File size: 6,354 Bytes
5d2263b 5cbd5ac 5d2263b 5cbd5ac |
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 |
import argparse
import torch
import os
os.chdir('..')
from dataloader import CellLoader
from matplotlib import pyplot as plt
from celle_main import instantiate_from_config
from omegaconf import OmegaConf
from celle.utils import process_image
def run_model(mode, sequence,
nucleus_image_path,
protein_image_path,
model_ckpt_path,
model_config_path,
device):
if mode == "image":
run_image_prediction(
sequence,
nucleus_image_path,
protein_image_path,
model_ckpt_path,
model_config_path,
device
)
elif mode == "sequence":
run_sequence_prediction(
sequence,
nucleus_image_path,
protein_image_path,
model_ckpt_path,
model_config_path,
device
)
def run_sequence_prediction(
sequence_input,
nucleus_image_path,
protein_image_path,
model_ckpt_path,
model_config_path,
device
):
"""
Run Celle model with provided inputs and display results.
:param sequence: Path to sequence file
:param nucleus_image_path: Path to nucleus image
:param protein_image_path: Path to protein image (optional)
:param model_ckpt_path: Path to model checkpoint
:param model_config_path: Path to model config
"""
# Instantiate dataset object
dataset = CellLoader(
sequence_mode="embedding",
vocab="esm2",
split_key="val",
crop_method="center",
resize=600,
crop_size=256,
text_seq_len=1000,
pad_mode="end",
threshold="median",
)
# Check if sequence is provided and valid
if len(sequence_input) == 0:
raise ValueError("Sequence must be provided.")
if "<mask>" not in sequence_input:
print("Warning: Sequence does not contain any masked positions to predict.")
# Convert SEQUENCE to sequence using dataset.tokenize_sequence()
sequence = dataset.tokenize_sequence(sequence_input)
# Check if nucleus image path is provided and valid
if not os.path.exists(nucleus_image_path):
# Use default nucleus image from dataset and print warning
nucleus_image_path = 'images/nucleus.jpg'
print(
"Warning: No nucleus image provided. Using default nucleus image from dataset."
)
else:
# Load nucleus image from provided path
nucleus_image = process_image(nucleus_image_path)
# Check if protein image path is provided and valid
if not os.path.exists(protein_image_path):
# Use default nucleus image from dataset and print warning
protein_image_path = 'images/protein.jpg'
print(
"Warning: No nucleus image provided. Using default protein image from dataset."
)
else:
# Load protein image from provided path
protein_image = process_image(protein_image_path)
protein_image = (protein_image > torch.median(protein_image,dim=0))*1.0
# Load model config and set ckpt_path if not provided in config
config = OmegaConf.load(model_config_path)
if config["model"]["params"]["ckpt_path"] is None:
config["model"]["params"]["ckpt_path"] = model_ckpt_path
# Set condition_model_path and vqgan_model_path to None
config["model"]["params"]["condition_model_path"] = None
config["model"]["params"]["vqgan_model_path"] = None
# Instantiate model from config and move to device
model = instantiate_from_config(config).to(device)
# Sample from model using provided sequence and nucleus image
_, predicted_sequence, _ = model.celle.sample_text(
text=sequence,
condition=nucleus_image,
image=protein_image,
force_aas=True,
timesteps=1,
temperature=1,
progress=True,
)
formatted_predicted_sequence = ""
for i in range(min(len(predicted_sequence), len(sequence))):
if predicted_sequence[i] != sequence[i]:
formatted_predicted_sequence += f"**{predicted_sequence[i]}**"
else:
formatted_predicted_sequence += predicted_sequence[i]
if len(predicted_sequence) > len(sequence):
formatted_predicted_sequence += f"**{predicted_sequence[len(sequence):]}**"
print("predicted_sequence:", formatted_predicted_sequence)
def run_image_prediction(
sequence_input,
nucleus_image,
model_ckpt_path,
model_config_path,
device
):
"""
Run Celle model with provided inputs and display results.
:param sequence: Path to sequence file
:param nucleus_image_path: Path to nucleus image
:param protein_image_path: Path to protein image (optional)
:param model_ckpt_path: Path to model checkpoint
:param model_config_path: Path to model config
"""
# Instantiate dataset object
dataset = CellLoader(
sequence_mode="embedding",
vocab="esm2",
split_key="val",
crop_method="center",
resize=600,
crop_size=256,
text_seq_len=1000,
pad_mode="end",
threshold="median",
)
# Convert SEQUENCE to sequence using dataset.tokenize_sequence()
sequence = dataset.tokenize_sequence(sequence_input)
# Load model config and set ckpt_path if not provided in config
config = OmegaConf.load(model_config_path)
if config["model"]["params"]["ckpt_path"] is None:
config["model"]["params"]["ckpt_path"] = model_ckpt_path
# Set condition_model_path and vqgan_model_path to None
config["model"]["params"]["condition_model_path"] = None
config["model"]["params"]["vqgan_model_path"] = None
# Instantiate model from config and move to device
model = instantiate_from_config(config).to(device)
# Sample from model using provided sequence and nucleus image
_, _, _, predicted_threshold, predicted_heatmap = model.celle.sample(
text=sequence,
condition=nucleus_image,
timesteps=1,
temperature=1,
progress=True,
)
# Move predicted_threshold and predicted_heatmap to CPU and select first element of batch
predicted_threshold = predicted_threshold.cpu()[0, 0]
predicted_heatmap = predicted_heatmap.cpu()[0, 0]
return predicted_threshold, predicted_heatmap |