deepliif 1.2.1__py3-none-any.whl → 1.2.3__py3-none-any.whl

deepliif/models/__init__ - run_dask_multi dev.py

@@ -1,943 +0,0 @@
-"""This package contains modules related to objective functions, optimizations, and network architectures.
-
-To add a custom model class called 'dummy', you need to add a file called 'dummy_model.py' and define a subclass DummyModel inherited from BaseModel.
-You need to implement the following five functions:
-    -- <__init__>:                      initialize the class; first call BaseModel.__init__(self, opt).
-    -- <set_input>:                     unpack data from dataset and apply preprocessing.
-    -- <forward>:                       produce intermediate results.
-    -- <optimize_parameters>:           calculate loss, gradients, and update network weights.
-    -- <modify_commandline_options>:    (optionally) add model-specific options and set default options.
-
-In the function <__init__>, you need to define four lists:
-    -- self.loss_names (str list):          specify the training losses that you want to plot and save.
-    -- self.model_names (str list):         define networks used in our training.
-    -- self.visual_names (str list):        specify the images that you want to display and save.
-    -- self.optimizers (optimizer list):    define and initialize optimizers. You can define one optimizer for each network. If two networks are updated at the same time, you can use itertools.chain to group them. See CycleGAN_model.py for an usage.
-
-Now you can use the model class by specifying flag '--model dummy'.
-See our template model class 'template_model.py' for more details.
-"""
-import base64
-import os
-import itertools
-import importlib
-from functools import lru_cache
-from io import BytesIO
-import json
-import math
-
-import requests
-import torch
-from PIL import Image
-Image.MAX_IMAGE_PIXELS = None
-
-import numpy as np
-from dask import delayed, compute
-import openslide
-
-from deepliif.util import *
-from deepliif.util.util import tensor_to_pil
-from deepliif.data import transform
-from deepliif.postprocessing import compute_final_results, compute_cell_results
-from deepliif.postprocessing import encode_cell_data_v4, decode_cell_data_v4
-from deepliif.options import Options, print_options
-
-from .base_model import BaseModel
-
-# import for init purpose, not used in this script
-from .DeepLIIF_model import DeepLIIFModel
-from .DeepLIIFExt_model import DeepLIIFExtModel
-
-
-@lru_cache
-def get_opt(model_dir, mode='test'):
-    """
-    mode: test or train, currently only functions used for inference utilize get_opt so it
-          defaults to test
-    """
-    if mode == 'train':
-        opt = Options(path_file=os.path.join(model_dir,'train_opt.txt'), mode=mode)
-    elif mode == 'test':
-        try:
-            opt = Options(path_file=os.path.join(model_dir,'test_opt.txt'), mode=mode)
-        except:
-            opt = Options(path_file=os.path.join(model_dir,'train_opt.txt'), mode=mode)
-        opt.use_dp = False
-        opt.gpu_ids = list(range(torch.cuda.device_count()))
-    return opt
-
-
-def find_model_using_name(model_name):
-    """Import the module "models/[model_name]_model.py".
-
-    In the file, the class called DatasetNameModel() will
-    be instantiated. It has to be a subclass of BaseModel,
-    and it is case-insensitive.
-    """
-    model_filename = "deepliif.models." + model_name + "_model"
-    modellib = importlib.import_module(model_filename)
-    model = None
-    target_model_name = model_name.replace('_', '') + 'model'
-    for name, cls in modellib.__dict__.items():
-        if name.lower() == target_model_name.lower() \
-                and issubclass(cls, BaseModel):
-            model = cls
-
-    if model is None:
-        print("In %s.py, there should be a subclass of BaseModel with class name that matches %s in lowercase." % (
-            model_filename, target_model_name))
-        exit(0)
-
-    return model
-
-
-def get_option_setter(model_name):
-    """Return the static method <modify_commandline_options> of the model class."""
-    model_class = find_model_using_name(model_name)
-    return model_class.modify_commandline_options
-
-
-def create_model(opt):
-    """Create a model given the option.
-
-    This function warps the class CustomDatasetDataLoader.
