some functions maybe mixed between 1.7.0 vs 1.7.2rc1

[hufanglq]

When update to 1.7.2rc1，code in tutorial can't run normally, such as plot_weights in _mofa.py file. The code has big difference.
v1.7.0...e08f2cb

[Starlitnightly]

sry, we will provided a new tutorial for mofa in new version

[Starlitnightly]

It seems that we need to some time to create a new tutorial, but now you can use the function to analysis and visualize directly.

t_mofa_bulk.md

[hufanglq]

It's pretty cool! There is a little question. Using same data, the muon.tl.mofa func raise OOM for cuda, while mofa_obj.mofa_run func done well.
mofa_obj_mofa_run:

mofa_obj.mofa_preprocess()
mofa_obj.mofa_run(outfile='models/chen_rna_atac.hdf5', gpu_device=0)

        #########################################################
        ###           __  __  ____  ______                    ### 
        ###          |  \/  |/ __ \|  ____/\    _             ### 
        ###          | \  / | |  | | |__ /  \ _| |_           ### 
        ###          | |\/| | |  | |  __/ /\ \_   _|          ###
        ###          | |  | | |__| | | / ____ \|_|            ###
        ###          |_|  |_|\____/|_|/_/    \_\              ###
        ###                                                   ### 
        ######################################################### 
       
 
        
Groups names not provided, using default naming convention:
- group1, group2, ..., groupG

Successfully loaded view='RNA' group='group0' with N=5802 samples and D=2000 features...
Successfully loaded view='ATAC' group='group0' with N=5802 samples and D=25488 features...


Warning: 15 features(s) in view 0 have zero variance, consider removing them before training the model...

Warning: 204 features(s) in view 1 have zero variance, consider removing them before training the model...

Model options:
- Automatic Relevance Determination prior on the factors: True
- Automatic Relevance Determination prior on the weights: True
- Spike-and-slab prior on the factors: False
- Spike-and-slab prior on the weights: True
Likelihoods:
- View 0 (RNA): gaussian
- View 1 (ATAC): gaussian



GPU mode is activated


Using GPU device 0



######################################
## Training the model with seed 112 ##
######################################


ELBO before training: -2115272814.25 

Iteration 1: time=47.26, ELBO=191213094.17, deltaELBO=2306485908.422 (109.03964221%), Factors=19
Iteration 2: time=41.02, ELBO=196818457.51, deltaELBO=5605363.336 (0.26499482%), Factors=18
Iteration 3: time=38.21, ELBO=196891451.69, deltaELBO=72994.187 (0.00345082%), Factors=17
Iteration 4: time=36.30, ELBO=196995605.43, deltaELBO=104153.733 (0.00492389%), Factors=16
Iteration 5: time=34.33, ELBO=197054469.84, deltaELBO=58864.412 (0.00278283%), Factors=15
Iteration 6: time=32.41, ELBO=197108040.24, deltaELBO=53570.400 (0.00253255%), Factors=14
Iteration 7: time=30.56, ELBO=197166806.53, deltaELBO=58766.287 (0.00277819%), Factors=13
Iteration 8: time=28.83, ELBO=197221227.29, deltaELBO=54420.761 (0.00257275%), Factors=12
Iteration 9: time=27.12, ELBO=197274435.32, deltaELBO=53208.035 (0.00251542%), Factors=11
Iteration 10: time=25.16, ELBO=197321461.12, deltaELBO=47025.800 (0.00222316%), Factors=10
Iteration 11: time=23.15, ELBO=197307721.09, deltaELBO=-13740.036 (0.00064956%), Factors=9
Warning, lower bound is decreasing...�
Iteration 12: time=21.24, ELBO=197345997.01, deltaELBO=38275.920 (0.00180950%), Factors=8
Iteration 13: time=19.19, ELBO=197354473.50, deltaELBO=8476.491 (0.00040073%), Factors=8
Iteration 14: time=19.12, ELBO=197361554.07, deltaELBO=7080.568 (0.00033474%), Factors=8

Converged!



