yinuozhang commited on
Commit
c860158
1 Parent(s): b84fc7f
Files changed (2) hide show
  1. README.md +4 -1
  2. config.json +1 -1
README.md CHANGED
@@ -25,12 +25,15 @@ You can try out the MetaLATTE model directly in your browser:
25
 
26
  ```python
27
  from transformers import AutoTokenizer, AutoModel, AutoConfig
 
 
 
28
 
29
  tokenizer = AutoTokenizer.from_pretrained("facebook/esm2_t33_650M_UR50D")
30
  config = AutoConfig.from_pretrained("ChatterjeeLab/MetaLATTE")
31
  model = AutoModel.from_pretrained("ChatterjeeLab/MetaLATTE", config=config)
32
 
33
- # Use the model
34
  sequence = "AVYNIGWSFNVNGARGKSFRAGDVLVFKYIKGQHNVVAVNGRGYASCSAPRGARTYSSGQDRIKLTRGQNYFICSFPGHCGGGMKIAINAK"
35
  inputs = tokenizer(sequence, return_tensors="pt")
36
  raw_probs, predictions = model(**inputs)
 
25
 
26
  ```python
27
  from transformers import AutoTokenizer, AutoModel, AutoConfig
28
+ from metalatte import MetaLATTEConfig, MetaLATTEModel
29
+ AutoConfig.register("metalatte", MetaLATTEConfig)
30
+ AutoModel.register(MetaLATTEConfig, MetaLATTEModel)
31
 
32
  tokenizer = AutoTokenizer.from_pretrained("facebook/esm2_t33_650M_UR50D")
33
  config = AutoConfig.from_pretrained("ChatterjeeLab/MetaLATTE")
34
  model = AutoModel.from_pretrained("ChatterjeeLab/MetaLATTE", config=config)
35
 
36
+ model.eval()
37
  sequence = "AVYNIGWSFNVNGARGKSFRAGDVLVFKYIKGQHNVVAVNGRGYASCSAPRGARTYSSGQDRIKLTRGQNYFICSFPGHCGGGMKIAINAK"
38
  inputs = tokenizer(sequence, return_tensors="pt")
39
  raw_probs, predictions = model(**inputs)
config.json CHANGED
@@ -2,7 +2,7 @@
2
  "architectures": [
3
  "MultitaskProteinModel"
4
  ],
5
- "model_type": "esm",
6
  "num_labels": 15,
7
  "hidden_size": 1280,
8
  "num_hidden_layers": 33,
 
2
  "architectures": [
3
  "MultitaskProteinModel"
4
  ],
5
+ "model_type": "metalatte",
6
  "num_labels": 15,
7
  "hidden_size": 1280,
8
  "num_hidden_layers": 33,