### Training Progress Example 2

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/schwann/modelcomparison.ipynb

Example output showing training progress, indicating the current epoch, completion percentage, and time.

```text
Epoch 1424/1500:  95%|█████████▍| 1424/1500 [46:44<02:29,  1.97s/it, v_num=1]
Monitored metric elbo_validation did not improve in the last 45 records. Best score: -4652.919. Signaling Trainer to stop.
```

--------------------------------

### Training Progress Example 3

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/schwann/modelcomparison.ipynb

Example output showing training progress, indicating the current epoch, completion percentage, and time.

```text
Epoch 1050/1500:  70%|███████   | 1050/1500 [35:02<15:00,  2.00s/it, v_num=1]
```

--------------------------------

### Training Progress Example 1

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/schwann/modelcomparison.ipynb

Example output showing the training progress, including the current epoch, percentage completion, and time elapsed.

```text
Epoch 1369/1500:  91%|█████████▏| 1369/1500 [45:44<04:22,  2.01s/it, v_num=1]
Monitored metric elbo_validation did not improve in the last 45 records. Best score: -4881.255. Signaling Trainer to stop.
```

--------------------------------

### Install Latest RegVelo Release

Source: https://github.com/theislab/regvelo/blob/main/README.md

Install the latest stable version of regvelo from PyPI. This is the recommended method for most users.

```bash
pip install regvelo
```

--------------------------------

### Training Output Example 1

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/schwann/modelcomparison.ipynb

Example output from the model training process, indicating seed, GPU availability, and potential data loading bottlenecks.

```text
[rank: 0] Seed set to 0
GPU available: True (cuda), used: True
TPU available: False, using: 0 TPU cores
HPU available: False, using: 0 HPUs
You are using a CUDA device ('NVIDIA A100 80GB PCIe') that has Tensor Cores. To properly utilize them, you should set `torch.set_float32_matmul_precision('medium' | 'high')` which will trade-off precision for performance. For more details, read https://pytorch.org/docs/stable/generated/torch.set_float32_matmul_precision.html#torch.set_float32_matmul_precision
LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0]
SLURM auto-requeueing enabled. Setting signal handlers.
/home/icb/yifan.chen/miniconda3/envs/regvelo_env/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/data_connector.py:425: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=7` in the `DataLoader` to improve performance.
/home/icb/yifan.chen/miniconda3/envs/regvelo_env/lib/python3.10/site-packages/lightning/pytorch/loops/fit_loop.py:310: The number of training batches (2) is smaller than the logging interval Trainer(log_every_n_steps=10). Set a lower value for log_every_n_steps if you want to see logs for the training epoch.
/home/icb/yifan.chen/miniconda3/envs/regvelo_env/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/data_connector.py:425: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=7` in the `DataLoader` to improve performance.
```

--------------------------------

### Training Output Example 2

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/schwann/modelcomparison.ipynb

Another example output from the model training process, similar to the first, detailing system configuration and potential performance warnings.

```text
[rank: 0] Seed set to 0
GPU available: True (cuda), used: True
TPU available: False, using: 0 TPU cores
HPU available: False, using: 0 HPUs
LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0]
SLURM auto-requeueing enabled. Setting signal handlers.
/home/icb/yifan.chen/miniconda3/envs/regvelo_env/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/data_connector.py:425: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=7` in the `DataLoader` to improve performance.
/home/icb/yifan.chen/miniconda3/envs/regvelo_env/lib/python3.10/site-packages/lightning/pytorch/loops/fit_loop.py:310: The number of training batches (2) is smaller than the logging interval Trainer(log_every_n_steps=10). Set a lower value for log_every_n_steps if you want to see logs for the training epoch.
/home/icb/yifan.chen/miniconda3/envs/regvelo_env/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/data_connector.py:425: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=7` in the `DataLoader` to improve performance.
```

--------------------------------

### Setup AnnData and Initialize RegVelo Model

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/murine/03_perturbation_tutorial_murine.ipynb

Sets up the AnnData object for RegVelo and initializes the VAE model with the prepared GRN.

```python
REGVELOVI.setup_anndata(adata_baseline, spliced_layer="Ms", unspliced_layer="Mu")
vae = REGVELOVI(adata_baseline, W=W, regulators=TF)
```

--------------------------------

### Install Latest RegVelo Development Version

Source: https://github.com/theislab/regvelo/blob/main/README.md

Install the latest development version of regvelo directly from its GitHub repository. Use this if you need the newest features or bug fixes.

```bash
pip install git+https://github.com/theislab/regvelo.git@main
```

--------------------------------

### Import Libraries for SCENIC Tutorial

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/murine/01_SCENIC_tutorial.ipynb

Imports necessary Python libraries for data manipulation, analysis, and SCENIC integration. Ensure these libraries are installed in your environment.

```python
import os
import numpy as np
import pandas as pd
import scanpy as sc
import loompy as lp

import glob
```

--------------------------------

### Model Download Confirmation

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/zebrafish/grn_tutorial.ipynb

This output indicates that a required model file has already been downloaded, confirming successful setup for subsequent analysis.

```text
Finished hoxc3a -> elf1\nINFO     File rgv_model/model.pt already downloaded                                                                \n
```

