upload json scripts

kundajelab · Aug 27, 2024 · b9d29a6 · b9d29a6
1 parent 2837ae5
commit b9d29a6
Show file tree

Hide file tree

Showing 12 changed files with 293 additions and 33 deletions.
diff --git a/upload_jsons/upload_scripts/README.md b/upload_jsons/upload_scripts/README.md
@@ -0,0 +1,3 @@
+python run_train_test_making_bias.py
+python run_conversion_bias.py
+
diff --git a/upload_jsons/upload_scripts/dnase_run_train_test_making_bias.py b/upload_jsons/upload_scripts/dnase_run_train_test_making_bias.py
@@ -0,0 +1,67 @@
+import pandas as pd
+import os
+
+#model_atac = pd.read_csv("/mnt/lab_data2/anusri/chrombpnet/logs/checkpoint/JAN_02_2023/model_dir_atac.csv",sep=",", header=None)
+#model_atac = pd.read_csv("/mnt/lab_data2/anusri/chrombpnet/logs/checkpoint/JAN_02_2023/model_dir_dnase.csv",sep=",", header=None)
+#model_atac=pd.read_csv("bias_models_atlas.csv", sep=',', header=None)
+model_atac=pd.read_csv("model_dir_dnase_v2.1_bias.csv", sep=',', header=None)
+
+
+
+
+encode_id = {"K562": "ENCSR868FGK",
+        "GM12878": "ENCSR637XSC",
+        "HEPG2": "ENCSR291GJU",
+        "IMR90": "ENCSR200OML",
+        "H1ESC": "ENCDUMMY"}
+
+encode_id = {"K562": "ENCSR000EOT",
+        "GM12878": "ENCSR000EMT",
+        "HEPG2": "ENCSR149XIL",
+        "IMR90": "ENCSR477RTP",
+        "H1ESC": "ENCSR000EMU"}
+
+
+for i,r in model_atac.iterrows():
+	fold = r[0]
+	name = r[1]
+	model_path = r[2]
+
+	#input_peaks=os.path.join(model_path,"chrombpnet_model/filtered.peaks.bed")
+	input_nonpeaks=os.path.join(model_path,"bias_model/filtered.bias_nonpeaks.bed")
+	#test_nonpeaks="/oak/stanford/groups/akundaje/projects/chromatin-atlas-2022/DNASE/"+encode_id[name]+"/negatives_data/test/test."+fold+".filtered.negatives_with_summit.bed"
+	fold="/mnt/lab_data2/anusri/chrombpnet/splits/"+fold+".json"
+	output_path=os.path.join(model_path,"train_test_regions_bias_may_7_2024/")
+
+	if not os.path.isfile(input_nonpeaks):
+		cellline=input_nonpeaks.split("/")[10]
+		biasth=input_nonpeaks.split("/")[11].split("_")[6]
+		foldn=input_nonpeaks.split("/")[11].split("_")[8]
+		#print(cellline,biasth,foldn)
+		if cellline in ["K562", "HEPG2"]:
+			ddatype="DNASE_PE"
+		elif cellline in ["H1ESC"]:
+			ddatype="DNASE_SE"
+		else:
+			print(cellline)
+			break
+		outputdir=os.path.join(model_path,"bias_model/newgen/")
+		if not  os.path.isfile(os.path.join(model_path,"bias_model/newgen/filtered.bias_nonpeaks.bed")):
+			os.makedirs(outputdir, exist_ok=True)
+			print(outputdir)
+			command = "bash make_missing_bed_regions.sh "+cellline+" "+biasth+" "+foldn+" "+outputdir+" "+ddatype
+			os.system(command)
+			print(command)
+			input_nonpeaks=os.path.join(model_path,"bias_model/newgen/filtered.bias_nonpeaks.bed")
+		else:
+			input_nonpeaks=os.path.join(model_path,"bias_model/newgen/filtered.bias_nonpeaks.bed")
+
+	if not os.path.isfile(output_path+"nonpeaks.validationset.bed.gz"):
+		print(output_path)
+		os.makedirs(output_path, exist_ok=True)
+		command=["python get_train_test_regions_bias.py "]+["-inp"]+[input_nonpeaks]+["-f"]+[fold]+["-o"]+[output_path]
+		command = " ".join(command)
+		print(command)
+		os.system(command)
+	else:
+		print(output_path)
diff --git a/upload_jsons/upload_scripts/get_new_tf_model_format.py b/upload_jsons/upload_scripts/get_new_tf_model_format.py
@@ -3,6 +3,7 @@
 import argparse
 from tensorflow.keras.utils import get_custom_objects
 import os
+from tensorflow.keras.layers import Input, Cropping1D, add, Conv1D, GlobalAvgPool1D, Dense, Add, Concatenate, Lambda, Flatten
 