-    This is the main interface between this package and 'train.py'/'test.py'
-
-    Example:
-        >>> from deepliif.models import create_model
-        >>> model = create_model(opt)
-    """
-    model = find_model_using_name(opt.model)
-    instance = model(opt)
-    print("model [%s] was created" % type(instance).__name__)
-    return instance
-
-
-def load_torchscript_model(model_pt_path, device):
-    net = torch.jit.load(model_pt_path, map_location=device)
-    net = disable_batchnorm_tracking_stats(net)
-    net.eval()
-    return net
-
-
-
-def load_eager_models(opt, devices=None):
-    # create a model given model and other options
-    model = create_model(opt)
-    # regular setup: load and print networks; create schedulers
-    model.setup(opt)
-
-    nets = {}
-    if devices:
-        model_names = list(devices.keys())
-    else:
-        model_names = model.model_names
-    
-    for name in model_names:#model.model_names:
-        if isinstance(name, str):
-            if '_' in name:
-                net = getattr(model, 'net' + name.split('_')[0])[int(name.split('_')[-1]) - 1]
-            else:
-                net = getattr(model, 'net' + name)
-    
-            if opt.phase != 'train':
-                net.eval()
-                net = disable_batchnorm_tracking_stats(net)
-            
-            # SDG models when loaded are still DP.. not sure why
-            if isinstance(net, torch.nn.DataParallel):
-                net = net.module
-
-            nets[name] = net
-            if devices:
-                nets[name].to(devices[name])
-            
-    return nets
-
-
-@lru_cache
-def init_nets(model_dir, eager_mode=False, opt=None, phase='test'):
-    """
-    Init DeepLIIF networks so that every net in
-    the same group is deployed on the same GPU
-    
-    opt_args: to overwrite opt arguments in train_opt.txt, typically used in inference stage
-              for example, opt_args={'phase':'test'}
-    """ 
-    if opt is None:
-        opt = get_opt(model_dir, mode=phase)
-    opt.use_dp = False
-    
-    if opt.model == 'DeepLIIF':
-        net_groups = [
-            ('G1', 'G52'),
-            ('G2', 'G53'),
-            ('G3', 'G54'),
-            ('G4', 'G55'),
-            ('G51',)
-        ]
-    elif opt.model in ['DeepLIIFExt','SDG']:
-        if opt.seg_gen:
-            net_groups = [(f'G_{i+1}',f'GS_{i+1}') for i in range(opt.modalities_no)]
-        else:
-            net_groups = [(f'G_{i+1}',) for i in range(opt.modalities_no)]
-    elif opt.model == 'CycleGAN':
-        if opt.BtoA:
-            net_groups = [(f'GB_{i+1}',) for i in range(opt.modalities_no)]
-        else:
-            net_groups = [(f'GA_{i+1}',) for i in range(opt.modalities_no)]
-    else:
-        raise Exception(f'init_nets() not implemented for model {opt.model}')
-
-    number_of_gpus_all = torch.cuda.device_count()
-    number_of_gpus = min(len(opt.gpu_ids),number_of_gpus_all)
-
-    if number_of_gpus > 0:
-        mapping_gpu_ids = {i:idx for i,idx in enumerate(opt.gpu_ids)}
-        chunks = [itertools.chain.from_iterable(c) for c in chunker(net_groups, number_of_gpus)]
-        # chunks = chunks[1:]
-        devices = {n: torch.device(f'cuda:{mapping_gpu_ids[i]}') for i, g in enumerate(chunks) for n in g}
-        # devices = {n: torch.device(f'cuda:{i}') for i, g in enumerate(chunks) for n in g}
-    else:
-        devices = {n: torch.device('cpu') for n in itertools.chain.from_iterable(net_groups)}
-
-    if eager_mode:
-        return load_eager_models(opt, devices)
-
-    return {
-        n: load_torchscript_model(os.path.join(model_dir, f'{n}.pt'), device=d)
-        for n, d in devices.items()
-    }
-
-
-def compute_overlap(img_size, tile_size):
-    w, h = img_size
-    if round(w / tile_size) == 1 and round(h / tile_size) == 1:
-        return 0
-
-    return tile_size // 4
-
-
-def run_torchserve(img, model_path=None, eager_mode=False, opt=None, seg_only=False):
-    """
-    eager_mode: not used in this function; put in place to be consistent with run_dask
-           so that run_wrapper() could call either this function or run_dask with
-           same syntax
-    opt: same as eager_mode
-    seg_only: same as eager_mode
-    """
-    buffer = BytesIO()
-    torch.save(transform(img.resize((opt.scale_size, opt.scale_size))), buffer)
-
-    torchserve_host = os.getenv('TORCHSERVE_HOST', 'http://localhost')
-    res = requests.post(
-        f'{torchserve_host}/wfpredict/deepliif',
-        json={'img': base64.b64encode(buffer.getvalue()).decode('utf-8')}
-    )
-
-    res.raise_for_status()
-
-    def deserialize_tensor(bs):
-        return torch.load(BytesIO(base64.b64decode(bs.encode())), map_location=torch.device('cpu'))
-
-    return {k: tensor_to_pil(deserialize_tensor(v)) for k, v in res.json().items()}
-
-
-def run_dask(img, model_path, eager_mode=False, opt=None, seg_only=False):
-    model_dir = os.getenv('DEEPLIIF_MODEL_DIR', model_path)
-    nets = init_nets(model_dir, eager_mode, opt)
-    use_dask = True if opt.norm != 'spectral' else False
-    
-    if opt.input_no > 1 or opt.model == 'SDG':