--------------------------------

### Progress Bar Example

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/humanlimb/Limb_ModelComp.ipynb

This snippet shows a typical progress bar output during a computation, indicating the percentage of completion and the rate of cells processed per second.

```text
0%|          | 0/12207 [00:00<?, ?cell/s]
```

--------------------------------

### Compute Transition Matrix and Define Indices

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/zebrafish/tutorial.ipynb

Computes the transition matrix and defines start and terminal indices for simulation. Requires the 'regvelo' and 'numpy' libraries.

```python
vk = cr.kernels.VelocityKernel(adata).compute_transition_matrix()
vk = cr.kernels.VelocityKernel(adata).compute_transition_matrix()
vkt = vk.transition_matrix.A
terminal_indices = np.where(adata.obs["term_states_fwd"].isin(TERMINAL_STATES))[0]
start_indices = np.where(adata.obs["cell_type"].isin(["NPB_nohox"]))[0]
```

--------------------------------

### Import RegVelo and Load Data

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/modelcomparison/ModelComp.ipynb

Imports necessary libraries, sets a random seed for reproducibility, and loads example zebrafish data and a prior gene regulatory network.

```python
import numpy as np
import scanpy as sc
import cellrank as cr
import scvi
import scvelo as scv
from regvelo import ModelComparison # Import ModelComparison
import regvelo as rgv

# Initialize random seed
scvi.settings.seed = 0

# Data loading
ada = rgv.datasets.zebrafish_nc()
prior_net = rgv.datasets.zebrafish_grn()
```

--------------------------------

### Import Libraries for RegVelo

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/hindbrain/preprocessing.ipynb

Imports necessary libraries for data manipulation, analysis, and the RegVelo pipeline. Ensure these are installed before running.

```python
import scanpy as sc
import numpy as np
import pandas as pd

import scvelo as scv
import scvi

import regvelo as rgv
```

--------------------------------

### Import Libraries for RegVelo

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/hindbrain/modelcomparison.ipynb

Imports necessary libraries including numpy, scanpy, cellrank, scvi, scvelo, and regvelo modules. Ensure these libraries are installed before running.

```python
import numpy as np
import scanpy as sc

import cellrank as cr
import scvi
import scvelo as scv

from regvelo import REGVELOVI
from regvelo import ModelComparison
import regvelo as rgv
```

--------------------------------

### Import Libraries

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/murine/03_perturbation_tutorial_murine.ipynb

Imports essential Python libraries for data manipulation, plotting, and regulatory velocity analysis. Ensure these libraries are installed.

```python
import numpy as np
import pandas as pd

import matplotlib.pyplot as plt

import cellrank as cr
import scanpy as sc
import scvelo as scv
import scvi
from regvelo import REGVELOVI

import regvelo as rgv

import torch
```

--------------------------------

### Analysis Completion and Model Status

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/zebrafish/grn_tutorial.ipynb

This output confirms the completion of an analysis task (e.g., 'sox9b -> elf1') and notes that the required model file is already downloaded, indicating efficient setup.

```text
Finished sox9b -> elf1\nINFO     File rgv_model/model.pt already downloaded                                                                \n
```

--------------------------------

### PETSC Error Handling Example (glb1l)

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/zebrafish/grn_tutorial.ipynb

This output shows a PETSC error related to a 'Broken Pipe' signal, likely occurring during socket operations. It suggests options for debugging and provides links to PETSC FAQs.

```text
100%|██████████| 697/697 [00:00<00:00, 2890.70cell/s]\n[0]PETSC ERROR: ------------------------------------------------------------------------\n[0]PETSC ERROR: Caught signal number 13 Broken Pipe: Likely while reading or writing to a socket\n[0]PETSC ERROR: Try option -start_in_debugger or -on_error_attach_debugger\n[0]PETSC ERROR: or see https://petsc.org/release/faq/#valgrind and https://petsc.org/release/faq/\n[0]PETSC ERROR: configure using --with-debugging=yes, recompile, link, and run \n[0]PETSC ERROR: to get more information on the crash.\n100%|██████████| 697/697 [00:00<00:00, 2874.07cell/s]\n[0]PETSC ERROR: ------------------------------------------------------------------------\n[0]PETSC ERROR: Caught signal number 13 Broken Pipe: Likely while reading or writing to a socket\n[0]PETSC ERROR: Try option -start_in_debugger or -on_error_attach_debugger\n[0]PETSC ERROR: or see https://petsc.org/release/faq/#valgrind and https://petsc.org/release/faq/\n[0]PETSC ERROR: configure using --with-debugging=yes, recompile, link, and run \n[0]PETSC ERROR: to get more information on the crash.\n100%|██████████| 4/4 [00:00<00:00, 57.42/s]\n
```

--------------------------------

### Initialize Stepwise Simulation

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/zebrafish/tutorial.ipynb

Sets up the simulation method and prepares data for a perturbed model by removing the TF 'nr2f5'. Computes the transition matrix for the perturbed system.

```python
method = "stepwise"

TF = "nr2f5"
ada_perturb = adata_perturb_dict[TF].copy()
ada_perturb.obs["cell_type"] = adata.obs["cell_type"].copy()

vk_p = cr.kernels.VelocityKernel(ada_perturb).compute_transition_matrix()
vkt_p = vk_p.transition_matrix.A
```