 parser = argparse.ArgumentParser(description="converting model types")
 parser.add_argument("-i", "--input_model")
@@ -13,7 +14,8 @@
 
 
 output_path=os.path.join(args.output_dir, args.file_path+"/")
-custom_objects={"tf": tf}
+#custom_objects={"tf": tf}
+custom_objects={"logcount_predictions":  Lambda(lambda x: tf.math.reduce_logsumexp(x, axis=-1, keepdims=True)), "tf": tf}
 get_custom_objects().update(custom_objects)
 model=load_model(args.input_model,compile=False)
 model.save(output_path)

diff --git a/upload_jsons/upload_scripts/get_train_test_regions.py b/upload_jsons/upload_scripts/get_train_test_regions.py
@@ -15,7 +15,7 @@
 
 peaks = pd.read_csv(args.input_peaks, sep="\t", header=None, names=NARROWPEAK_SCHEMA)
 nonpeaks = pd.read_csv(args.input_nonpeaks, sep="\t", header=None, names=NARROWPEAK_SCHEMA)
-nonpeaks_test =  pd.read_csv(args.input_nonpeaks_test, sep="\t", header=None, names=NARROWPEAK_SCHEMA)
+#nonpeaks_test =  pd.read_csv(args.input_nonpeaks_test, sep="\t", header=None, names=NARROWPEAK_SCHEMA)
 
 splits_dict=json.load(open(args.chr_fold_path))
 
@@ -41,7 +41,7 @@
 path_nonpeaks_valid = os.path.join(args.output_path,"nonpeaks.validationset.bed.gz")
 nonpeaks_valid.to_csv(path_nonpeaks_valid,sep="\t", index=False, header=False, compression='gzip')
 
-path_nonpeaks_test = os.path.join(args.output_path,"nonpeaks.testset.bed.gz")
-nonpeaks_test.to_csv(path_nonpeaks_test,sep="\t", index=False, header=False, compression='gzip')
+#path_nonpeaks_test = os.path.join(args.output_path,"nonpeaks.testset.bed.gz")
+#nonpeaks_test.to_csv(path_nonpeaks_test,sep="\t", index=False, header=False, compression='gzip')
 