-        l_ts = [transform(img_i.resize((opt.scale_size,opt.scale_size))) for img_i in img]
-        ts = torch.cat(l_ts, dim=1)
-    else:
-        ts = transform(img.resize((opt.scale_size, opt.scale_size)))
-    
-
-    if use_dask:
-        @delayed
-        def forward(input, model):
-            with torch.no_grad():
-                return model(input.to(next(model.parameters()).device))
-    else: # some train settings like spectral norm some how in inference mode is not compatible with dask
-        def forward(input, model):
-            with torch.no_grad():
-                return model(input.to(next(model.parameters()).device))
-    
-    if opt.model == 'DeepLIIF':
-        weights = {
-            'G51': 0.25, # IHC
-            'G52': 0.25, # Hema
-            'G53': 0.25, # DAPI
-            'G54': 0.00, # Lap2
-            'G55': 0.25, # Marker
-        }
-
-        seg_map = {'G1': 'G52', 'G2': 'G53', 'G3': 'G54', 'G4': 'G55'}
-        if seg_only:
-            seg_map = {k: v for k, v in seg_map.items() if weights[v] != 0}
-        
-        lazy_gens = {k: forward(ts, nets[k]) for k in seg_map}
-        gens = compute(lazy_gens)[0]
-        
-        lazy_segs = {v: forward(gens[k], nets[v]).to(torch.device('cpu')) for k, v in seg_map.items()}
-        if weights['G51'] != 0:
-            lazy_segs['G51'] = forward(ts, nets['G51']).to(torch.device('cpu'))
-        segs = compute(lazy_segs)[0]
-    
-        seg = torch.stack([torch.mul(segs[k], weights[k]) for k in segs.keys()]).sum(dim=0)
-    
-        if seg_only:
-            res = {'G4': tensor_to_pil(gens['G4'])} if 'G4' in gens else {}
-        else:
-            res = {k: tensor_to_pil(v) for k, v in gens.items()}
-            res.update({k: tensor_to_pil(v) for k, v in segs.items()})
-        res['G5'] = tensor_to_pil(seg)
-    
-        return res
-    elif opt.model in ['DeepLIIFExt','SDG','CycleGAN']:
-        if opt.model == 'CycleGAN':
-            seg_map = {f'GB_{i+1}':None for i in range(opt.modalities_no)} if opt.BtoA else {f'GA_{i+1}':None for i in range(opt.modalities_no)}
-        else:
-            seg_map = {'G_' + str(i): 'GS_' + str(i) for i in range(1, opt.modalities_no + 1)}
-        
-        if use_dask:
-            lazy_gens = {k: forward(ts, nets[k]) for k in seg_map}
-            gens = compute(lazy_gens)[0]
-        else:
-            gens = {k: forward(ts, nets[k]) for k in seg_map}
-        
-        res = {k: tensor_to_pil(v) for k, v in gens.items()}
-    
-        if opt.seg_gen:
-            if use_dask:
-                lazy_segs = {v: forward(torch.cat([ts.to(torch.device('cpu')), gens[next(iter(seg_map))].to(torch.device('cpu')), gens[k].to(torch.device('cpu'))], 1), nets[v]).to(torch.device('cpu')) for k, v in seg_map.items()}
-                segs = compute(lazy_segs)[0]
-            else:
-                segs = {v: forward(torch.cat([ts.to(torch.device('cpu')), gens[next(iter(seg_map))].to(torch.device('cpu')), gens[k].to(torch.device('cpu'))], 1), nets[v]).to(torch.device('cpu')) for k, v in seg_map.items()}
-            res.update({k: tensor_to_pil(v) for k, v in segs.items()})
-    
-        return res
-    else:
-        raise Exception(f'run_dask() not fully implemented for {opt.model}')
-
-
-def run_dask_multi(imgs, model_path, eager_mode=False, opt=None, seg_only=False):
-    model_dir = os.getenv('DEEPLIIF_MODEL_DIR', model_path)
-    nets = init_nets(model_dir, eager_mode, opt)
-    use_dask = True if opt.norm != 'spectral' else False
-    
-    if opt.input_no > 1 or opt.model == 'SDG':
-        raise Exception(f'run_dask_multi() not fully implemented for {opt.model}')
-        #l_ts = [transform(img_i.resize((opt.scale_size,opt.scale_size))) for img_i in img]
-        #ts = torch.cat(l_ts, dim=1)
-    else:
-        tss = [transform(img.resize((opt.scale_size, opt.scale_size))) for img in imgs]
-    
-
-    if use_dask:
-        @delayed
-        def forward(input, model):
-            with torch.no_grad():
-                return model(input.to(next(model.parameters()).device))
-    else: # some train settings like spectral norm some how in inference mode is not compatible with dask
-        def forward(input, model):
-            with torch.no_grad():
-                return model(input.to(next(model.parameters()).device))
-    
-    if opt.model == 'DeepLIIF':
-        weights = {
-            'G51': 0.25, # IHC
-            'G52': 0.25, # Hema
-            'G53': 0.25, # DAPI
-            'G54': 0.00, # Lap2
-            'G55': 0.25, # Marker
-        }
-
-        seg_map = {'G1': 'G52', 'G2': 'G53', 'G3': 'G54', 'G4': 'G55'}
-        if seg_only:
-            seg_map = {k: v for k, v in seg_map.items() if weights[v] != 0}
-
-        lazy_gens = [{k: forward(ts, nets[k]) for k in seg_map} for ts in tss]
-        gens = compute(*lazy_gens)
-        
-        lazy_segs = [{v: forward(g[k], nets[v]).to(torch.device('cpu')) for k, v in seg_map.items()} for g in gens]
-        if weights['G51'] != 0:
-            for i in range(len(lazy_segs)):
-                lazy_segs[i]['G51'] = forward(tss[i], nets['G51']).to(torch.device('cpu'))
-        segs = compute(*lazy_segs)
-    