#######################
## Training finished ##
#######################


Warning: Output file models/chen_rna_atac.hdf5 already exists, it will be replaced
Saving model in models/chen_rna_atac.hdf5...

muon.tl.mofa:

muon.tl.mofa(mdata, use_obs='union',gpu_mode=True, gpu_device=0,
           n_factors=20, convergence_mode='fast', seed=112, 
           outfile="models/chen_rna_atac.hdf5", verbose=True)

        #########################################################
        ###           __  __  ____  ______                    ### 
        ###          |  \/  |/ __ \|  ____/\    _             ### 
        ###          | \  / | |  | | |__ /  \ _| |_           ### 
        ###          | |\/| | |  | |  __/ /\ \_   _|          ###
        ###          | |  | | |__| | | / ____ \|_|            ###
        ###          |_|  |_|\____/|_|/_/    \_\              ###
        ###                                                   ### 
        ######################################################### 
       
 
        
Loaded view='RNA' group='group1' with N=5802 samples and D=2000 features...
Loaded view='ATAC' group='group1' with N=5802 samples and D=25488 features...


Warning: 4 features(s) in view 0 have zero variance, consider removing them before training the model...

Warning: 204 features(s) in view 1 have zero variance, consider removing them before training the model...

Model options:
- Automatic Relevance Determination prior on the factors: True
- Automatic Relevance Determination prior on the weights: True
- Spike-and-slab prior on the factors: False
- Spike-and-slab prior on the weights: True
Likelihoods:
- View 0 (rna): gaussian
- View 1 (atac): poisson



GPU mode is activated


Using GPU device 0



######################################
## Training the model with seed 112 ##
######################################


---------------------------------------------------------------------------
OutOfMemoryError                          Traceback (most recent call last)
Cell In[9], line 1
----> 1 muon.tl.mofa(mdata, use_obs='union',gpu_mode=True, gpu_device=0,
      2            n_factors=20, convergence_mode='fast', seed=112, 
      3            outfile="models/chen_rna_atac.hdf5", verbose=True)

File ~/conda/envs/omicverse_py311/lib/python3.11/site-packages/muon/_core/tools.py:586, in mofa(data, groups_label, use_raw, use_layer, use_var, use_obs, likelihoods, n_factors, scale_views, scale_groups, center_groups, ard_weights, ard_factors, spikeslab_weights, spikeslab_factors, n_iterations, convergence_mode, use_float32, gpu_mode, gpu_device, svi_mode, svi_batch_size, svi_learning_rate, svi_forgetting_rate, svi_start_stochastic, smooth_covariate, smooth_warping, smooth_kwargs, save_parameters, save_data, save_metadata, seed, outfile, expectations, save_interrupted, verbose, quiet, copy)
    584 ent.build()
    585 logging.info(f"[{datetime.now().strftime('%Y-%m-%d %H:%M:%S')}] Running the model...")
--> 586 ent.run()
    588 if (
    589     smooth_kwargs is not None
    590     and "new_values" in smooth_kwargs
    591     and smooth_kwargs["new_values"]
    592     and smooth_covariate
    593 ):
    594     logging.info(f"[{datetime.now().strftime('%Y-%m-%d %H:%M:%S')}] Interpolating factors...")

File ~/conda/envs/omicverse_py311/lib/python3.11/site-packages/mofapy2/run/entry_point.py:26, in keyboardinterrupt_saver.<locals>.saver(self, *args, **kwargs)
     23 @wraps(func)
     24 def saver(self, *args, **kwargs):
     25     try:
---> 26         func(self, *args, **kwargs)
     27     # Internal methods will raise TypeError when interrupted
     28     except (KeyboardInterrupt, TypeError):

File ~/conda/envs/omicverse_py311/lib/python3.11/site-packages/mofapy2/run/entry_point.py:1400, in entry_point.run(self)
   1397 self.model.setTrainOptions(self.train_opts)
   1399 # Train the model
-> 1400 train_model(self.model)