 
diff --git a/upload_jsons/upload_scripts/make_missing_bed_regions.sh b/upload_jsons/upload_scripts/make_missing_bed_regions.sh
@@ -0,0 +1,35 @@
+cellline=$1
+biasth=$2
+foldn=$3
+outputdir=$4
+ddatype=$5
+echo "python /mnt/lab_data2/anusri/chrombpnet/src/helpers/hyperparameters/find_bias_hyperparams.py \\
+       --genome=/mnt/lab_data2/anusri/chrombpnet/reference/hg38.genome.fa \\
+       --bigwig=/mnt/lab_data2/anusri/chrombpnet/results/chrombpnet/$ddatype/$cellline/data/$cellline"_unstranded.bw" \\
+       --peaks=/mnt/lab_data2/anusri/chrombpnet/results/chrombpnet/$ddatype/$cellline/data/peaks_no_blacklist.bed \\
+       --nonpeaks=/mnt/lab_data2/anusri/chrombpnet/results/chrombpnet/$ddatype/$cellline/negatives_data_$foldn/negatives_with_summit.be>
+       --outlier_threshold=0.99 \\
+       --chr_fold_path=/mnt/lab_data2/anusri/chrombpnet/splits/fold_$foldn.json \\
+       --inputlen=2114 \\
+       --outputlen=1000 \\
+       --max_jitter=0 \\
+       --filters=128 \\
+       --n_dilation_layers=4 \\
+       --bias_threshold_factor=$biasth \\
+       --output_dir $outputdir"
+
+python /mnt/lab_data2/anusri/chrombpnet/src/helpers/hyperparameters/find_bias_hyperparams.py \
+       --genome=/mnt/lab_data2/anusri/chrombpnet/reference/hg38.genome.fa \
+       --bigwig=/mnt/lab_data2/anusri/chrombpnet/results/chrombpnet/$ddatype/$cellline/data/$cellline"_unstranded.bw" \
+       --peaks=/mnt/lab_data2/anusri/chrombpnet/results/chrombpnet/$ddatype/$cellline/data/peaks_no_blacklist.bed \
+       --nonpeaks=/mnt/lab_data2/anusri/chrombpnet/results/chrombpnet/$ddatype/$cellline/negatives_data_$foldn/negatives_with_summit.bed \
+       --outlier_threshold=0.99 \
+       --chr_fold_path=/mnt/lab_data2/anusri/chrombpnet/splits/fold_$foldn.json \
+       --inputlen=2114 \
+       --outputlen=1000 \
+       --max_jitter=0 \
+       --filters=128 \
+       --n_dilation_layers=4 \
+       --bias_threshold_factor=$biasth \
+       --output_dir $outputdir
+
diff --git a/upload_jsons/upload_scripts/model_dir_dnase_v2.1_bias.csv b/upload_jsons/upload_scripts/model_dir_dnase_v2.1_bias.csv
@@ -0,0 +1,19 @@
+fold_0,K562,/mnt/lab_data2/anusri/chrombpnet/results/chrombpnet/DNASE_PE/K562/nautilus_runs_may18/K562_05.13.2022_bias_128_4_1234_0.5_fold_0/
+fold_1,K562,/oak/stanford/groups/akundaje/projects/chromatin-atlas-2022/chrombpnet/folds/DNASE/K562/K562_07.17.2022_bias_128_4_1234_0.4_fold_1_data_type_DNASE_PE/
+fold_2,K562,/oak/stanford/groups/akundaje/projects/chromatin-atlas-2022/chrombpnet/folds/DNASE/K562/K562_07.07.2022_bias_128_4_1234_0.5_fold_2_data_type_DNASE_PE/
+fold_3,K562,/oak/stanford/groups/akundaje/projects/chromatin-atlas-2022/chrombpnet/folds/DNASE/K562/K562_07.07.2022_bias_128_4_1234_0.5_fold_3_data_type_DNASE_PE/
+fold_4,K562,/oak/stanford/groups/akundaje/projects/chromatin-atlas-2022/chrombpnet/folds/DNASE/K562/K562_07.07.2022_bias_128_4_1234_0.5_fold_4_data_type_DNASE_PE/
+fold_0,HEPG2,/mnt/lab_data2/anusri/chrombpnet/results/chrombpnet/DNASE_PE/HEPG2/copy_HEPG2_06.08.2022_bias_128_4_1234_0.8_fold_0/
+fold_1,HEPG2,/oak/stanford/groups/akundaje/projects/chromatin-atlas-2022/chrombpnet/folds/DNASE/HEPG2/HEPG2_07.13.2022_bias_128_4_1234_0.8_fold_1_data_type_DNASE_PE/
