Quick Start: Run astroflow on FRB180417¶

1) Get the test file¶

Option A: Zenodo (recommended)

wget -O FRB180417.fil "https://zenodo.org/record/3905426/files/FRB180417.fil"

Option B: GitHub raw (alternate mirror)

curl -L -o FRB180417.fil   "https://raw.githubusercontent.com/lintian233/astroflow/main/tests/FRB180417.fil"

Quick check:

ls -lh FRB180417.fil

-rw-r--r-- 1 root root 1.7M Aug 26 20:12 FRB180417.fil

2) Create a run config¶

Save as frb180417.yaml:

# frb180417.yaml

input: FRB180417.fil      # Input file path (case-sensitive)
output: frb180417         # Output directory prefix (plots & logs go here)

mode: single              # One of [single, muti, dataset, monitor]

timedownfactor: 1         # Downsample factor for time axis
confidence: 0.4           # Detector confidence threshold

dedgpu: 0                 # GPU index for dedispersion
detgpu: 0                 # GPU index for detection
cputhread: 8              # CPU workers (I/O, prep)

plotworker: 1             # Plotting workers

modelname: yolov11n       # Detection model; currently only yolov11n is supported

# maskfile: none.txt      # RFI mask file (or  disable)
gencand: true             # generate candidate info csv file

tsample:
  - name: t1
    t: 0.5                 # seconds per detection slice 
    # (pulse_width ~ 1-10 ms 0.5s is recommended )
    # (pulsewith ~ 0.1-0.5 ms <0.05s is is recommended )


# human-readable label for DM limits
dm_limt:
  - name: limt1
    dm_low: 0
    dm_high: 800

# Actual DM search grid FRB181407 DM is 474.8 
dmrange:
  - name: dm_10_600
    dm_low: 10
    dm_high: 600
    dm_step: 1  #DM trails > 500 is strong recommended

# Frequency window (MHz)
freqrange:
  - name: fullband
    freq_start: 1130
    freq_end: 1465
#   - name: lowband
#     freq_start: 1130
#     freq_end: 1250


# DMT (DM–time) plot scaling
dmtconfig:
  minpercentile: 0
  maxpercentile: 99.99

# Dynamic spectrum plot scaling
specconfig:
  minpercentile: 0.1
  maxpercentile: 99.9
  tband: 120                # time bands
  mode: subband             # one of [subband, standard, detrend]
  dtrend: true             # optional: linear detrending per subband (default false)
  norm: false                # optional: normalize each subband (default true)
  subfreq: 128              # optional: subbands in subband mode (default 128)
  subtsamp: 2               # optional: time binning factor (default 4)
  savetype: png             # optional: image format png/jpg

Tip: subfreq, subtsamp, dtrend, norm, and savetype are optional. Remove any line to fall back to the default behaviour shown in the comment.

Notes

If you run on a multi-GPU box, switch dedgpu/detgpu to the indices you want (or use CUDA_VISIBLE_DEVICES).

3) Run¶

astroflow frb180417.yaml

You should see logs like:

❯ astroflow ./frb20180417.yaml 
Detected mode: single
Starting single file search for: ....
[TIMER] IO : 0.001 seconds, 1387.844 MB/s
[INFO] Using device 0: ....
[INFO] RFICONFIG use_iqrm=0, use_zero_dm=0, use_mask=0
....
[INFO]: INPUT SHAPE = [591, 5120], T: 6.484 s, SDM1: 591, ST2: 394, N: 13
....
Saving: 9.07_2.86_473.591_2.042_FRB180417_T_2.0s_2.5s_DM_10_600_F_1130_1465.png

4) View the results¶

Images are written under the output prefix, e.g.:

ls frb180417/detect/*.png 

display frb180417/detect/*.png

You should see png like:

5) Troubleshooting¶

Model Download Failure¶

astroflow will automatically download the required AI model on its first run. If you see an error message like the one below, it means the automatic download failed after several attempts, likely due to network issues or file corruption.

Warning: Model file checksum mismatch on attempt 1/3. Expected d4305e273fec6f5733f3c0a823fa5275064b015d549fda26529e0b1b8f59c124, but got xxxxxxx. The file is corrupted and will be re-downloaded.
.......
Error: Failed to download model from https://github.com/lintian233/astroflow/releases/download/v0.1.1/yolo11n_0816_v1.pt after 3 attempts.

Solution: Manual Download

You can manually download the model and place it in the correct directory.

Create the configuration directory if it doesn’t exist:
```
mkdir -p ~/.config/astroflow
```
Download the model file using wget or curl:
```
wget -O ~/.config/astroflow/yolo11n_0816_v1.pt "https://github.com/lintian233/astroflow/releases/download/v0.1.1/yolo11n_0816_v1.pt"
```
Alternatively, you can download it from your browser and move it to ~/.config/astroflow/yolo11n_0816_v1.pt.

AstroFlow

Navigation

Related Topics