-        seg = [torch.stack([torch.mul(s[k], weights[k]) for k in s.keys()]).sum(dim=0) for s in segs]
-    
-        if seg_only:
-            res = [{'G4': tensor_to_pil(g['G4'])} if 'G4' in g else {} for g in gens]
-        else:
-            raise Exception(f'run_dask_multi() not fully implemented for {opt.model}')
-            #res = {k: tensor_to_pil(v) for k, v in gens.items()}
-            #res.update({k: tensor_to_pil(v) for k, v in segs.items()})
-        for i in range(len(seg)):
-            res[i]['G5'] = tensor_to_pil(seg[i])
-    
-        return res
-
-    else:
-        raise Exception(f'run_dask_multi() not fully implemented for {opt.model}')
-
-
-def run_dask_multi2(imgs, model_path, eager_mode=False, opt=None, seg_only=False):
-    model_dir = os.getenv('DEEPLIIF_MODEL_DIR', model_path)
-    nets = init_nets(model_dir, eager_mode, opt)
-    use_dask = True if opt.norm != 'spectral' else False
-    
-    if opt.input_no > 1 or opt.model == 'SDG':
-        raise Exception(f'run_dask_multi() not fully implemented for {opt.model}')
-        #l_ts = [transform(img_i.resize((opt.scale_size,opt.scale_size))) for img_i in img]
-        #ts = torch.cat(l_ts, dim=1)
-    else:
-        tss = [transform(img.resize((opt.scale_size, opt.scale_size))) for img in imgs]
-    
-
-    if use_dask:
-        @delayed
-        def forward(input, model):
-            with torch.no_grad():
-                return model(input.to(next(model.parameters()).device))
-    else: # some train settings like spectral norm some how in inference mode is not compatible with dask
-        def forward(input, model):
-            with torch.no_grad():
-                return model(input.to(next(model.parameters()).device))
-    
-    if opt.model == 'DeepLIIF':
-        weights = {
-            'G51': 0.25, # IHC
-            'G52': 0.25, # Hema
-            'G53': 0.25, # DAPI
-            'G54': 0.00, # Lap2
-            'G55': 0.25, # Marker
-        }
-
-        seg_map = {'G1': 'G52', 'G2': 'G53', 'G3': 'G54', 'G4': 'G55'}
-        if seg_only:
-            seg_map = {k: v for k, v in seg_map.items() if weights[v] != 0}
-
-        lazy_gens = {k: [forward(ts, nets[k]) for ts in tss] for k in seg_map}
-        gens = compute(lazy_gens)[0]
-        
-        lazy_segs = {v: [forward(g, nets[v]).to(torch.device('cpu')) for g in gens[k]] for k, v in seg_map.items()}
-        if weights['G51'] != 0:
-            lazy_segs['G51'] = [forward(ts, nets['G51']).to(torch.device('cpu')) for ts in tss]
-        segs = compute(lazy_segs)[0]
-    
-        seg = [torch.stack([torch.mul(segs[k][i], weights[k]) for k in segs.keys()]).sum(dim=0) for i in range(len(tss))]
-    
-        if seg_only:
-            res = [{'G4': tensor_to_pil(gens['G4'][i])} if 'G4' in gens else {} for i in range(len(tss))]
-        else:
-            raise Exception(f'run_dask_multi() not fully implemented for {opt.model}')
-            #res = {k: tensor_to_pil(v) for k, v in gens.items()}
-            #res.update({k: tensor_to_pil(v) for k, v in segs.items()})
-        for i in range(len(seg)):
-            res[i]['G5'] = tensor_to_pil(seg[i])
-    
-        return res
-
-    else:
-        raise Exception(f'run_dask_multi() not fully implemented for {opt.model}')
-
-
-def run_dask_multi3(imgs, model_path, eager_mode=False, opt=None, seg_only=False):
-    model_dir = os.getenv('DEEPLIIF_MODEL_DIR', model_path)
-    nets = init_nets(model_dir, eager_mode, opt)
-    #print(nets['G1'], flush=True)
-    #print(next(nets['G1'].parameters()), flush=True)
-    use_dask = True if opt.norm != 'spectral' else False
-
-    '''
-    with torch.no_grad():
-        ts = [transform(img.resize((512, 512))) for img in imgs]
-        #ts = [transform(imgs[0]), transform(imgs[0]), transform(imgs[0]), transform(imgs[0])]
-        for i in range(0):
-            for img in imgs:
-                ts.append(transform(img.resize((512, 512))))
-
-        num_iters = 100
-
-        tstart = time.time()
-        for i in range(num_iters):
-            tcat = torch.cat(ts)
-            rs = nets['G1'](tcat.to('cuda:0'))
-            rs = torch.split(rs, 1)
-        tend = time.time()
-        for i, r in enumerate(rs):
-            print('cat result size:', r.size(), flush=True)
-            im = tensor_to_pil(r)
-            im.save(f'test_wsi_G51_{i}_cat.png')
-        print('time:', tend-tstart, 'sec.', flush=True)
-
-        tstart = time.time()
-        for i in range(num_iters):
-            rs = []
-            for t in ts:
-                rs.append(nets['G1'](t.to('cuda:0')))
-        tend = time.time()
-        for i, r in enumerate(rs):
-            print('ind result size:', r.size(), flush=True)
-            im = tensor_to_pil(r)
-            im.save(f'test_wsi_G51_{i}_ind.png')
-        print('time:', tend-tstart, 'sec.', flush=True)
-    '''    
-
-    if opt.input_no > 1 or opt.model == 'SDG':
-        raise Exception(f'run_dask_multi() not fully implemented for {opt.model}')
-        #l_ts = [transform(img_i.resize((opt.scale_size,opt.scale_size))) for img_i in img]
-        #ts = torch.cat(l_ts, dim=1)
-    else:
-        #ts = transform(img.resize((opt.scale_size, opt.scale_size)))
-        ts = torch.cat([transform(img.resize((opt.scale_size, opt.scale_size))) for img in imgs])
-    
-
-    if use_dask:
-        @delayed
-        def forward(input, model):
-            with torch.no_grad():
-                return model(input.to(next(model.parameters()).device))
-    else: # some train settings like spectral norm some how in inference mode is not compatible with dask
-        def forward(input, model):
-            with torch.no_grad():
-                return model(input.to(next(model.parameters()).device))
-    
-    if opt.model == 'DeepLIIF':
-        weights = {
-            'G51': 0.25, # IHC
-            'G52': 0.25, # Hema
-            'G53': 0.25, # DAPI
-            'G54': 0.00, # Lap2
-            'G55': 0.25, # Marker
-        }
-
-        seg_map = {'G1': 'G52', 'G2': 'G53', 'G3': 'G54', 'G4': 'G55'}
-        if seg_only:
-            seg_map = {k: v for k, v in seg_map.items() if weights[v] != 0}
-        
-        lazy_gens = {k: forward(ts, nets[k]) for k in seg_map}
-        gens = compute(lazy_gens)[0]
-        
-        lazy_segs = {v: forward(gens[k], nets[v]).to(torch.device('cpu')) for k, v in seg_map.items()}
-        if weights['G51'] != 0:
-            lazy_segs['G51'] = forward(ts, nets['G51']).to(torch.device('cpu'))
-        segs = compute(lazy_segs)[0]
-    
-        seg = torch.stack([torch.mul(segs[k], weights[k]) for k in segs.keys()]).sum(dim=0)
-    
-        if seg_only:
-            #res = {'G4': tensor_to_pil(gens['G4'])} if 'G4' in gens else {}
-            g4 = torch.split(gens['G4'], 1) # SHOULD DO CHECK IF G4 IN gens
-            #print(type(g4), g4[0].size(), flush=True)
-            res = [{'G4': tensor_to_pil(g)} for g in g4] # SHOULD DO CHECK IF G4 IN gens
-        else:
-            raise Exception(f'run_dask_multi() not fully implemented for {opt.model}')
-            #res = {k: tensor_to_pil(v) for k, v in gens.items()}
-            #res.update({k: tensor_to_pil(v) for k, v in segs.items()})
-        #res['G5'] = tensor_to_pil(seg)
-        g5 = torch.split(seg, 1)
-        for i in range(len(g5)):
-            res[i]['G5'] = tensor_to_pil(g5[i])
-    
-        return res
-
-    else:
-        raise Exception(f'run_dask_multi3() not fully implemented for {opt.model}')
-      
-
-def is_empty(tile):
-    thresh = 15
-    if isinstance(tile, list): # for pair of tiles, only mark it as empty / no need for prediction if ALL tiles are empty
-        return all([True if np.max(image_variance_rgb(t)) < thresh else False for t in tile])
-    else:
-        return True if np.max(image_variance_rgb(tile)) < thresh else False
-
-
-def run_wrapper(tile, run_fn, model_path, eager_mode=False, opt=None, seg_only=False):
-    if opt.model == 'DeepLIIF':
-        if is_empty(tile):
-            return {
-                'G1': Image.new(mode='RGB', size=(512, 512), color=(201, 211, 208)),
-                'G2': Image.new(mode='RGB', size=(512, 512), color=(10, 10, 10)),
-                'G3': Image.new(mode='RGB', size=(512, 512), color=(0, 0, 0)),
-                'G4': Image.new(mode='RGB', size=(512, 512), color=(10, 10, 10)),
-                'G5': Image.new(mode='RGB', size=(512, 512), color=(0, 0, 0)),
-                'G51': Image.new(mode='RGB', size=(512, 512), color=(0, 0, 0)),
-                'G52': Image.new(mode='RGB', size=(512, 512), color=(0, 0, 0)),
-                'G53': Image.new(mode='RGB', size=(512, 512), color=(0, 0, 0)),
-                'G54': Image.new(mode='RGB', size=(512, 512), color=(0, 0, 0)),
-                'G55': Image.new(mode='RGB', size=(512, 512), color=(0, 0, 0)),
-            }
-        else:
-            return run_fn(tile, model_path, eager_mode, opt, seg_only)
-    elif opt.model in ['DeepLIIFExt', 'SDG']:
-        if is_empty(tile):
-            res = {'G_' + str(i): Image.new(mode='RGB', size=(512, 512)) for i in range(1, opt.modalities_no + 1)}
-            res.update({'GS_' + str(i): Image.new(mode='RGB', size=(512, 512)) for i in range(1, opt.modalities_no + 1)})
-            return res
-        else:
-            return run_fn(tile, model_path, eager_mode, opt)
-    elif opt.model in ['CycleGAN']:
-        if is_empty(tile):
-            net_names = ['GB_{i+1}' for i in range(opt.modalities_no)] if opt.BtoA else [f'GA_{i+1}' for i in range(opt.modalities_no)]
-            res = {net_name: Image.new(mode='RGB', size=(512, 512)) for net_name in net_names}
-            return res
-        else:
-            return run_fn(tile, model_path, eager_mode, opt)
-    else:
-        raise Exception(f'run_wrapper() not implemented for model {opt.model}')
-
-
-def inference(img, tile_size, overlap_size, model_path, use_torchserve=False,
-              eager_mode=False, color_dapi=False, color_marker=False, opt=None,
-              return_seg_intermediate=False, seg_only=False):
-    if not opt:
-        opt = get_opt(model_path)
-        #print_options(opt)
-
-    run_fn = run_torchserve if use_torchserve else run_dask
-
-    if opt.model == 'SDG':
-        # SDG could have multiple input images/modalities, hence the input could be a rectangle.
-        # We split the input to get each modality image then create tiles for each set of input images.
-        w, h = int(img.width / opt.input_no), img.height
-        orig = [img.crop((w * i, 0, w * (i+1), h)) for i in range(opt.input_no)]
-    else:
-        # Otherwise expect a single input image, which is used directly.
-        orig = img
-
-    tiler = InferenceTiler(orig, tile_size, overlap_size)
-    for tile in tiler:
-        tiler.stitch(run_wrapper(tile, run_fn, model_path, eager_mode, opt, seg_only))
-    results = tiler.results()
-
-    if opt.model == 'DeepLIIF':
-        if seg_only:
-            images = {'Seg': results['G5']}
-            if 'G4' in results:
-                images.update({'Marker': results['G4']})
-        else:
-            images = {
-                'Hema': results['G1'],
-                'DAPI': results['G2'],
-                'Lap2': results['G3'],
-                'Marker': results['G4'],
-                'Seg': results['G5'],
-            }
-        
-        if return_seg_intermediate and not seg_only:
-            images.update({'IHC_s':results['G51'],
-                          'Hema_s':results['G52'],
-                          'DAPI_s':results['G53'],
-                          'Lap2_s':results['G54'],
-                          'Marker_s':results['G55'],})
-        
-        if color_dapi and not seg_only:
-            matrix = (       0,        0,        0, 0,
-                      299/1000, 587/1000, 114/1000, 0,
-                      299/1000, 587/1000, 114/1000, 0)
-            images['DAPI'] = images['DAPI'].convert('RGB', matrix)
-        if color_marker and not seg_only:
-            matrix = (299/1000, 587/1000, 114/1000, 0,
-                      299/1000, 587/1000, 114/1000, 0,
-                             0,        0,        0, 0)
-            images['Marker'] = images['Marker'].convert('RGB', matrix)
-        return images
-
-    elif opt.model == 'DeepLIIFExt':
-        images = {f'mod{i}': results[f'G_{i}'] for i in range(1, opt.modalities_no + 1)}
-        if opt.seg_gen:
-            images.update({f'Seg{i}': results[f'GS_{i}'] for i in range(1, opt.modalities_no + 1)})
-        return images
-
-    elif opt.model == 'SDG':
-        images = {f'mod{i}': results[f'G_{i}'] for i in range(1, opt.modalities_no + 1)}
-        return images
-
-    else:
-        #raise Exception(f'inference() not implemented for model {opt.model}')
-        return results # return result images with default key names (i.e., net names)
-
-
-def postprocess(orig, images, tile_size, model, seg_thresh=150, size_thresh='default', marker_thresh=None, size_thresh_upper=None):
-    if model == 'DeepLIIF':
-        resolution = '40x' if tile_size > 384 else ('20x' if tile_size > 192 else '10x')
-        overlay, refined, scoring = compute_final_results(
-            orig, images['Seg'], images.get('Marker'), resolution,
-            size_thresh, marker_thresh, size_thresh_upper, seg_thresh)
-        processed_images = {}
-        processed_images['SegOverlaid'] = Image.fromarray(overlay)
-        processed_images['SegRefined'] = Image.fromarray(refined)
-        return processed_images, scoring
-
-    elif model in ['DeepLIIFExt','SDG']:
-        resolution = '40x' if tile_size > 768 else ('20x' if tile_size > 384 else '10x')
-        processed_images = {}
-        scoring = {}
-        for img_name in list(images.keys()):
-            if 'Seg' in img_name:
-                seg_img = images[img_name]
-                overlay, refined, score = compute_final_results(
-                    orig, images[img_name], None, resolution,
-                    size_thresh, marker_thresh, size_thresh_upper, seg_thresh)
-    
-                processed_images[img_name + '_Overlaid'] = Image.fromarray(overlay)
-                processed_images[img_name + '_Refined'] = Image.fromarray(refined)
-                scoring[img_name] = score
-        return processed_images, scoring
-
-    else:
-        raise Exception(f'postprocess() not implemented for model {model}')
-
-
-def infer_modalities(img, tile_size, model_dir, eager_mode=False,
-                     color_dapi=False, color_marker=False, opt=None,
-                     return_seg_intermediate=False):
-    """
-    This function is used to infer modalities for the given image using a trained model.
