Upload README.md with huggingface_hub

README.md

@@ -3,8 +3,9 @@
 tags:
 - LoRA
 - protein language model
+base_model: facebook/esm2_t12_35M_UR50D
 datasets:
-- SoftDis
+- CQSB/SoftDis
 ---
 
 # LoRA-DR-suite
@@ -70,21 +71,24 @@ import torch.nn.functional as F
 device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
 model = model.to(device)
 
-# example sequence (DP02948)
-sequence = "METLTLLWRASSSCLLVVLSHSLLRLLGVRCLEKSGAQPNKLFRPPCCQKGPSFARHSRCVYYTQSRE"
+# example sequence 
+sequence = "TAIWEQHTVTLHRAPGFGFGIAISGGRDNPHFQSGETSIVISDVLKG"
 
 # each pre-trained model adds its own special tokens to the tokenized sequence,  
-# special_tokens_mask allows to deal with them (padding included, for batched inputs)
-# without changing the code
-inputs = tokenizer([sequence], return_tensors="pt", return_special_tokens_mask=True)
+# special_tokens_mask allows to deal with them (padding included, for batched 
+# inputs) without changing the code
+inputs = tokenizer(
+    [sequence], return_tensors="pt", return_special_tokens_mask=True
+)
 input_ids = inputs['input_ids'].to(device)
 attention_mask = inputs['attention_mask'].to(device)
 special_tokens_mask = inputs['special_tokens_mask'].bool()
 
 # extract predicted disorder probability
 with torch.inference_mode():
-    output = model(input_ids=input_ids,attention_mask=attention_mask).squeeze(0).cpu()
-disorder_proba = F.softmax(output[~special_tokens_mask], dim=-1)[:, 1]
+    output = model(input_ids, attention_mask).logits.cpu()
+output = output[~special_tokens_mask, :]
+disorder_proba = F.softmax(output, dim=-1)[:, 1]
 ```
 
 ## How to cite