--------------------------------

### Prepare data for one-step simulation

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/zebrafish/tutorial.ipynb

Sets up the simulation method and prepares a copy of the AnnData object for perturbation analysis. This involves copying data and computing the transition matrix.

```python
method = "one-step"

TF = "nr2f5"
adata_perturb = adata_perturb_dict[TF].copy()
adata_perturb.obs["cell_type"] = adata.obs["cell_type"].copy()

vk_p = cr.kernels.VelocityKernel(adata_perturb).compute_transition_matrix()
vkt_p = vk_p.transition_matrix.A
```

--------------------------------

### Load and Set Up Model for Perturbation Analysis

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/humanlimb/Limb_ModelComp.ipynb

Reloads a pre-trained model and sets up the necessary parameters for subsequent analyses, including the number of samples and batch size.

```python
vae_sr0 = REGVELOVI.load('vae_sr0', adata)
rgv.tl.set_output(adata, vae_sr0, n_samples=30, batch_size=adata.n_obs)
```

--------------------------------

### Run SCENIC Context Generation

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/murine/01_SCENIC_tutorial.ipynb

Executes the SCENIC context generation step using specified databases and input files. Ensure all file paths and database names are correctly set.

```python
!pyscenic ctx "adj.csv" \
    {f_db_names} \
    --annotations_fname {f_motif_path} \
    --expression_mtx_fname {f_loom_path_scenic} \
    --output "reg.csv" \
    --all_modules \
    --num_workers 24
```

--------------------------------

### Training Output and Progress

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/humanlimb/Limb_ModelComp.ipynb

This output shows the progress of the model training process, including seed initialization, GPU availability, and training progress bars. It indicates that the training is running and monitoring metrics.

```text
Output:
[rank: 0] Seed set to 0
GPU available: True (cuda), used: True
TPU available: False, using: 0 TPU cores
HPU available: False, using: 0 HPUs
You are using a CUDA device ('NVIDIA A100 80GB PCIe') that has Tensor Cores. To properly utilize them, you should set `torch.set_float32_matmul_precision('medium' | 'high')` which will trade-off precision for performance. For more details, read https://pytorch.org/docs/stable/generated/torch.set_float32_matmul_precision.html#torch.set_float32_matmul_precision
LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0]
SLURM auto-requeueing enabled. Setting signal handlers.
/home/icb/yifan.chen/miniconda3/envs/regvelo-py310-v2/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/data_connector.py:425: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=7` in the `DataLoader` to improve performance.
/home/icb/yifan.chen/miniconda3/envs/regvelo-py310-v2/lib/python3.10/site-packages/lightning/pytorch/loops/fit_loop.py:310: The number of training batches (1) is smaller than the logging interval Trainer(log_every_n_steps=10). Set a lower value for log_every_n_steps if you want to see logs for the training epoch.
/home/icb/yifan.chen/miniconda3/envs/regvelo-py310-v2/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/data_connector.py:425: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=7` in the `DataLoader` to improve performance.

```

```text
Result:
Training:   0%|          | 0/1500 [00:00<?, ?it/s]
```

--------------------------------

### Set Scanpy Verbosity and Print Versions

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/murine/01_SCENIC_tutorial.ipynb

Configures the verbosity level for Scanpy logging and prints the versions of installed packages. This is useful for debugging and ensuring compatibility.

```python
sc.settings.verbosity = 3
sc.logging.print_versions()
```

--------------------------------

### Download and Verify Model

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/zebrafish/grn_tutorial.ipynb

This snippet indicates that the rgv_model.pt file has already been downloaded and is ready for use. No action is required if the file exists.

```text
Finished elf1 -> sash1b
INFO     File rgv_model/model.pt already downloaded                                                                

```

--------------------------------

### Download and Verify Model

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/zebrafish/grn_tutorial.ipynb

This snippet indicates that the rgv_model.pt file has already been downloaded and is ready for use. No action is required if the file exists.

```text
Finished elf1 -> sema3d
INFO     File rgv_model/model.pt already downloaded                                                                

```

--------------------------------

### Identify Terminal and Start Indices

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/zebrafish/grn_tutorial.ipynb

Identifies indices for terminal states and specific cell types (e.g., 'NPB_nohox') from the AnnData object. These are used in subsequent simulations.

```python
terminal_indices = np.where(adata.obs["term_states_fwd"].isin(TERMINAL_STATES))[0]
start_indices = np.where(adata.obs["cell_type"].isin(["NPB_nohox"]))[0]
```

--------------------------------

### Download and Verify Model

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/zebrafish/grn_tutorial.ipynb

This snippet indicates that the rgv_model.pt file has already been downloaded and is ready for use. No action is required if the file exists.

```text
Finished elf1 -> sema4ba
INFO     File rgv_model/model.pt already downloaded                                                                

```

--------------------------------

### Train Models and Compare Results

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/hindbrain/modelcomparison.ipynb