-    :param img: The input image.
-    :param tile_size: The tile size.
-    :param model_dir: The directory containing serialized model files.
-    :return: The inferred modalities and the segmentation mask.
-    """
-    if opt is None:
-        opt = get_opt(model_dir)
-        opt.use_dp = False
-        #print_options(opt)
-    
-    # for those with multiple input modalities, find the correct size to calculate overlap_size
-    input_no = opt.input_no if hasattr(opt, 'input_no') else 1
-    img_size = (img.size[0] / input_no, img.size[1]) # (width, height)
-
-    images = inference(
-        img,
-        tile_size=tile_size,
-        #overlap_size=compute_overlap(img_size, tile_size),
-        overlap_size=tile_size//16,
-        model_path=model_dir,
-        eager_mode=eager_mode,
-        color_dapi=color_dapi,
-        color_marker=color_marker,
-        opt=opt,
-        return_seg_intermediate=return_seg_intermediate
-    )
-    
-    if not hasattr(opt,'seg_gen') or (hasattr(opt,'seg_gen') and opt.seg_gen): # the first condition accounts for old settings of deepliif; the second refers to deepliifext models
-        post_images, scoring = postprocess(img, images, tile_size, opt.model)
-        images = {**images, **post_images}
-        return images, scoring
-    else:
-        return images, None
-
-
-def infer_results_for_wsi(input_dir, filename, output_dir, model_dir, tile_size, region_size=20000):
-    """
-    This function infers modalities and segmentation mask for the given WSI image. It
-
-    :param input_dir: The directory containing the WSI.
-    :param filename: The WSI name.
-    :param output_dir: The directory for saving the inferred modalities.
-    :param model_dir: The directory containing the serialized model files.
-    :param tile_size: The tile size.
-    :param region_size: The size of each individual region to be processed at once.
-    :return:
-    """
-    basename, _ = os.path.splitext(filename)
-    results_dir = os.path.join(output_dir, basename)
-    if not os.path.exists(results_dir):
-        os.makedirs(results_dir)
-    size_x, size_y, size_z, size_c, size_t, pixel_type = get_information(os.path.join(input_dir, filename))
-    rescale = (pixel_type != 'uint8')
-    print(filename, size_x, size_y, size_z, size_c, size_t, pixel_type)
-
-    results = {}
-    scoring = None
-
-    # javabridge already set up from previous call to get_information()
-    with bioformats.ImageReader(os.path.join(input_dir, filename)) as reader:
-        start_x, start_y = 0, 0
-
-        while start_x < size_x:
-            while start_y < size_y:
-                print(start_x, start_y)
-                region_XYWH = (start_x, start_y, min(region_size, size_x - start_x), min(region_size, size_y - start_y))
-                region = reader.read(XYWH=region_XYWH, rescale=rescale)
-                img = Image.fromarray((region * 255).astype(np.uint8)) if rescale else Image.fromarray(region)
-
-                region_modalities, region_scoring = infer_modalities(img, tile_size, model_dir)
-                if region_scoring is not None:
-                    if scoring is None:
-                        scoring = {
-                            'num_pos': region_scoring['num_pos'],
-                            'num_neg': region_scoring['num_neg'],
-                        }
-                    else:
-                        scoring['num_pos'] += region_scoring['num_pos']
-                        scoring['num_neg'] += region_scoring['num_neg']
-
-                for name, img in region_modalities.items():
-                    if name not in results:
-                        results[name] = np.zeros((size_y, size_x, 3), dtype=np.uint8)
-                    results[name][region_XYWH[1]: region_XYWH[1] + region_XYWH[3],
-                    region_XYWH[0]: region_XYWH[0] + region_XYWH[2]] = np.array(img)
-                start_y += region_size
-            start_y = 0
-            start_x += region_size
-
-    # write_results_to_pickle_file(os.path.join(results_dir, "results.pickle"), results)
-    # read_results_from_pickle_file(os.path.join(results_dir, "results.pickle"))
-
-    for name, img in results.items():
-        write_big_tiff_file(os.path.join(results_dir, f'{basename}_{name}.ome.tiff'), img, tile_size)
-
-    if scoring is not None:
-        scoring['num_total'] = scoring['num_pos'] + scoring['num_neg']
-        scoring['percent_pos'] = round(scoring['num_pos'] / scoring['num_total'] * 100, 1) if scoring['num_pos'] > 0 else 0
-        with open(os.path.join(results_dir, f'{basename}.json'), 'w') as f:
-            json.dump(scoring, f, indent=2)
-
-    javabridge.kill_vm()
-
-
-def get_wsi_resolution(filename):
-    """
-    Use OpenSlide to get the resolution (magnification) of the slide
-    and the corresponding tile size to use by default for DeepLIIF.
