Spaces:
Running
Running
File size: 6,501 Bytes
df89a31 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 |
import pickle
from pathlib import Path
import numpy as np
import h5py
import faiss
import click
def getFlatIP():
test_index = faiss.IndexFlatIP(768)
return test_index
def getFlatL2():
test_index = faiss.IndexFlatL2(768)
return test_index
def getIVFFlat(all_keys, seen_test, unseen_test, seen_val, unseen_val):
quantizer = faiss.IndexFlatIP(768)
test_index = faiss.IndexIVFFlat(quantizer, 768, 128)
test_index.train(all_keys)
test_index.train(seen_test)
test_index.train(unseen_test)
test_index.train(seen_val)
test_index.train(unseen_val)
return test_index
def getHNSW():
# 16: connections for each vertex. efSearch: depth of search during search. efConstruction: depth of search during build
test_index = faiss.IndexHNSWFlat(768, 16)
test_index.hnsw.efSearch = 32
test_index.hnsw.efConstruction = 64
return test_index
def getLSH():
test_index = faiss.IndexLSH(768, 768 * 2)
return test_index
def getIdToEmbedding(allid, stid, utid, svalid, uvalid, all_keys, seen_test, unseen_test, seen_val, unseen_val):
id_to_emb_dict = dict()
i = 0
for id in allid:
id_to_emb_dict[id] = np.array([all_keys[i]])
i += 1
for id in stid:
id_to_emb_dict[id] = np.array([seen_test[i]])
i += 1
for id in utid:
id_to_emb_dict[id] = np.array([unseen_test[i]])
i += 1
for id in svalid:
id_to_emb_dict[id] = np.array([seen_val[i]])
i += 1
for id in uvalid:
id_to_emb_dict[id] = np.array([unseen_val[i]])
i += 1
return id_to_emb_dict
@click.command()
@click.option(
"--input",
type=click.Path(path_type=Path),
default="bioscan-clip-scripts/extracted_features",
help="Path to extracted features",
)
@click.option(
"--metadata", type=click.Path(path_type=Path), default="data/BIOSCAN_5M/BIOSCAN_5M.hdf5", help="Path to metadata"
)
@click.option(
"--output", type=click.Path(path_type=Path), default="bioscan-clip-scripts/index", help="Path to save the index"
)
def main(input, metadata, output):
# initialize data
all_keys = h5py.File(input / "extracted_features_of_all_keys.hdf5", "r", libver="latest")
all_keys_dna = all_keys["encoded_dna_feature"][:]
all_keys_im = all_keys["encoded_image_feature"][:]
seen_test = h5py.File(input / "extracted_features_of_seen_test.hdf5", "r", libver="latest")
seen_test_dna = seen_test["encoded_dna_feature"][:]
seen_test_im = seen_test["encoded_image_feature"][:]
unseen_test = h5py.File(input / "extracted_features_of_unseen_test.hdf5", "r", libver="latest")
unseen_test_dna = unseen_test["encoded_dna_feature"][:]
unseen_test_im = unseen_test["encoded_image_feature"][:]
seen_val = h5py.File(input / "extracted_features_of_seen_val.hdf5", "r", libver="latest")
seen_val_dna = seen_val["encoded_dna_feature"][:]
seen_val_im = seen_val["encoded_image_feature"][:]
unseen_val = h5py.File(input / "extracted_features_of_unseen_val.hdf5", "r", libver="latest")
unseen_val_dna = unseen_val["encoded_dna_feature"][:]
unseen_val_im = unseen_val["encoded_image_feature"][:]
dataset = h5py.File(metadata, "r", libver="latest")
id_field = "sampleid" # "processid"
allid = [item.decode("utf-8") for item in dataset["all_keys"][id_field][:]]
stid = [item.decode("utf-8") for item in dataset["test_seen"][id_field][:]]
utid = [item.decode("utf-8") for item in dataset["test_unseen"][id_field][:]]
svalid = [item.decode("utf-8") for item in dataset["val_seen"][id_field][:]]
uvalid = [item.decode("utf-8") for item in dataset["val_unseen"][id_field][:]]
all_keys = dataset["all_keys"]
seen_test = dataset["test_seen"]
unseen_test = dataset["test_unseen"]
seen_val = dataset["val_seen"]
unseen_val = dataset["val_unseen"]
# d = getIdToEmbedding(allid, stid, utid, svalid, uvalid, all_keys_dna, seen_test_dna, unseen_test_dna, seen_val_dna, unseen_val_dna)
# d = getIdToEmbedding(allid, stid, utid, svalid, uvalid, all_keys_im, seen_test_im, unseen_test_im, seen_val_im, unseen_val_im)
big_id_to_image_emb_dict = dict()
i = 0
for object in allid:
big_id_to_image_emb_dict[object] = np.array([all_keys_im[i]])
i += 1
i = 0
for object in stid:
big_id_to_image_emb_dict[object] = np.array([seen_test_im[i]])
i += 1
i = 0
for object in utid:
big_id_to_image_emb_dict[object] = np.array([unseen_test_im[i]])
i += 1
i = 0
for object in svalid:
big_id_to_image_emb_dict[object] = np.array([seen_val_im[i]])
i += 1
i = 0
for object in uvalid:
big_id_to_image_emb_dict[object] = np.array([unseen_val_im[i]])
i += 1
###
big_id_to_dna_emb_dict = dict()
i = 0
for object in allid:
big_id_to_dna_emb_dict[object] = np.array([all_keys_dna[i]])
i += 1
i = 0
for object in stid:
big_id_to_dna_emb_dict[object] = np.array([seen_test_dna[i]])
i += 1
i = 0
for object in utid:
big_id_to_dna_emb_dict[object] = np.array([unseen_test_dna[i]])
i += 1
i = 0
for object in svalid:
big_id_to_dna_emb_dict[object] = np.array([seen_val_dna[i]])
i += 1
i = 0
for object in uvalid:
big_id_to_dna_emb_dict[object] = np.array([unseen_val_dna[i]])
i += 1
###
processid_to_indx = dict()
big_indx_to_id_dict = dict()
i = 0
for object in allid:
big_indx_to_id_dict[i] = object
processid_to_indx[object] = i
i += 1
for object in stid:
big_indx_to_id_dict[i] = object
processid_to_indx[object] = i
i += 1
for object in utid:
big_indx_to_id_dict[i] = object
processid_to_indx[object] = i
i += 1
for object in svalid:
big_indx_to_id_dict[i] = object
processid_to_indx[object] = i
i += 1
for object in uvalid:
big_indx_to_id_dict[i] = object
processid_to_indx[object] = i
i += 1
###
with open(output / "big_id_to_image_emb_dict.pickle", "wb") as f:
pickle.dump(big_id_to_image_emb_dict, f)
with open(output / "big_id_to_dna_emb_dict.pickle", "wb") as f:
pickle.dump(big_id_to_dna_emb_dict, f)
with open(output / "big_indx_to_id_dict.pickle", "wb") as f:
pickle.dump(big_indx_to_id_dict, f)
if __name__ == "__main__":
main()
|