This snippet shows the training process for different model types and displays the resulting labels. It's useful for initial exploration of model performance.

```python
adata = sc.read_csv("adata.csv")
adata.obs["cell_type"] = adata.obs["cell_type"].astype("category")
adata.var_names = [f"gene_{i}" for i in range(adata.shape[1])]
adata.write("adata.h5ad")

adata = sc.read_h5ad("adata.h5ad")

# Train soft clustering model
adata = sc.tl.regvelo.train(adata, model="soft_0")

# Train hard clustering model
adata = sc.tl.regvelo.train(adata, model="hard_0")

# Train regularized soft clustering model
adata = sc.tl.regvelo.train(adata, model="soft_regularized\nlam2:1.0_0")

print(adata.obs["regvelo_model"])

```

--------------------------------

### Initialize ModelComparison Object

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/hindbrain/modelcomparison.ipynb

Initializes the ModelComparison object from the regvelo library. This object is used to compare different model setups on the provided AnnData object and state transitions.

```python
comp = ModelComparison(adata = adata, state_transition=STATE_TRANSITIONS)
```

--------------------------------

### Download Pre-trained Model

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/zebrafish/grn_tutorial.ipynb

This command downloads a pre-trained GRN model. Ensure the file path is correct.

```bash
Finished elf1 -> fhl3a\nINFO     File rgv_model/model.pt already downloaded                                                                \n\n
```

--------------------------------

### PETSC Error Output (variant)

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/zebrafish/grn_tutorial.ipynb

This snippet presents another instance of the PETSC 'Broken Pipe' error, similar to the first example. It also provides guidance on debugging and troubleshooting.

```text
100%|██████████| 697/697 [00:00<00:00, 2405.58cell/s]\n[0]PETSC ERROR: ------------------------------------------------------------------------\n[0]PETSC ERROR: Caught signal number 13 Broken Pipe: Likely while reading or writing to a socket\n[0]PETSC ERROR: Try option -start_in_debugger or -on_error_attach_debugger\n[0]PETSC ERROR: or see https://petsc.org/release/faq/#valgrind and https://petsc.org/release/faq/\n[0]PETSC ERROR: configure using --with-debugging=yes, recompile, link, and run \n[0]PETSC ERROR: to get more information on the crash.\n100%|██████████| 697/697 [00:00<00:00, 2907.64cell/s]\n[0]PETSC ERROR: ------------------------------------------------------------------------\n[0]PETSC ERROR: Caught signal number 13 Broken Pipe: Likely while reading or writing to a socket\n[0]PETSC ERROR: Try option -start_in_debugger or -on_error_attach_debugger\n[0]PETSC ERROR: or see https://petsc.org/release/faq/#valgrind and https://petsc.org/release/faq/\n[0]PETSC ERROR: configure using --with-debugging=yes, recompile, link, and run \n[0]PETSC ERROR: to get more information on the crash.\n100%|██████████| 4/4 [00:00<00:00, 63.70/s]\n[0]PETSC ERROR: ------------------------------------------------------------------------\n[0]PETSC ERROR: Caught signal number 13 Broken Pipe: Likely while reading or writing to a socket\n[0]PETSC ERROR: Try option -start_in_debugger or -on_error_attach_debugger\n[0]PETSC ERROR: or see https://petsc.org/release/faq/#valgrind and https://petsc.org/release/faq/\n[0]PETSC ERROR: configure using --with-debugging=yes, recompile, link, and run \n[0]PETSC ERROR: to get more information on the crash.
```

--------------------------------

### PETSC Error during GRN Analysis (final variant)

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/zebrafish/grn_tutorial.ipynb

This snippet shows a PETSC error related to a 'Broken Pipe' signal, similar to previous examples. It includes progress bars and debugging suggestions.

```text
100%|██████████| 697/697 [00:00<00:00, 3109.97cell/s]
[0]PETSC ERROR: ------------------------------------------------------------------------
[0]PETSC ERROR: Caught signal number 13 Broken Pipe: Likely while reading or writing to a socket
[0]PETSC ERROR: Try option -start_in_debugger or -on_error_attach_debugger
[0]PETSC ERROR: or see https://petsc.org/release/faq/#valgrind and https://petsc.org/release/faq/
[0]PETSC ERROR: configure using --with-debugging=yes, recompile, link, and run 
[0]PETSC ERROR: to get more information on the crash.
100%|██████████| 697/697 [00:00<00:00, 3333.38cell/s]
[0]PETSC ERROR: ------------------------------------------------------------------------
[0]PETSC ERROR: Caught signal number 13 Broken Pipe: Likely while reading or writing to a socket

```

--------------------------------

### Download Pre-trained Model (fli1a)

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/zebrafish/grn_tutorial.ipynb

This command downloads a pre-trained GRN model for the gene 'fli1a'. The 'INFO' message indicates the model file is already present.

```bash
Finished elf1 -> fli1a\nINFO     File rgv_model/model.pt already downloaded                                                                \n\n
```

--------------------------------

### PETSC Error Handling Example

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/zebrafish/grn_tutorial.ipynb

