Merge branch 'clean' into main

clean_selection/analyze_new_price_pppl.ipynb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:be031f05f2b7d92cc5ee89671a8ddd9d844ea0c8e9b803f5dcb70bdcab2b67a5
+size 228782

clean_selection/compute_new_price_pppl.py

@@ -0,0 +1,34 @@
+from transformers import AutoModelForMaskedLM, AutoTokenizer
+from protein_conformal.scope_utils import calculate_pppl
+import torch
+
+torch.cuda.set_device(0)
+device = torch.device("cuda:0")
+
+# Load the model and tokenizer
+model_name = "facebook/esm1b_t33_650M_UR50S" ## this is the backbone ESM model used for CLEAN
+tokenizer = AutoTokenizer.from_pretrained(model_name)
+model = AutoModelForMaskedLM.from_pretrained(model_name)
+model = model.cuda()
+
+# load sequences from pa][ndas dataframe "data/new.csv" under "Sequence" column
+import pandas as pd
+df = pd.read_csv("data/new.csv", sep = "\t")
+new_sequences = df["Sequence"]
+
+# same for "data/price.csv"
+df = pd.read_csv("data/price.csv", sep = "\t")
+price_sequences = df["Sequence"]
+
+# Calculate PPPL for all sequences
+pppl_new = [calculate_pppl(model, tokenizer, seq, device) for seq in new_sequences]
+pppl_price = [calculate_pppl(model, tokenizer, seq, device) for seq in price_sequences]
+
+# Save the results
+df["PPPL"] = pppl_price
+df.to_csv("data/price_w_pppl.csv", sep = "\t", index = False)
+
+df = pd.read_csv("data/new.csv", sep = "\t")
+df["PPPL"] = pppl_new
+
+df.to_csv("data/new_w_pppl.csv", sep = "\t", index = False)

clean_selection/get_clean_dists.ipynb

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:aedd07adb2c5b2e7a599c94faceb57d292444f3b72f270bba83e26c42290492b
-size 9023
+oid sha256:c7b40ac630e522634e58c21508a00602879325bd25938288f731573a379f4e3b
+size 41395

clean_selection/process_clean_ec.ipynb

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:744eae0442a80d617d24342f3e00cf49c3460edb260054e9f0dca173c87e5968
-size 5854
+oid sha256:699e8fd4cf4f956d420f2a6ed9ab3c9c38c5afa605d77add397b6638c189e6a8
+size 10824