+fold_2,HEPG2,/oak/stanford/groups/akundaje/projects/chromatin-atlas-2022/chrombpnet/folds/DNASE/HEPG2/HEPG2_07.07.2022_bias_128_4_1234_0.8_fold_2_data_type_DNASE_PE/
+fold_3,HEPG2,/oak/stanford/groups/akundaje/projects/chromatin-atlas-2022/chrombpnet/folds/DNASE/HEPG2/HEPG2_07.07.2022_bias_128_4_1234_0.8_fold_3_data_type_DNASE_PE/
+fold_4,HEPG2,/oak/stanford/groups/akundaje/projects/chromatin-atlas-2022/chrombpnet/folds/DNASE/HEPG2/HEPG2_07.07.2022_bias_128_4_1234_0.8_fold_4_data_type_DNASE_PE/
+fold_0,H1ESC,/mnt/lab_data2/anusri/chrombpnet/results/chrombpnet/DNASE_SE/H1ESC/nautilus_runs_apr12/H1ESC_04.09.2022_bias_128_4_1234_0.8_fold_0/
+fold_1,H1ESC,/oak/stanford/groups/akundaje/projects/chromatin-atlas-2022/chrombpnet/folds/DNASE/H1ESC/H1ESC_07.07.2022_bias_128_4_1234_0.8_fold_1_data_type_DNASE_SE/
+fold_2,H1ESC,/oak/stanford/groups/akundaje/projects/chromatin-atlas-2022/chrombpnet/folds/DNASE/H1ESC/H1ESC_07.07.2022_bias_128_4_1234_0.8_fold_2_data_type_DNASE_SE/
+fold_3,H1ESC,/oak/stanford/groups/akundaje/projects/chromatin-atlas-2022/chrombpnet/folds/DNASE/H1ESC/H1ESC_07.07.2022_bias_128_4_1234_0.8_fold_3_data_type_DNASE_SE/
+fold_4,H1ESC,/oak/stanford/groups/akundaje/projects/chromatin-atlas-2022/chrombpnet/folds/DNASE/H1ESC/H1ESC_07.07.2022_bias_128_4_1234_0.8_fold_4_data_type_DNASE_SE/
+
+
+
+
diff --git a/upload_jsons/upload_scripts/new_run_train_test_making.py b/upload_jsons/upload_scripts/new_run_train_test_making.py
@@ -0,0 +1,38 @@
+import pandas as pd
+import os
+
+#model_atac = pd.read_csv("/mnt/lab_data2/anusri/chrombpnet/logs/checkpoint/JAN_02_2023/model_dir_atac.csv",sep=",", header=None)
+#model_atac = pd.read_csv("/mnt/lab_data2/anusri/chrombpnet/logs/checkpoint/JAN_02_2023/model_dir_dnase.csv",sep=",", header=None)
+model_atac = pd.read_csv("/mnt/lab_data2/anusri/chrombpnet/logs/checkpoint/JAN_02_2023/v1/model_dir_dnase_v2.2.csv",sep=",",header=None)
+
+#encode_id = {"K562": "ENCSR868FGK",
+#        "GM12878": "ENCSR637XSC",
+#        "HEPG2": "ENCSR291GJU",
+#        "IMR90": "ENCSR200OML",
+#         "H1ESC": "ENCDUMMY"}
+
+encode_id = {"K562": "ENCSR000EOT",
+        "GM12878": "ENCSR000EMT",
+        "HEPG2": "ENCSR149XIL",
+        "IMR90": "ENCSR477RTP",
+        "H1ESC": "ENCSR000EMU"}
+
+
+for i,r in model_atac.iterrows():
+	fold = r[0]
+	name = r[1]
+	model_path = r[2]
+
+	input_peaks=os.path.join(model_path,"filtered.peaks.bed")
+	input_nonpeaks=os.path.join(model_path,"filtered.nonpeaks.bed")
+	test_nonpeaks="/oak/stanford/groups/akundaje/projects/chromatin-atlas-2022/ATAC/"+encode_id[name]+"/negatives_data/test/test."+fold+".filtered.negatives_with_summit.bed"
+	fold="/mnt/lab_data2/anusri/chrombpnet/splits/"+fold+".json"
+	output_path=os.path.join(model_path,"train_test_regions_may_7_2024/")
+
+	if not os.path.isfile(output_path+"nonpeaks.validationset.bed.gz"):
+		os.makedirs(output_path, exist_ok=True)
+		command=["python get_train_test_regions.py  -ip"]+[input_peaks]+["-inp"]+[input_nonpeaks]+["-inpt"]+[test_nonpeaks]+["-f"]+[fold]+["-o"]+[output_path]
+		command = " ".join(command)
+		print(command)
+		os.system(command)
+
diff --git a/upload_jsons/upload_scripts/run_conversion.py b/upload_jsons/upload_scripts/run_conversion.py
@@ -1,39 +1,46 @@
 import pandas as pd
 import os
 