This output shows a PETSC error related to a 'Broken Pipe' signal, likely occurring during socket operations. It suggests options for debugging and provides links to PETSC FAQs.

```text
100%|██████████| 697/697 [00:00<00:00, 1644.58cell/s]\n[0]PETSC ERROR: ------------------------------------------------------------------------\n[0]PETSC ERROR: Caught signal number 13 Broken Pipe: Likely while reading or writing to a socket\n[0]PETSC ERROR: Try option -start_in_debugger or -on_error_attach_debugger\n[0]PETSC ERROR: or see https://petsc.org/release/faq/#valgrind and https://petsc.org/release/faq/\n[0]PETSC ERROR: configure using --with-debugging=yes, recompile, link, and run \n[0]PETSC ERROR: to get more information on the crash.\n100%|██████████| 697/697 [00:00<00:00, 2734.33cell/s]\n[0]PETSC ERROR: ------------------------------------------------------------------------\n[0]PETSC ERROR: Caught signal number 13 Broken Pipe: Likely while reading or writing to a socket\n[0]PETSC ERROR: Try option -start_in_debugger or -on_error_attach_debugger\n[0]PETSC ERROR: or see https://petsc.org/release/faq/#valgrind and https://petsc.org/release/faq/\n[0]PETSC ERROR: configure using --with-debugging=yes, recompile, link, and run \n[0]PETSC ERROR: to get more information on the crash.\n100%|██████████| 4/4 [00:00<00:00, 61.10/s]\n[0]PETSC ERROR: ------------------------------------------------------------------------\n[0]PETSC ERROR: Caught signal number 13 Broken Pipe: Likely while reading or writing to a socket\n[0]PETSC ERROR: Try option -start_in_debugger or -on_error_attach_debugger\n[0]PETSC ERROR: or see https://petsc.org/release/faq/#valgrind and https://petsc.org/release/faq/\n[0]PETSC ERROR: configure using --with-debugging=yes, recompile, link, and run \n[0]PETSC ERROR: to get more information on the crash.\n
```

--------------------------------

### Initialize and Train Model Comparison

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/humanlimb/Limb_ModelComp.ipynb

Initialize the `ModelComparison` object with the annotated data, terminal states, and number of states. Then, train the models using a list of specified model types, lambda values, and number of repeats.

```python
comp = ModelComparison(adata = adata,terminal_states=TERMINAL_STATES, n_states=n_STATES)
comp.train(model_list=['soft','hard','soft_regularized'],
           lam2=[1.0],
           n_repeat=3)
```

--------------------------------

### Instantiate RegVelo Model (REGVELOVI)

Source: https://context7.com/theislab/regvelo/llms.txt

Creates a REGVELOVI model that incorporates a GRN prior into the velocity VAE. The weight matrix `W` encodes the GRN skeleton, and `lam`/`lam2` control regularization strength.

```python
import torch
import regvelo as rgv

# Example instantiation (parameters may vary)
# model = rgv.REGVELOVI(adata, W=adata.uns["skeleton"], lam=1.0, lam2=0.0)

```

--------------------------------

### PETSC Error Handling Example (fzd3a)

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/zebrafish/grn_tutorial.ipynb

This output shows a PETSC error related to a 'Broken Pipe' signal, likely occurring during socket operations. It suggests options for debugging and provides links to PETSC FAQs.

```text
100%|██████████| 697/697 [00:00<00:00, 3138.32cell/s]\n[0]PETSC ERROR: ------------------------------------------------------------------------\n[0]PETSC ERROR: Caught signal number 13 Broken Pipe: Likely while reading or writing to a socket\n[0]PETSC ERROR: Try option -start_in_debugger or -on_error_attach_debugger\n[0]PETSC ERROR: or see https://petsc.org/release/faq/#valgrind and https://petsc.org/release/faq/\n[0]PETSC ERROR: configure using --with-debugging=yes, recompile, link, and run \n[0]PETSC ERROR: to get more information on the crash.\n100%|██████████| 697/697 [00:00<00:00, 2760.82cell/s]\n[0]PETSC ERROR: ------------------------------------------------------------------------\n[0]PETSC ERROR: Caught signal number 13 Broken Pipe: Likely while reading or writing to a socket\n[0]PETSC ERROR: Try option -start_in_debugger or -on_error_attach_debugger\n[0]PETSC ERROR: or see https://petsc.org/release/faq/#valgrind and https://petsc.org/release/faq/\n[0]PETSC ERROR: configure using --with-debugging=yes, recompile, link, and run \n[0]PETSC ERROR: to get more information on the crash.\n100%|██████████| 4/4 [00:00<00:00, 57.42/s]\n[0]PETSC ERROR: ------------------------------------------------------------------------\n[0]PETSC ERROR: Caught signal number 13 Broken Pipe: Likely while reading or writing to a socket\n[0]PETSC ERROR: Try option -start_in_debugger or -on_error_attach_debugger\n[0]PETSC ERROR: or see https://petsc.org/release/faq/#valgrind and https://petsc.org/release/faq/\n[0]PETSC ERROR: configure using --with-debugging=yes, recompile, link, and run \n[0]PETSC ERROR: to get more information on the crash.\n
```

--------------------------------

### PyTorch Lightning Trainer Configuration and Warnings

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/modelcomparison/ModelComp.ipynb