-    If it cannot be found, return (None, None) instead.
-
-    Parameters
-    ----------
-    filename : str
-        Full path to the file.
-
-    Returns
-    -------
-    str :
-        Magnification (objective power) as found by OpenSlide.
-    int :
-        Corresponding tile size for DeepLIIF.
-    """
-    try:
-        image = openslide.OpenSlide(filename)
-        mag = image.properties.get(openslide.PROPERTY_NAME_OBJECTIVE_POWER)
-        tile_size = round((float(mag) / 40) * 512)
-        return mag, tile_size
-    except Exception as e:
-        return None, None
-
-
-def infer_cells_for_wsi(filename, model_dir, tile_size, region_size=20000, version=3, print_log=False):
-    """
-    Perform inference on a slide and get the results individual cell data.
-
-    Parameters
-    ----------
-    filename : str
-        Full path to the file.
-    model_dir : str
-        Full path to the directory with the DeepLIIF model files.
-    tile_size : int
-        Size of tiles to extract and perform inference on.
-    region_size : int
-        Maximum size to split the slide for processing.
-    version : int
-        Version of cell data to return (3 or 4).
-    print_log : bool
-        Whether or not to print updates while processing.
-
-    Returns
-    -------
-    dict :
-        Individual cell data and associated values.
-    """
-
-    def print_info(*args):
-        if print_log:
-            print(*args, flush=True)
-
-    resolution = '40x' if tile_size > 384 else ('20x' if tile_size > 192 else '10x')
-
-    size_x, size_y, size_z, size_c, size_t, pixel_type = get_information(filename)
-    rescale = (pixel_type != 'uint8')
-    print_info('Info:', size_x, size_y, size_z, size_c, size_t, pixel_type)
-
-    num_regions_x = math.ceil(size_x / region_size)
-    num_regions_y = math.ceil(size_y / region_size)
-    stride_x = math.ceil(size_x / num_regions_x)
-    stride_y = math.ceil(size_y / num_regions_y)
-    print_info('Strides:', stride_x, stride_y)
-
-    data = None
-    default_marker_thresh, count_marker_thresh = 0, 0
-    default_size_thresh, count_size_thresh = 0, 0
-
-    # javabridge already set up from previous call to get_information()
-    with bioformats.ImageReader(filename) as reader:
-        start_x, start_y = 0, 0
-
-        while start_y < size_y:
-            while start_x < size_x:
-                region_XYWH = (start_x, start_y, min(stride_x, size_x-start_x), min(stride_y, size_y-start_y))
-                print_info('Region:', region_XYWH)
-
-                region = reader.read(XYWH=region_XYWH, rescale=rescale)
-                print_info(region.shape, region.dtype)
-                img = Image.fromarray((region * 255).astype(np.uint8)) if rescale else Image.fromarray(region)
-                print_info(img.size, img.mode)
-
-                images = inference(
-                    img,
-                    tile_size=tile_size,
-                    overlap_size=tile_size//16,
-                    model_path=model_dir,
-                    eager_mode=False,
-                    color_dapi=False,
-                    color_marker=False,
-                    opt=None,
-                    return_seg_intermediate=False,
-                    seg_only=True,
-                )
-                region_data = compute_cell_results(images['Seg'], images.get('Marker'), resolution, version=version)
-
-                if start_x != 0 or start_y != 0:
-                    for i in range(len(region_data['cells'])):
-                        cell = decode_cell_data_v4(region_data['cells'][i]) if version == 4 else region_data['cells'][i]
-                        for j in range(2):
-                            cell['bbox'][j] = (cell['bbox'][j][0] + start_x, cell['bbox'][j][1] + start_y)
-                        cell['centroid'] = (cell['centroid'][0] + start_x, cell['centroid'][1] + start_y)
-                        for j in range(len(cell['boundary'])):
-                            cell['boundary'][j] = (cell['boundary'][j][0] + start_x, cell['boundary'][j][1] + start_y)
-                        region_data['cells'][i] = encode_cell_data_v4(cell) if version == 4 else cell
-
-                if data is None:
-                    data = region_data
-                else:
-                    data['cells'] += region_data['cells']
-
-                if region_data['settings']['default_marker_thresh'] is not None and region_data['settings']['default_marker_thresh'] != 0:
-                    default_marker_thresh += region_data['settings']['default_marker_thresh']
-                    count_marker_thresh += 1
-                if region_data['settings']['default_size_thresh'] != 0:
-                    default_size_thresh += region_data['settings']['default_size_thresh']
-                    count_size_thresh += 1
-
-                start_x += stride_x
-
-            start_x = 0
-            start_y += stride_y
-
-    javabridge.kill_vm()
-
-    if count_marker_thresh == 0:
-        count_marker_thresh = 1
-    if count_size_thresh == 0:
-        count_size_thresh = 1
-    data['settings']['default_marker_thresh'] = round(default_marker_thresh / count_marker_thresh)
-    data['settings']['default_size_thresh'] = round(default_size_thresh / count_size_thresh)
-
-    return data