-model_atac = pd.read_csv("/mnt/lab_data2/anusri/chrombpnet/logs/checkpoint/JAN_02_2023/model_dir_atac.csv",sep=",",header=None)
+#model_atac = pd.read_csv("/mnt/lab_data2/anusri/chrombpnet/logs/checkpoint/JAN_02_2023/model_dir_atac.csv",sep=",",header=None)
 #model_atac = pd.read_csv("/mnt/lab_data2/anusri/chrombpnet/logs/checkpoint/JAN_02_2023/model_dir_dnase.csv",sep=",",header=None)
+#model_atac = pd.read_csv("/mnt/lab_data2/anusri/chrombpnet/logs/checkpoint/JAN_02_2023/v1/model_dir_dnase_v2.1.csv",sep=",",header=None)
+#model_atac = pd.read_csv("/mnt/lab_data2/anusri/chrombpnet/logs/checkpoint/JAN_02_2023/v1/model_dir_subsample_atac.csv",sep=",",header=None)
+model_atac = pd.read_csv("/mnt/lab_data2/anusri/chrombpnet/upload_jsons/upload_scripts/model_dir_dnase_v2.1_bias.csv", sep=",", header=None)
+
+
 
 for i,r in model_atac.iterrows():
 	fold = r[0]	
 	name = r[1]
+	#model_path = r[3]
 	model_path = r[2]
+
 	input_path=os.path.join(model_path,"chrombpnet_model/chrombpnet.h5")
-	output_path=os.path.join(model_path,"chrombpnet_model/new_model_formats/chrombpnet")
+	output_path=os.path.join(model_path,"chrombpnet_model/new_model_formats_may_7_24/chrombpnet")
+	output_dir=os.path.join(model_path,"chrombpnet_model/new_model_formats_may_7_24/")
 	print(output_path)
 	if not os.path.isfile(output_path+".tar"):
 
-		os.makedirs(os.path.join(model_path,"chrombpnet_model/new_model_formats/"), exist_ok=True)
-		command = "CUDA_VISIBLE_DEVICES=5 python get_new_tf_model_format.py -i "+input_path+" -o "+output_path
+		os.makedirs(os.path.join(model_path,"chrombpnet_model/new_model_formats_may_7_24/"), exist_ok=True)
+		command = "CUDA_VISIBLE_DEVICES=1 python get_new_tf_model_format.py -i "+input_path+" -o "+output_dir+" -f chrombpnet"
 		print(command)	
 		os.system(command)
 
 	input_path=os.path.join(model_path,"chrombpnet_model/chrombpnet_wo_bias.h5")
-	output_path=os.path.join(model_path,"chrombpnet_model/new_model_formats/chrombpnet_wo_bias")
+	output_path=os.path.join(model_path,"chrombpnet_model/new_model_formats_may_7_24/chrombpnet_wo_bias")
 