Shows the configuration of the PyTorch Lightning Trainer, including GPU availability and potential performance bottlenecks related to DataLoader workers and logging intervals. It's recommended to increase `num_workers` for DataLoaders and adjust `log_every_n_steps` if needed.

```text
GPU available: True (cuda), used: True
TPU available: False, using: 0 TPU cores
HPU available: False, using: 0 HPUs
LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0]
SLURM auto-requeueing enabled. Setting signal handlers.
/home/icb/yifan.chen/miniconda3/envs/regvelo-test/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/data_connector.py:425: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=7` in the `DataLoader` to improve performance.
/home/icb/yifan.chen/miniconda3/envs/regvelo-test/lib/python3.10/site-packages/lightning/pytorch/loops/fit_loop.py:310: The number of training batches (1) is smaller than the logging interval Trainer(log_every_n_steps=10). Set a lower value for log_every_n_steps if you want to see logs for the training epoch.
/home/icb/yifan.chen/miniconda3/envs/regvelo-test/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/data_connector.py:425: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=7` in the `DataLoader` to improve performance.
```

--------------------------------

### Analysis Completion and Model Download

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/zebrafish/grn_tutorial.ipynb

This output confirms the completion of an analysis step (e.g., 'sox10 -> elf1') and indicates that the necessary model file was already downloaded, streamlining the process.

```text
Finished sox10 -> elf1\nINFO     File rgv_model/model.pt already downloaded                                                                \n
```

--------------------------------

### Initialize REGVELOVI Model

Source: https://context7.com/theislab/regvelo/llms.txt

Loads the GRN skeleton and initializes the REGVELOVI model with specified parameters. Ensure AnnData is set up with spliced and unspliced layers.

```python
W = torch.tensor(adata.uns["skeleton"].values, dtype=torch.float32)
TF_list = adata.uns["regulators"].tolist()

rgv.REGVELOVI.setup_anndata(adata, spliced_layer="Ms", unspliced_layer="Mu")

model = rgv.REGVELOVI(
    adata,
    W=W,                    # GRN weight matrix (targets x regulators)
    regulators=TF_list,     # list of TF gene names
    soft_constraint=True,   # use soft GRN constraint (recommended)
    lam=1.0,                # GRN prior regularization strength
    lam2=0.0,               # L1 Jacobian regularization
    n_hidden=256,
    n_latent=10,
    n_layers=1,
    dropout_rate=0.1,
)
print(model)
```

--------------------------------

### Initialize ModelComparison object

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/schwann/modelcomparison.ipynb

Initializes the `ModelComparison` class from `regvelo` with the loaded AnnData object.

```python
comp = ModelComparison(adata = adata)
```

--------------------------------

### Seed Initialization

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/modelcomparison/ModelComp.ipynb

Confirms that a random seed has been set for the training process, ensuring reproducibility.

```text
[rank: 0] Seed set to 2
```

--------------------------------

### Download Model File

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/zebrafish/grn_tutorial.ipynb

Indicates that the model file 'rgv_model/model.pt' has already been downloaded and is ready for use.

```text
Output:
Finished tfap2a -> elf1
INFO     File rgv_model/model.pt already downloaded
```

--------------------------------

### Model Training Output

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/zebrafish/tutorial.ipynb

This output indicates the progress and status of the model training process, including GPU availability and potential performance bottlenecks.

```text
training model...


```

```text
GPU available: True (cuda), used: True
TPU available: False, using: 0 TPU cores
HPU available: False, using: 0 HPUs
LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0]
SLURM auto-requeueing enabled. Setting signal handlers.
/home/icb/yifan.chen/miniconda3/envs/regvelo-py310-v2/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/data_connector.py:425: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=7` in the `DataLoader` to improve performance.
/home/icb/yifan.chen/miniconda3/envs/regvelo-py310-v2/lib/python3.10/site-packages/lightning/pytorch/loops/fit_loop.py:310: The number of training batches (1) is smaller than the logging interval Trainer(log_every_n_steps=10). Set a lower value for log_every_n_steps if you want to see logs for the training epoch.
/home/icb/yifan.chen/miniconda3/envs/regvelo-py310-v2/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/data_connector.py:425: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=7` in the `DataLoader` to improve performance.


```

```text
Result:
  0%|          | 0/1500 [00:00<?, ?it/s]

```

```text
Output:
Monitored metric elbo_validation did not improve in the last 45 records. Best score: -2400.929. Signaling Trainer to stop.
save model...
inferring perturbation...
INFO     File                                                                                                      
         /ictstr01/groups/ml01/workspace/yifan.chen/regvelo_reproducibility/notebooks/tutorials/zebrafish/perturb_r
         epeat_runs/rgv_model_0/model.pt already downloaded                                                        
Computing transition matrix using `'deterministic'` model


```

```text
Result:
  0%|          | 0/697 [00:00<?, ?cell/s]

```

```text
Output:
Using `softmax_scale=9.3675`


```

```text
Result:
  0%|          | 0/697 [00:00<?, ?cell/s]

```

```text
Output:
    Finish (0:00:01)
Computing transition matrix based on `adata.obsp['connectivities']`
    Finish (0:00:00)
Computing Schur decomposition
Adding `adata.uns['eigendecomposition_fwd']`
       `.schur_vectors`
       `.schur_matrix`
       `.eigendecomposition`
    Finish (0:00:00)
Computing `8` macrostates
Adding `.macrostates`
       `.macrostates_memberships`
       `.coarse_T`
       `.coarse_initial_distribution
       `.coarse_stationary_distribution`
       `.schur_vectors`
       `.schur_matrix`
       `.eigendecomposition`
    Finish (0:00:01)
Adding `adata.obs['term_states_fwd']`
       `adata.obs['term_states_fwd_probs']`
       `.terminal_states`
       `.terminal_states_probabilities`
       `.terminal_states_memberships
    Finish`
Computing fate probabilities
Adding `adata.obsm['lineages_fwd']`
       `.fate_probabilities`
    Finish (0:00:00)
INFO     File                                                                                                      
         /ictstr01/groups/ml01/workspace/yifan.chen/regvelo_reproducibility/notebooks/tutorials/zebrafish/perturb_r
         epeat_runs/rgv_model_0/model.pt already downloaded                                                        
Computing transition matrix using `'deterministic'` model


```

```text
Result:
  0%|          | 0/697 [00:00<?, ?cell/s]

```

```text
Output:
Using `softmax_scale=9.6567`


```

```text
Result:
  0%|          | 0/697 [00:00<?, ?cell/s]

```

```text
Output:
    Finish (0:00:00)
Computing transition matrix based on `adata.obsp['connectivities']`
    Finish (0:00:00)
Computing Schur decomposition
Adding `adata.uns['eigendecomposition_fwd']`

```

--------------------------------

### REGVELOVI.__init__

Source: https://context7.com/theislab/regvelo/llms.txt

Instantiates the RegVelo model (REGVELOVI), which is a VAE that incorporates a GRN prior into the velocity modeling. It takes the AnnData object and GRN information as input.

```APIDOC
## REGVELOVI.__init__

### Description
Creates a REGVELOVI model incorporating a GRN prior into the velocity VAE. The weight matrix `W` encodes the GRN skeleton (targets × regulators), and `lam` / `lam2` control regularization strength.

### Method
`REGVELOVI(adata, W, lam=1.0, lam2=1.0, **kwargs)`

### Parameters
- **adata** (AnnData) - AnnData object registered with `REGVELOVI.setup_anndata`.
- **W** (torch.Tensor or np.ndarray) - The GRN skeleton weight matrix (targets × regulators).
- **lam** (float, optional) - Regularization strength for the GRN prior. Defaults to 1.0.
- **lam2** (float, optional) - Second regularization parameter. Defaults to 1.0.
- **kwargs** - Additional keyword arguments passed to the underlying scvi-tools model.

### Returns
- **REGVELOVI** - An instance of the REGVELOVI model.
```

--------------------------------

### Initialize and Prepare GRN Velocity Model

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/zebrafish/tutorial.ipynb

Initializes the RegVeloVI model with the AnnData object, GRN prior, and transcription factor list. It sets up the model for training with soft constraints disabled.

```python
REGVELOVI.setup_anndata(adata, spliced_layer="Ms", unspliced_layer="Mu")
reg_vae = REGVELOVI(adata, W=W.T, regulators=TF_list, soft_constraint=False)
```

--------------------------------

### Download Pre-trained Model (fhod1)

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/zebrafish/grn_tutorial.ipynb

This command downloads a pre-trained GRN model for the gene 'fhod1'. The 'INFO' message indicates the model file is already present.

```bash
Finished elf1 -> fhod1\nINFO     File rgv_model/model.pt already downloaded                                                                \n\n
```

--------------------------------

### Load Schwann cell dataset

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/schwann/modelcomparison.ipynb

Loads the preprocessed mouse neural crest and Schwann cell dataset using `regvelo.datasets.schwann()`.

```python
adata = rgv.datasets.schwann()
```

--------------------------------

### Load and Prepare Prior GRN

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/murine/03_perturbation_tutorial_murine.ipynb

Loads the prior GRN from adata_baseline.uns['skeleton'], converts it to a PyTorch tensor, and transposes it for RegVelo.

```python
W = adata_baseline.uns["skeleton"].copy()
W = torch.tensor(np.array(W)).int()
W = W.T
```

--------------------------------

### Configure Plotting Parameters

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/zebrafish/tutorial.ipynb

Sets up matplotlib parameters for generating publication-quality figures, including font type, DPI, transparency, and colormap.

```python
plt.rcParams["svg.fonttype"] = "none"
scv.settings.set_figure_params("scvelo", dpi=100, transparent=True, fontsize=14, color_map="viridis")
```

--------------------------------

### Train Models

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/schwann/modelcomparison.ipynb

Initiates the training process for specified models. It's recommended to set batch_size to min(adata.shape[0], 5000) to prevent out-of-memory errors.

```python
comp.train(model_list=['soft','hard','soft_regularized'],
           lam2=[1.0],
           n_repeat=1, 
           batch_size=min(adata.shape[0], 5000))
```

--------------------------------

### Import necessary libraries

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/schwann/modelcomparison.ipynb

Imports all required libraries for data analysis, including scanpy, cellrank, scvi, scvelo, and regvelo.

```python
import numpy as np
import scanpy as sc
import cellrank as cr
import scvi
import scvelo as scv

import regvelo as rgv
from regvelo import REGVELOVI
from regvelo import ModelComparison 
```

