Skip to content

Cnn training and binning#1

Open
catr1na wants to merge 76 commits into
SBC-Collaboration:masterfrom
catr1na:cnn-rebinning
Open

Cnn training and binning#1
catr1na wants to merge 76 commits into
SBC-Collaboration:masterfrom
catr1na:cnn-rebinning

Conversation

@catr1na

@catr1na catr1na commented Jul 30, 2025

Copy link
Copy Markdown

No description provided.

cericdahl and others added 30 commits June 24, 2025 16:03
changed brackets and spacing in // Apply softmax 
printf("DEBUG: Forward pass completed\n");
return model.dense_output_2;
}   // ← this closing brace is critical
Optimized conv2dforward for timing
deleted second useless conv2dforward function
Changed conv2d forward
changed back to before
going back to earlier, faster version
changed: max_pool2d_forward: spatial-major ordering
changed: conv2d_forward: filter-major ordering
- restrict keyword for pointer aliasing for conv2d forward
static void conv2d_forward(
    float * restrict output,
    const float * restrict input,
- unrolled innermost channel loop of conv2d forward
changed makefile as well: -03 -march=native -ffast=math
catr1na added 30 commits July 22, 2025 11:30
changed 513 to 17
Added new function  create_log_binning_ranges and made new log_scale_spectrogram_c function. Commented out the original log_scale_spectrogram_c function.
changeed filter numbers
x2 for each filter number
nvm last one was wrong
sanity check bak to nromal
changed back to smaller for testing
getting bigger filters for testing again
back to nromal 64
all the changes are functions save_spectrogram_csv
not sur eif this is facts
Lots of this code is still commented out and replaced with dummy variables.
Lots of this code is still commented out and/or using dummy variables
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants