencoder
|
24 |
+
from huggingface_hub import hf_hub_download
|
25 |
|
26 |
+
ESM_DIM = 1280
|
27 |
+
SAE_DIM = 4096
|
28 |
+
LAYER = 24
|
|
|
29 |
|
30 |
+
# Load ESM model
|
|
|
|
|
|
|
|
|
|
|
31 |
tokenizer = AutoTokenizer.from_pretrained("facebook/esm2_t33_650M_UR50D")
|
32 |
esm_model = EsmModel.from_pretrained("facebook/esm2_t33_650M_UR50D")
|
33 |
|
34 |
+
# Load SAE model
|
35 |
+
checkpoint_path = hf_hub_download(
|
36 |
+
repo_id="liambai/InterProt-ESM2-SAEs",
|
37 |
+
filename="esm2_plm1280_l24_sae4096.safetensors"
|
38 |
+
)
|
39 |
+
sae_model = SparseAutoencoder(ESM_DIM, SAE_DIM)
|
40 |
+
sae_model.load_state_dict(load_file(checkpoint_path))
|
41 |
+
```
|
42 |
+
|
43 |
+
ESM -> SAE inference on an amino acid sequence of length `L`
|
44 |
+
```
|
45 |
seq = "MKTVRQERLKSIVRILERSKEPVSGAQLAEELSVSRQVVAAIVQDIAYLRSLGYNIVATPRGYVLAGG"
|
|
|
46 |
|
47 |
+
# Tokenize sequence and run ESM inference
|
48 |
+
inputs = tokenizer(seq, padding=True, return_tensors="pt")
|
49 |
with torch.no_grad():
|
50 |
outputs = esm_model(**inputs, output_hidden_states=True)
|
|
|
51 |
|
52 |
+
# esm_layer_acts has shape (L+2, ESM_DIM), +2 for BoS and EoS tokens
|
53 |
+
esm_layer_acts = outputs.hidden_states[LAYER][0]
|
54 |
+
|
55 |
+
# Using ESM embeddings from LAYER, run SAE inference
|
56 |
+
sae_acts = sae_model.get_acts(esm_layer_acts) # (L+2, SAE_DIM)
|
57 |
+
sae_acts
|
58 |
```
|