File ~/conda/envs/omicverse_py311/lib/python3.11/site-packages/mofapy2/build_model/train_model.py:26, in train_model(model)
     23 print("#" * 38)
     24 print("\n")
---> 26 model.iterate()
     28 print("\n")
     29 print("#" * 23)

File ~/conda/envs/omicverse_py311/lib/python3.11/site-packages/mofapy2/core/BayesNet.py:265, in BayesNet.iterate(self)
    263 converged = False
    264 convergence_token = 1
--> 265 elbo.iloc[0] = self.precompute()
    266 number_factors[0] = self.dim["K"]
    267 iter_time[0] = 0.0

File ~/conda/envs/omicverse_py311/lib/python3.11/site-packages/mofapy2/core/BayesNet.py:230, in BayesNet.precompute(self)
    228 for node in self.nodes["Y"].getNodes():
    229     node.TauTrick = False  # important to do this for ELBO computation
--> 230 elbo = self.calculateELBO()
    231 for node in self.nodes["Y"].getNodes():
    232     node.TauTrick = True

File ~/conda/envs/omicverse_py311/lib/python3.11/site-packages/mofapy2/core/BayesNet.py:530, in BayesNet.calculateELBO(self, *nodes)
    528 for node in nodes:
    529     if isinstance(self.nodes[node], Multiview_Variational_Node):
--> 530         elbo[node] = float(self.nodes[node].calculateELBO(weights=weights))
    531     else:
    532         elbo[node] = float(self.nodes[node].calculateELBO())

File ~/conda/envs/omicverse_py311/lib/python3.11/site-packages/mofapy2/core/nodes/multiview_nodes.py:186, in Multiview_Mixed_Node.calculateELBO(self, weights)
    184 for m in self.activeM:
    185     if isinstance(self.nodes[m], Variational_Node):
--> 186         lb += self.nodes[m].calculateELBO() * weights[m]
    187 return lb

File ~/conda/envs/omicverse_py311/lib/python3.11/site-packages/mofapy2/core/nodes/nongaussian_nodes.py:257, in Poisson_PseudoY.calculateELBO(self)
    255 # term1 = 0.5*tau*(ZW - zeta)**2
    256 obs = gpu_utils.array(self.obs)
--> 257 term1 = 0.5 * tau * (ZZWW - 2 * ZW * zeta + gpu_utils.square(zeta))
    258 term2 = (ZW - zeta) * (
    259     gpu_utils.sigmoid(zeta) * (1.0 - obs / self.ratefn(zeta))
    260 )
    261 term3 = self.ratefn(zeta) - obs * gpu_utils.log(self.ratefn(zeta))

File cupy/_core/core.pyx:1371, in cupy._core.core._ndarray_base.__add__()

File cupy/_core/_kernel.pyx:1349, in cupy._core._kernel.ufunc.__call__()

File cupy/_core/_kernel.pyx:645, in cupy._core._kernel._get_out_args_from_optionals()

File cupy/_core/core.pyx:3006, in cupy._core.core._ndarray_init()

File cupy/_core/core.pyx:272, in cupy._core.core._ndarray_base._init_fast()

File cupy/cuda/memory.pyx:875, in cupy.cuda.memory.alloc()

File cupy/cuda/memory.pyx:1579, in cupy.cuda.memory.MemoryPool.malloc()

File cupy/cuda/memory.pyx:1600, in cupy.cuda.memory.MemoryPool.malloc()

File cupy/cuda/memory.pyx:1271, in cupy.cuda.memory.SingleDeviceMemoryPool.malloc()

File cupy/cuda/memory.pyx:1292, in cupy.cuda.memory.SingleDeviceMemoryPool._malloc()

File cupy/cuda/memory.pyx:1537, in cupy.cuda.memory.SingleDeviceMemoryPool._try_malloc()

File cupy/cuda/memory.pyx:1540, in cupy.cuda.memory.SingleDeviceMemoryPool._try_malloc()

OutOfMemoryError: Out of memory allocating 1,183,051,264 bytes (allocated so far: 11,243,994,112 bytes).

[hufanglq]

It's pretty cool!