 	if not os.path.isfile(output_path+".tar"):
-		os.makedirs(os.path.join(model_path,"chrombpnet_model/new_model_formats/"), exist_ok=True)
-		command = "CUDA_VISIBLE_DEVICES=5 python get_new_tf_model_format.py -i "+input_path+" -o "+output_path
+		os.makedirs(os.path.join(model_path,"chrombpnet_model/new_model_formats_may_7_24/"), exist_ok=True)
+		command = "CUDA_VISIBLE_DEVICES=1 python get_new_tf_model_format.py -i "+input_path+" -o "+output_dir+" -f chrombpnet_wo_bias"
 		print(command)	
 		os.system(command)
 
 	input_path=os.path.join(model_path,"chrombpnet_model/bias_model_scaled.h5")
-	output_path=os.path.join(model_path,"chrombpnet_model/new_model_formats/bias_model_scaled")
+	output_path=os.path.join(model_path,"chrombpnet_model/new_model_formats_may_7_24/bias_model_scaled")
 
 	if not os.path.isfile(output_path+".tar"):
-		os.makedirs(os.path.join(model_path,"chrombpnet_model/new_model_formats/"), exist_ok=True)
-		command = "CUDA_VISIBLE_DEVICES=5 python get_new_tf_model_format.py -i "+input_path+" -o "+output_path
+		os.makedirs(os.path.join(model_path,"chrombpnet_model/new_model_formats_may_7_24/"), exist_ok=True)
+		command = "CUDA_VISIBLE_DEVICES=1 python get_new_tf_model_format.py -i "+input_path+" -o "+output_dir+" -f bias_model_scaled"
 		print(command)	
 		os.system(command)
 
-
diff --git a/upload_jsons/upload_scripts/run_conversion_bias.py b/upload_jsons/upload_scripts/run_conversion_bias.py
@@ -1,28 +1,36 @@
 import pandas as pd
 import os
+import tensorflow as tf
+from keras.models import load_model
 
-#model_atac = pd.read_csv("/mnt/lab_data2/anusri/chrombpnet/logs/checkpoint/JAN_02_2023/model_dir_atac.csv",sep=",",header=None)
-model_atac = pd.read_csv("bias_models_atlas.csv",sep=",",header=None)
+model_atac = pd.read_csv("/mnt/lab_data2/anusri/chrombpnet/logs/checkpoint/JAN_02_2023/model_dir_atac.csv",sep=",",header=None)
+#model_atac = pd.read_csv("bias_models_atlas.csv",sep=",",header=None)
 #model_atac = pd.read_csv("/mnt/lab_data2/anusri/chrombpnet/logs/checkpoint/JAN_02_2023/model_dir_dnase.csv",sep=",",header=None)
+#model_atac = pd.read_csv("model_dir_dnase_v2.1_bias.csv",sep=",",header=None)
 
 
 for i,r in model_atac.iterrows():
 	fold = r[0]	
 	name = r[1]
 	model_path = r[2]
 	input_path=os.path.join(model_path,"bias_model/bias.h5")
-	output_path=os.path.join(model_path,"bias_model/new_model_formats_v1/bias")
-	output_dir=os.path.join(model_path,"bias_model/new_model_formats_v1/")
+	output_path=os.path.join(model_path,"bias_model/new_model_formats_vf/bias")
+	output_dir=os.path.join(model_path,"bias_model/new_model_formats_vf/")
 	file_path="bias"
 	if not os.path.isfile(input_path):
+		#if "IMR90_07.17.2022_bias_128_4_1234_0.3_fold_1_data_type_ATAC_PE" in input_path:
+		#	continue 
 		print("ERROR bias model not found")
 		print(input_path)
+		#ppath=os.path.join(model_path,"bias_model/new_model_formats/bias/")
+		#modelf = tf.keras.models.load_model(ppath)
+		#modelf.save(input_path)
 		continue
 
 	print(output_path)
 	if not os.path.isfile(output_path+".tar"):
 