--------------------------------

### Load a Trained Model and Set Output

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/schwann/modelcomparison.ipynb

Load a previously saved model and set its output for trajectory analysis. This involves specifying the number of samples and batch size for the computation.

```python
vae_s = REGVELOVI.load('vae_s', adata)
rgv.tl.set_output(adata, vae_s, n_samples=30, batch_size=5000)
```

--------------------------------

### Import necessary libraries

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/murine/02_RegVelo_preparation.ipynb

Imports the required libraries for data manipulation and analysis with RegVelo and Scanpy.

```python
import scvelo as scv
import scanpy as sc
import pandas as pd
import numpy as np

import regvelo as rgv
```

--------------------------------

### REGVELOVI.setup_anndata

Source: https://context7.com/theislab/regvelo/llms.txt

Registers the AnnData object for use with the REGVELOVI model. This class method must be called before instantiating the model, specifying the layers for spliced and unspliced counts.

```APIDOC
## REGVELOVI.setup_anndata

### Description
Class method that registers the spliced and unspliced layers with the scvi-tools data manager. Must be called before instantiating the model.

### Method
`REGVELOVI.setup_anndata(adata, spliced_layer='Ms', unspliced_layer='Mu')`

### Parameters
- **adata** (AnnData) - The AnnData object to register.
- **spliced_layer** (str, optional) - The layer containing spliced counts. Defaults to 'Ms'.
- **unspliced_layer** (str, optional) - The layer containing unspliced counts. Defaults to 'Mu'.

### Returns
- None. Modifies the AnnData object in place for scvi-tools compatibility.
```

--------------------------------

### Download Model File

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/zebrafish/grn_tutorial.ipynb

This code snippet downloads the 'rgv_model.pt' file if it doesn't already exist. It indicates that the file is already present.

```python
Finished elf1 -> hsp70.2\nINFO     File rgv_model/model.pt already downloaded                                                                \n
```

```python
Finished elf1 -> hspa5\nINFO     File rgv_model/model.pt already downloaded                                                                \n
```

```python
Finished elf1 -> hspb8\nINFO     File rgv_model/model.pt already downloaded                                                                \n
```

--------------------------------

### Initialize and Compute Macrostates with GPCCA

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/murine/03_perturbation_tutorial_murine.ipynb

Initializes a GPCCA estimator with the computed VelocityKernel and computes macrostates based on a specified number of states and a cluster key.

```python
estimator = cr.estimators.GPCCA(vk)
estimator.compute_macrostates(n_states=10, cluster_key="celltype_update")
```

--------------------------------

### Load and Process Regulon Matrix

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/murine/01_SCENIC_tutorial.ipynb

Loads the saved regulon matrix and preprocesses it for RegVelo. This includes extracting transcription factor names and collapsing duplicate TFs by summing their rows.

```python
# load saved regulon-target matrix
reg = pd.read_csv("regulon_mat_all_regulons.csv", index_col = 0)

reg.index = reg.index.str.extract(r"(\w+)")[0]
reg = reg.groupby(reg.index).sum()
```

--------------------------------

### Import necessary libraries

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/zebrafish/grn_tutorial.ipynb

Imports all required libraries for RegVelo analysis, including scanpy, scvi, scvelo, and matplotlib.

```python
import numpy as np
import pandas as pd
import scanpy as sc
import cellrank as cr
import scvi
import scvelo as scv

import regvelo as rgv

import matplotlib.pyplot as plt
import mplscience
import seaborn as sns
```

--------------------------------

### Download Pre-trained Model (hat1)

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/zebrafish/grn_tutorial.ipynb

Downloads a pre-trained model for gene regulatory network analysis, specifically for the gene hat1. This is useful for quickly setting up the environment for GRN analysis.

```python
Finished elf1 -> hat1
INFO     File rgv_model/model.pt already downloaded                                                                


```

```python
100%|██████████| 697/697 [00:00<00:00, 2225.21cell/s]
[0]PETSC ERROR: ------------------------------------------------------------------------
[0]PETSC ERROR: Caught signal number 13 Broken Pipe: Likely while reading or writing to a socket
[0]PETSC ERROR: Try option -start_in_debugger or -on_error_attach_debugger
[0]PETSC ERROR: or see https://petsc.org/release/faq/#valgrind and https://petsc.org/release/faq/
[0]PETSC ERROR: configure using --with-debugging=yes, recompile, link, and run 
[0]PETSC ERROR: to get more information on the crash.
100%|██████████| 697/697 [00:00<00:00, 2742.22cell/s]

```

--------------------------------

### Download Model and Run Inference

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/zebrafish/grn_tutorial.ipynb

Downloads a pre-trained model and runs inference for a specific gene. This process may involve significant data transfer and processing.

```python
Output:
Finished elf1 -> ildr2
INFO     File rgv_model/model.pt already downloaded                                                                


```

--------------------------------

### Download Model and Run Inference

Source: https://github.com/theislab/regvelo/blob/main/docs/tutorials/zebrafish/grn_tutorial.ipynb

Downloads a pre-trained model and runs inference for a specific gene. This process may involve significant data transfer and processing.

```python
Output:
Finished elf1 -> inka1a
INFO     File rgv_model/model.pt already downloaded                                                                


```