eqcctpro 0.6.3

EQCCTPro: A powerful seismic event detection toolkit

EQCCTPro: powerful seismic event detection toolkit

EQCCTPro is a high-performace seismic event detection and processing framework that leverages DL-pickers, like EQCCT, to process seismic data efficiently. It enables users to fully leverage the computational ability of their computing resources for maximum performance for simultaneous seismic waveform processing, achieving real-time performance by identifying and utilizing the optimal computational configurations for their hardware. More information about the development, capabilities, and real-world applications about EQCCTPro can be read about in our research publication here.

Features

• Supports both CPU and GPU execution
• Configurable parallelism execution for optimized performance
• Includes tools for evaluating system performance for optimal usecase configurations
• Automatic selection of best-usecase configurations
• Efficient handling of large-scale seismic data

Installation Guide

There are two installation methods for EQCCTPro:

1. Method 1: Install EQCCTPro out of the box (for experienced users)
2. Method 2: Install EQCCTPro with sample waveform data (recommended for first-time users)

It is highly recommended that first-time users pull the EQCCTPro folder, which includes sample waveform data and code to help get acquainted with EQCCTPro.

---

Method 1: Install EQCCTPro (No Sample Data)

This method installs only the EQCCTPro package without the sample waveform data.

Step 1: Create a Clean Conda Environment for the Install

EQCCTPro requires Python 3.10.14 or higher as well as minimum Tensorflow packages. If you have a clean working environment, you can simply run pip install eqcctpro. However, if you have a nonclean environment, its highly recommended to create a new conda environment so that you can install the necessary packages safely with no issues. You can create a new conda environment with the correct Python version by using the following commands:

conda create --name yourenvironemntname python=3.10.14 -y
conda activate yourenvironemntname 
python3 --version

Expected output:

Python 3.10.14

After activating your new conda environment, run the following command:

pip install eqcctpro

You will have access to EQCCTPro and its functionality. However you will not have immediate access to the provided sample waveform data to use for testing. You can pull the waveform data either by downloading the .zip file from the repository or by following step 3.

Step 3 (Optional): Pull the EQCCTPro Folder

Although not required, it is highly recommended to pull the EQCCTPro folder to gain access to sample waveform data for testing.

mkdir my_work_directory
cd my_work_directory
git clone --depth 1 --filter=tree:0 https://github.com/ut-beg-texnet/eqcct.git --sparse
cd eqcct
git sparse-checkout set eqcctpro

---

Method 2: Install EQCCTPro with Sample Data (Recommended for First-Time Users)

This method sets up EQCCTPro with a pre-created conda environment and sample waveform data.

Step 1: Clone the EQCCTPro Repository

mkdir my_work_directory
cd my_work_directory
git clone --depth 1 --filter=tree:0 https://github.com/ut-beg-texnet/eqcct.git --sparse
cd eqcct
git sparse-checkout set eqcctpro

Step 2: Create and Activate the Conda Environment

A pre-configured conda environment is included in the repository to handle all dependencies.

conda env create -f environment.yml
conda activate eqcctpro

More Information

For additional package updates, continue to check either this repository or visit the EQCCTPro PyPI page:
🔗 EQCCTPro on PyPI

---

Using Sample Waveform Data

To understand how EQCCTPro works, it is highly recommended to use provided sample seismic waveform data as the data source when testing the package.

1-minute long sample seismic waveforms from 229 TexNet stations have been provided in the repository under the 230_stations_1_min_dt.zip file.

Step 1: Unzip the Sample Wavefrom Data

After downloading the .zip file through the GitHub methods above, run:

unzip 230_stations_1_min_dt.zip

Step 2: Check and Understand the Directory Structure

The extracted data will contain a timechunk subdirectories, comprised of multiple station directories:

[skevofilaxc 230_stations_1_min_dt]$ ls
20241215T120000Z_20241215T120100Z
[skevofilaxc 230_stations_1_min_dt]$ cd 20241215T120000Z_20241215T120100Z
237B  BP01  CT02  DG02  DG10  EE04  EF07  EF54  EF63  EF69  EF77   FOAK3  FW06  FW14  HBVL  LWM2  MB05  MB12  MB19   MBBB3  MID03  NM01  OG02  PB05  PB11  PB19  PB26  PB34  PB41  PB51  PB57  PH03  SA06  SGCY  SN02  SN10  WB03  WB09  YK01
435B  BRDY  CV01  DG04  DKNS  EF02  EF08  EF56  EF64  EF71  ELG6   FOAK4  FW07  FW15  HNDO  LWM3  MB06  MB13  MB21   MBBB5  MLDN   NM02  OG04  PB06  PB12  PB21  PB28  PB35  PB42  PB52  PB58  PL01  SA07  SM01  SN03  SNAG  WB04  WB10
ALPN  BW01  CW01  DG05  DRIO  EF03  EF09  EF58  EF65  EF72  ET02   FW01   FW09  GV01  HP01  MB01  MB07  MB15  MB22   MBBB6  MNHN   NM03  OZNA  PB07  PB14  PB22  PB29  PB37  PB43  PB53  PB59  PLPT  SA09  SM02  SN04  TREL  WB05  WB11
APMT  CF01  DB02  DG06  DRZT  EF04  EF51  EF59  EF66  EF74  FLRS   FW02   FW11  GV02  HP02  MB02  MB08  MB16  MB25   MG01   MO01   ODSA  PB01  PB08  PB16  PB23  PB30  PB38  PB44  PB54  PCOS  POST  SAND  SM03  SN07  VHRN  WB06  WB12
AT01  CRHG  DB03  DG07  EE02  EF05  EF52  EF61  EF67  EF75  FOAK1  FW04   FW12  GV03  INDO  MB03  MB09  MB17  MBBB1  MID01  NGL01  OE01  PB03  PB09  PB17  PB24  PB32  PB39  PB46  PB55  PECS  SA02  SD01  SM04  SN08  VW01  WB07  WTFS
BB01  CT01  DB04  DG09  EE03  EF06  EF53  EF62  EF68  EF76  FOAK2  FW05   FW13  GV04  LWM1  MB04  MB11  MB18  MBBB2  MID02  NGL02  OG01  PB04  PB10  PB18  PB25  PB33  PB40  PB47  PB56  PH02  SA04  SE01  SMWD  SN09  WB02  WB08  WW01

Each subdirectory contains mSEED files of different waveform components:

[skevofilaxc PB35]$ ls
TX.PB35.00.HH1__20241215T115800Z__20241215T120100Z.mseed  TX.PB35.00.HHZ__20241215T115800Z__20241215T120100Z.mseed
TX.PB35.00.HH2__20241215T115800Z__20241215T120100Z.mseed

EQCCT (i.e., the ML model) requires at least one pose per station for detection, but using multiple poses enhances P and S wave directionality.

You have successfully installed EQCCTPro and set up the required sample waveform dataset for testing.

---

Using EQCCTPro

There are three main capabilities of EQCCTPro:

1. Process mSEED data from singular or multiple seismic stations using either CPUs or GPUs
2. Evaluate your system to identify the optimal parallelization configurations needed to get the minimum runtime performance out of your system
3. Identify and return back the optimal parallelization configurations for both specific and general-use usecases for both CPU (a) and GPU applications (b)

These capabilities are achieved using the following core functions:

• EQCCTMSeedRunner (for processing mSEED data)

• EvaluateSystem (for system evaluation)

• OptimalCPUConfigurationFinder (for CPU configuration optimization)

• OptimalGPUConfigurationFinder (for GPU configuration optimization)

---

Processing mSEED data using EQCCTPro (EQCCTMSeedRunner)

To process mSEED from various seismic stations, use the EQCCTMSeedRunner class. EQCCTMSeedRunner enables users to process multiple mSEED from a given input directory, which consists of station directories formatted as follows:

[skevofilaxc 230_stations_1_min_dt]$ ls
20241215T120000Z_20241215T120100Z
[skevofilaxc 230_stations_1_min_dt]$ cd 20241215T120000Z_20241215T120100Z
237B  BP01  CT02  DG02  DG10  EE04  EF07  EF54  EF63  EF69  EF77   FOAK3  FW06  FW14  HBVL  LWM2  MB05  MB12  MB19   MBBB3  MID03  NM01  OG02  PB05  PB11  PB19  PB26  PB34  PB41  PB51  PB57  PH03  SA06  SGCY  SN02  SN10  WB03  WB09  YK01
435B  BRDY  CV01  DG04  DKNS  EF02  EF08  EF56  EF64  EF71  ELG6   FOAK4  FW07  FW15  HNDO  LWM3  MB06  MB13  MB21   MBBB5  MLDN   NM02  OG04  PB06  PB12  PB21  PB28  PB35  PB42  PB52  PB58  PL01  SA07  SM01  SN03  SNAG  WB04  WB10
ALPN  BW01  CW01  DG05  DRIO  EF03  EF09  EF58  EF65  EF72  ET02   FW01   FW09  GV01  HP01  MB01  MB07  MB15  MB22   MBBB6  MNHN   NM03  OZNA  PB07  PB14  PB22  PB29  PB37  PB43  PB53  PB59  PLPT  SA09  SM02  SN04  TREL  WB05  WB11
APMT  CF01  DB02  DG06  DRZT  EF04  EF51  EF59  EF66  EF74  FLRS   FW02   FW11  GV02  HP02  MB02  MB08  MB16  MB25   MG01   MO01   ODSA  PB01  PB08  PB16  PB23  PB30  PB38  PB44  PB54  PCOS  POST  SAND  SM03  SN07  VHRN  WB06  WB12
AT01  CRHG  DB03  DG07  EE02  EF05  EF52  EF61  EF67  EF75  FOAK1  FW04   FW12  GV03  INDO  MB03  MB09  MB17  MBBB1  MID01  NGL01  OE01  PB03  PB09  PB17  PB24  PB32  PB39  PB46  PB55  PECS  SA02  SD01  SM04  SN08  VW01  WB07  WTFS
BB01  CT01  DB04  DG09  EE03  EF06  EF53  EF62  EF68  EF76  FOAK2  FW05   FW13  GV04  LWM1  MB04  MB11  MB18  MBBB2  MID02  NGL02  OG01  PB04  PB10  PB18  PB25  PB33  PB40  PB47  PB56  PH02  SA04  SE01  SMWD  SN09  WB02  WB08  WW01

Where each subdirectory is named after station code. If you wish to use create your own input directory with custom waveform mSEED files, please follow the above naming conventions. Otherwise, EQCCTPro will not work. Create subdirectories for each timechunk (sub-parent directories) and for each station (child directories). The station directories should be named as shown above. Each timechunk directory spans from the start of the analysis period minus the waveform overlap to the end of the analysis period, based on the defined timechunk duration.

For example:

[skevofilaxc 230_stations_2hr_1_hr_dt]$ ls
20241215T115800Z_20241215T130000Z  20241215T125800Z_20241215T140000Z

The timechunk time length is 1 hour long. At the same time, we use a waveform overlap of 2 minutes. Hence: 20241215T115800Z_20241215T130000Z spans from 11:58:00 to 13:00:00 UTC on Dec 15, 2024 and 20241215T125800Z_20241215T140000Z spans from 12:58:00 to 14:00:00 UTC on Dec 15, 2024

Each station subdirectory, such as PB35, are made up of mSEED files from seismometer different poses (EX. N, E, Z):

[skevofilaxc PB35]$ ls
TX.PB35.00.HH1__20241215T115800Z__20241215T120100Z.mseed  TX.PB35.00.HHZ__20241215T115800Z__20241215T120100Z.mseed
TX.PB35.00.HH2__20241215T115800Z__20241215T120100Z.mseed

EQCCT only needs one pose for the detection to occur, however more poses allow for better detection of the direction of the P and S waves.

After setting up or utilizing the provided sample waveform directory, and install eqcctpro, import EQCCTMseedRunner as show below:

from eqcctpro import EQCCTMSeedRunner

eqcct_runner = EQCCTMSeedRunner(
    use_gpu=False,
    intra_threads=1,
    inter_threads=1,
    cpu_id_list=[0,1,2,3,4],
    input_dir='/path/to/mseed',
    output_dir='/path/to/outputs',
    log_filepath='/path/to/outputs/eqcctpro.log',
    P_threshold=0.001,
    S_threshold=0.02,
    p_model_filepath='/path/to/model_p.h5',
    s_model_filepath='/path/to/model_s.h5',
    number_of_concurrent_station_predictions=5,
    number_of_concurrent_timechunk_predictions=2
    best_usecase_config=True,
    csv_dir='/path/to/csv',
    selected_gpus=[0],
    set_vram_mb=24750,
    specific_stations='AT01, BP01, DG05',
    start_time='2024-12-14 12:00:00',
    end_time='2024-12-15 12:00:00',
    timechunk_dt=1, 
    waveform_overlap=2)

eqcct_runner.run_eqcctpro()

EQCCTMseedRunner has multiple input parameters that need to be configured and are defined below:

• use_gpu (bool): True or False
  • Tells Ray to use either the GPU(s) (True) or CPUs (False) on your computer to process the waveforms in the entire workflow
  • Further specification of which GPU(s) and CPU(s) are provided in the parameters below
• intra_threads (int): default = 1
  • Controls how many intra-parallelism threads Tensorflow can use
• inter_threads (int): default = 1
  • Controls how many inter-parallelism threads Tensorflow can use
• cpu_id_list (list): default = [1]
  • List that defines which specific CPU cores that sched_setaffinity will allocate for executing the current EQCCTPro process.
  • Allows for specific allocation and limitation of CPUs for a given EQCCTPro process
    • "I want this program to run only on these specific cores."
• input_dir (str)
  • Directory path to the the mSEED directory
  • EX. /home/skevofilaxc/my_work_directory/eqcct/eqcctpro/230_stations_1_min_dt
• output_dir (str)
  • Directory path to where the output picks and logs will be sent
  • Doesn't need to exist, will be created if doesn't exist
  • Recommended to be in the same working directory as the input directory for convience
• log_filepath (str)
  • Filepath to where the EQCCTPro log will be written to and stored
  • Doesn't need to exist, will be created if doesn't exist
  • Recommended to be in the output directory and called eqcctpro.log, however the name can be changed for your own purposes
• P_threshold (float): default = 0.001
  • Threshold in which the P probabilities above it will be considered as P arrival
• S_threshold (float): default = 0.02
  • Threshold in which the S probabilities above it will be considered as S arrival
• p_model_filepath (str)
  • Filepath to where the P EQCCT detection model is stored
• s_model_filepath (str)
  • Filepath to where the S EQCCT detection model is stored
• number_of_concurrent_station_predictions (int)
  • The number of concurrent EQCCT detection tasks that can happen simultaneously on a given number of resources
  • EX. if number_of_concurrent_station_predictions = 5, up to 5 EQCCT instances can simultaneously analyze waveforms from 5 distinct seismic stations
  • To use the optimal parameter value for this param, use the EvaluateSystem class (can be found below)
• number_of_concurrent_timechunk_predictions (int): default = None
  • The number of timechunks running in parallel
  • Avoids the sequential processing of timechunks by processing multiple timechunks in parallel, exponetially reducing runtime
• best_usecase_config (bool): default = False
  • If True, will override inputted cpu_id_list, number_of_concurrent_predictions, intra_threads, inter_threads values for the best overall use-case configurations
  • Best overall use-case configurations are defined as the best overall input configurations that minimize runtime while doing the most amount of processing with your available hardware
  • Can only be used if EvaluateSystem has been run
• csv_dir (str)
  • Directory path containing the CSV's outputted by EvaluateSystem that contain the trial data that will be used to find the best_usecase_config
  • Script will look for specific files, will only exist if EvaluateSystem has been run
• selected_gpus (list): default = None
  • List of GPU IDs on your computer you want to use if use_gpu = True
  • None existing GPU IDs will cause the code to exit
• set_vram_mb (float)
  • Value of the maximum amount of VRAM EQCCTPro can use
  • Must be a real value that is based on your hardware's physical memory space, if it exceeds the space the code will break due to OutOfMemoryError
• specific_stations (str): default = None
  • String that contains the "list" of stations you want to only analyze
  • EX. Out of the 50 sample stations in 230_stations_1_min_dt, if I only want to analyze AT01, BP01, DG05, then specific_stations='AT01, BP01, DG05'.
  • Removes the need to move station directories around to be used as input, can contain all stations in one directory for access
• start_time (str): default = None
  • The start time of the area of time that is being analyzed
  • EX. 2024-12-14 12:00:00
  • Must follow the following convention YYYY-MO-DA HR:MI:SC
  • Used to create a list of defined timechunks from the defined analysis timeframe
  • Must be the exact start time of the analysis time period (does not include the prior waveform overlap time IE. 2024-12-15 11:58:00 for a 2 minute waveform overlap time)
  • Also used in the EvaluateSystem() class to help users note the analysis timeframe in the results CSV file for future result review
• end_time (str): default = None
  • The end time of the area of time that is being analyzed
  • EX. 2024-12-15 12:00:00
  • Must follow the following convention YYYY-MO-DA HR:MI:SC
  • Used to create a list of defined timechunks from the defined analysis timeframe
  • Must be the exact end time of the analysis time period
  • Also used in the EvaluateSystem() class to help users note the analysis timeframe in the results CSV file for future result review
• timechunk_dt (int): default = None
  • The length each time chunk is (in minutes)
  • EX. timechunk_dt = 10 and the analysis period is 30 minutes, then three 10-minute long timechunks will be created
• waveform_overlap (int): default = None
  • The duration (in minutes) for which each waveform overlaps with the others

---

Evaluating Your Systems Runtime Performance Capabilites

To evaluate your system’s runtime performance capabilites for both your CPU(s) and GPU(s), the EvaluateSystem class allows you to autonomously evaluate your system:

from eqcctpro import EvaluateSystem

eval_gpu = EvaluateSystem(
                mode='gpu',
                intra_threads=1,
                inter_threads=1,
                input_dir='/path/to/mseed',
                output_dir='/path/to/outputs',
                log_filepath='/path/to/outputs/eqcctpro.log',
                csv_dir='/path/to/csv',
                P_threshold=0.001,
                S_threshold=0.02,
                p_model_filepath='/path/to/model_p.h5',
                s_model_filepath='/path/to/model_s.h5',
                cpu_id_list=[0,1],
                set_vram_mb=24750,
                selected_gpus=[0],
                stations2use=2
)
eval_gpu.evaluate()

from eqcctpro import EvaluateSystem

eval_cpu = EvaluateSystem(
                mode='cpu',
                intra_threads=1,
                inter_threads=1,
                input_dir='/path/to/mseed',
                output_dir='/path/to/outputs',
                log_filepath='/path/to/outputs/eqcctpro.log',
                csv_dir='/path/to/csv',
                P_threshold=0.001,
                S_threshold=0.02,
                p_model_filepath='/path/to/model_p.h5',
                s_model_filepath='/path/to/model_s.h5',
                cpu_id_list=range(0,49), 
                min_cpu_amount=20,
                cpu_test_step_size=1, 
                stations2use=50,
                starting_amount_of_stations=25, 
                station_list_step_size=1,
                min_conc_stations=25,
                conc_station_tasks_step_size=5,
                start_time='2024-12-15 12:00:00',
                end_time='2024-12-15 14:00:00',
                conc_timechunk_tasks_step_size=1,
                timechunk_dt=30, 
                waveform_overlap=2,
                tmp_dir=tmp_dir)
eval_cpu.evaluate()

EvaluateSystem will iterate through different combinations of CPU(s), Concurrent Timechunk and Station Tasks, as well as GPU(s), and the amount of VRAM (MB) each Concurrent Prediction can use. EvaluateSystem will take time, depending on the number of CPU/GPUs, the amount of VRAM available, and the total workload that needs to be tested. However, after doing the testing once for most if not all usecases, the trial data will be available and can be used to identify the optimal input parallelization configurations for EQCCTMSeedRunner to use to get the maximum amount of processing out of your system in the shortest amonut of time.

The following input parameters need to be configurated for EvaluateSystem to evaluate your system based on your desired utilization of EQCCTPro:

• eval_mode (str)
  • Can be either cpu or gpu
  • Tells EvaluateSystem which computing approach the trials should it iterate with
• intra_threads (int): default = 1
  • Controls how many intra-parallelism threads Tensorflow can use
• inter_threads (int): default = 1
  • Controls how many inter-parallelism threads Tensorflow can use
• input_dir (str)
  • Directory path to the the mSEED directory
  • EX. /home/skevofilaxc/my_work_directory/eqcct/eqcctpro/230_stations_1_min_dt
• output_dir (str)
  • Directory path to where the output picks and logs will be sent
  • Doesn't need to exist, will be created if doesn't exist
  • Recommended to be in the same working directory as the input directory for convience
• log_filepath (str)
  • Filepath to where the EQCCTPro log will be written to and stored
  • Doesn't need to exist, will be created if doesn't exist
  • Recommended to be in the output directory and called eqcctpro.log, however the name can be changed for your own purposes
• csv_dir (str)
  • Directory path where the CSV's outputted by EvaluateSystem will be saved
  • Doesn't need to exist, will be created if doesn't exist
• P_threshold (float): default = 0.001
  • Threshold in which the P probabilities above it will be considered as P arrival
• S_threshold (float): default = 0.02
  • Threshold in which the S probabilities above it will be considered as S arrival
• p_model_filepath (str)
  • Filepath to where the P EQCCT detection model is stored
• s_model_filepath (str)
  • Filepath to where the S EQCCT detection model is stored
• cpu_id_list (list): default = [1]
  • List that defines which specific CPU cores that sched_setaffinity will allocate for executing the current EQCCTPro process and is the maximum amount of cores EvaluteSystem can use in its trial iterations
  • Allows for specific allocation and limitation of CPUs for a given EQCCTPro process
    • "I want this program to run only on these specific cores."
  • Must be at least 1 CPU if using GPUs (Ray needs CPUs to manage the Raylets (concurrent tasks), however the processing of the waveform is done on the GPU)
• min_cpu_amount (int): default = 1
  • Is the minimum amount of CPUs you want to start your trials with
  • By default, trials will start iterating with 1 CPU up to the maximum allocated
  • Can now set a value as the starting point, such as 15 CPUs up to the maximum of for instance 25
• cpu_test_step_size: default = 1
  • Is the desired step size for the trials will march from min_cpu_amount to len(cpu_id_list)
• stations2use (int): default = None
  • Controls the maximum amount of stations EvaluateSystem can use in its trial iterations
  • Sample data has been provided so that the maximum is 50, however, if using custom data, configure for your specific usecase
• starting_amount_of_stations (int): default = 1
  • For evaluating your system, you have the option to set a starting amount of stations you want to use in the test
  • By default, the test will start using 1 station but now is configurable
• station_list_step_size (int): default = 1
  • You can set a step size for the station list that is generated
  • For example if the stepsize is set to 10 and you start with 50 stations with a max of 100, then your list would be: [50, 60, 70, 80, 80, 100]
  • Using 1 will use the default step size of 1-10, then step size of 5 up to stations2use
• min_conc_stations (int): default = 1
  • Is the minimum amount of concurrent stations predictions you want each trial iteration to start with
  • By default, if min_conc_predictions and conc_predictions_step_size are set to 1, a custom step size iteration will be applied to test the 50 sample waveforms. The sequence follows: [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 15, 20, n+5, 50].
• conc_station_tasks_step_size (int): default = 1
  • Is the concurrent station predictions step size you want each trial iteration to iterate with
  • By default, if min_conc_predictions and conc_predictions_step_size are set to 1, a custom step size iteration will be applied to test the 50 sample waveforms. The sequence follows: [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 15, 20, n+5, 50]
• start_time (str): default = None
  • The start time of the area of time that is being analyzed
  • EX. 2024-12-14 12:00:00
  • Must follow the following convention YYYY-MO-DA HR:MI:SC
  • Used to create a list of defined timechunks from the defined analysis timeframe
  • Must be the exact start time of the analysis time period (does not include the prior waveform overlap time IE. 2024-12-15 11:58:00 for a 2 minute waveform overlap time)
  • Also used in the EvaluateSystem() class to help users note the analysis timeframe in the results CSV file for future result review
• end_time (str): default = None
  • The end time of the area of time that is being analyzed
  • EX. 2024-12-15 12:00:00
  • Must follow the following convention YYYY-MO-DA HR:MI:SC
  • Used to create a list of defined timechunks from the defined analysis timeframe
  • Must be the exact end time of the analysis time period
  • Also used in the EvaluateSystem() class to help users note the analysis timeframe in the results CSV file for future result review
• conc_timechunk_tasks_step_size (int): default = 1
  • Is the concurrent timechunk predictions step size you want each trial iteration to iterate with
• timechunk_dt (int): default = None
  • The length each time chunk is (in minutes)
  • EX. timechunk_dt = 10 and the analysis period is 30 minutes, then three 10-minute long timechunks will be created
• waveform_overlap (int): default = None
  • The duration (in minutes) for which each waveform oself.start_timeverlaps with the others
• tmp_dir (str): default = 1
  • A temporary directory to store all temp files produced by EQCCTPro
  • Used to help ease system cleanup and to not write to system's default temporary directory
• set_vram_mb (float)
  • Maximum amount of VRAM each Raylet can use (float).
  • Must be a real value that is based on your GPU's physical VRAM space, if it exceeds the space the code will break due to OutOfMemoryError
  • Good rule of thumb for calculating set_vram_mb = (GPU VRAM * .90 (to be safe)) / number_of_concurrent_station_predictions * number_of_concurrent_timechunk_predictions
• selected_gpus (list): default = None
  • List of GPU IDs on your computer you want to use if mode = 'gpu'
  • Non-existing GPU IDs will cause the code to exit

---

Finding Optimal CPU/GPU Configurations

After running EvalutateSystem, you can use either the OptimalCPUConfigurationFinder or the OptimalGPUConfigurationFinder determine the best CPU or GPU configurations (respectively) for your specific usecase:

from eqcctpro import OptimalCPUConfigurationFinder, OptimalGPUConfigurationFinder

csv_filepath = '/path/to/csv'

cpu_finder = OptimalCPUConfigurationFinder(csv_filepath)
best_cpu_config = cpu_finder.find_best_overall_usecase()
print(best_cpu_config)

optimal_cpu_config = cpu_finder.find_optimal_for(cpu=3, station_count=2)
print(optimal_cpu_config)

gpu_finder = OptimalGPUConfigurationFinder(csv_filepath)
best_gpu_config = gpu_finder.find_best_overall_usecase()
print(best_gpu_config)

optimal_gpu_config = gpu_finder.find_optimal_for(num_cpus=1, gpu_list=[0], station_count=1)
print(optimal_gpu_config)

Both OptimalCPUConfigurationFinder and OptimalGPUConfigurationFinder each have two usecases:

1. find_best_overall_usecase

• Returns the best overall usecase configuration
  • Uses middle 50% of CPUs for moderate, balanced CPU usage, with the maximum amount of stations processed with the minimum runtime

2. find_optimal_for

• Return the paralleliztion configurations (EX. concurrent predictions, intra/inter thread counts, vram, etc.) for a given number of CPU(s)/GPU(s) and stations
  • Enables users to quickly identify which input parameters should be used for the given amount of resources and workload they have for the minimum runtime possible on their computer

A input CSV directory path must be passed for the classes to use as a reference point:

• csv_filepath (str)
  • Directory path where the CSV's outputted by EvaluateSystem are

Using OptimalCPUConfigurationFinder.find_best_overall_usecase(), no input parameters are needed. It will return back the best usecase parameters.

For OptimalCPUConfigurationFinder.find_optimal_for(), the function requires two input parameters:

• cpu (int)
  • The number of CPU(s) you want to use in your application
• station_count (int)
  • The number of station(s) you want to use in your application

OptimalCPUConfigurationFinder.find_optimal_for() will return back a trial data point containing the mimimum runtime based on your input paramters

Similar to OptimalCPUConfigurationFinder.find_best_overall_usecase(), OptimalGPUConfigurationFinder.find_best_overall_usecase() will return back the best usecase parameters and no input parameters are needed.

For OptimalGPUConfigurationFinder.find_optimal_for(), the function requires three input parameters:

• cpu (int)
  • The number of CPU(s) you want to use in your application
• gpu_list (list)
  • The specific GPU ID(s) you want to use in your application
  • Useful if you have multiple GPUs available and want to use/dedicate a specific one to using EQCCTPro
• station_count (int)
  • The number of station(s) you want to use in your application

Configuration

The environment.yml file specifies the dependencies required to run EQCCTPro. Ensure you have the correct versions installed by using the provided conda environment setup.

License

EQCCTPro is provided under an open-source license. See LICENSE for details.

Contact

For inquiries or issues, please contact [e-mail] or [e-mail].