-		os.makedirs(os.path.join(model_path,"bias_model/new_model_formats_v1/"), exist_ok=True)
+		os.makedirs(os.path.join(model_path,"bias_model/new_model_formats_vf/"), exist_ok=True)
 		command = "CUDA_VISIBLE_DEVICES=2 python get_new_tf_model_format.py -i "+input_path+" -o "+output_dir+" -f "+file_path
 		print(command)	
 		os.system(command)

diff --git a/upload_jsons/upload_scripts/run_conversion_new.py b/upload_jsons/upload_scripts/run_conversion_new.py
@@ -0,0 +1,50 @@
+import pandas as pd
+import os
+
+#model_atac = pd.read_csv("/mnt/lab_data2/anusri/chrombpnet/logs/checkpoint/JAN_02_2023/model_dir_atac.csv",sep=",",header=None)
+#model_atac = pd.read_csv("/mnt/lab_data2/anusri/chrombpnet/logs/checkpoint/JAN_02_2023/model_dir_dnase.csv",sep=",",header=None)
+
+#model_atac = pd.read_csv("/mnt/lab_data2/anusri/chrombpnet/logs/checkpoint/JAN_02_2023/v1/model_dir_dnase_v2.2.csv",sep=",",header=None)
+model_atac = pd.read_csv("/mnt/lab_data2/anusri/chrombpnet/upload_jsons/upload_scripts/model_dir_dnase_v2.1_bias.csv", sep=",", header=None)
+
+for i,r in model_atac.iterrows():
+	fold = r[0]	
+	name = r[1]
+	model_path = r[2]
+	#input_path=os.path.join(model_path,"new_model_formats_may_7_24/chrombpnet/")
+	#input_path=os.path.join(model_path,"chrombpnet/")
+	input_path=os.path.join(model_path,"chrombpnet_model/chrombpnet.h5")
+
+	output_path=os.path.join(model_path,"new_model_formats_may_7_24_vf/chrombpnet")
+	output_dir=os.path.join(model_path,"new_model_formats_may_7_24_vf/")
+	print(output_path)
+	if not os.path.isfile(output_path+".tar"):
+
+		os.makedirs(os.path.join(model_path,"new_model_formats_may_7_24_vf/"), exist_ok=True)
+		command = "CUDA_VISIBLE_DEVICES=1 python get_new_tf_model_format.py -i "+input_path+" -o "+output_dir+" -f chrombpnet"
+		print(command)	
+		os.system(command)
+
+	#input_path=os.path.join(model_path,"chrombpnet_wo_bias.h5")
+	input_path=os.path.join(model_path,"chrombpnet_model/chrombpnet_wo_bias.h5")
+
+	output_path=os.path.join(model_path,"new_model_formats_may_7_24_vf/chrombpnet_wo_bias")
+
+	if not os.path.isfile(output_path+".tar"):
+		os.makedirs(os.path.join(model_path,"new_model_formats_may_7_24_vf/"), exist_ok=True)
+		command = "CUDA_VISIBLE_DEVICES=1 python get_new_tf_model_format.py -i "+input_path+" -o "+output_dir+" -f chrombpnet_wo_bias"
+		print(command)	
+		os.system(command)
+
+	#input_path=os.path.join(model_path,"bias_model_scaled.h5")
+	input_path=os.path.join(model_path,"chrombpnet_model/bias_model_scaled.h5")
+
+	output_path=os.path.join(model_path,"new_model_formats_may_7_24_vf/bias_model_scaled")
+
+	if not os.path.isfile(output_path+".tar"):
+		os.makedirs(os.path.join(model_path,"new_model_formats_may_7_24_vf/"), exist_ok=True)
+		command = "CUDA_VISIBLE_DEVICES=1 python get_new_tf_model_format.py -i "+input_path+" -o "+output_dir+" -f bias_model_scaled"
+		print(command)	
+		os.system(command)
+
+