Feat/backend dispatch mode#35
Merged
Merged
Conversation
Written SUB, SQUARE, MEAN
…ve that in some time Auto grad almost complete
FAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA AUTOGRAD WORKS 4NN XOR WORKS AND LIFE IS BEST
…h will need to see
…was some config and default return error
VishalVerma249
approved these changes
Apr 19, 2026
souls-syntax
added a commit
that referenced
this pull request
Apr 20, 2026
* Added time field in make.sh * chore/ Running summarizer * Completed MSE, passed smoke tests Written SUB, SQUARE, MEAN * chore/ Running summarizer * made assign grad memory function * chore/ Running summarizer * chore/ Running summarizer * chore/ Running summarizer * Somewhat work not tested * Implemented brackprop_b.cpp and added switch statements for op handling * Corrected the enum missnaming mistake * BROKEN : THE FUCKING NN IS NOT LEARNING WRRRRRYYYYYYYYY FUCK * BROKEN 2 : THE FUCKING NN IS NOT LEARNING WRRRRRYYYYYYYYY FUCK * Working operations there is still problem with routing but will resolve that in some time Auto grad almost complete * FAAAAAAAAAAAAAAAAAAAAAAAA: I HAVE COMPLETED THE AUTOGRAD FAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA AUTOGRAD WORKS 4NN XOR WORKS AND LIFE IS BEST * GPU fallback is working but i feel something is wrong with hybrid path will need to see * Broken push don't try it it's super broken likely * CORRECTED: The backend switch and GPU CPU thrashing was resolved, it was some config and default return error --------- Co-authored-by: Aakarsh Kashyap <souls-syntax@gmail.com>
VishalVerma249
pushed a commit
that referenced
this pull request
Apr 21, 2026
* Added time field in make.sh * chore/ Running summarizer * Completed MSE, passed smoke tests Written SUB, SQUARE, MEAN * chore/ Running summarizer * made assign grad memory function * chore/ Running summarizer * chore/ Running summarizer * chore/ Running summarizer * Somewhat work not tested * Implemented brackprop_b.cpp and added switch statements for op handling * Corrected the enum missnaming mistake * BROKEN : THE FUCKING NN IS NOT LEARNING WRRRRRYYYYYYYYY FUCK * BROKEN 2 : THE FUCKING NN IS NOT LEARNING WRRRRRYYYYYYYYY FUCK * Working operations there is still problem with routing but will resolve that in some time Auto grad almost complete * FAAAAAAAAAAAAAAAAAAAAAAAA: I HAVE COMPLETED THE AUTOGRAD FAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA AUTOGRAD WORKS 4NN XOR WORKS AND LIFE IS BEST * GPU fallback is working but i feel something is wrong with hybrid path will need to see * Broken push don't try it it's super broken likely * CORRECTED: The backend switch and GPU CPU thrashing was resolved, it was some config and default return error * Some works and config fallbakc * Feat/backend dispatch mode (#35) * Added time field in make.sh * chore/ Running summarizer * Completed MSE, passed smoke tests Written SUB, SQUARE, MEAN * chore/ Running summarizer * made assign grad memory function * chore/ Running summarizer * chore/ Running summarizer * chore/ Running summarizer * Somewhat work not tested * Implemented brackprop_b.cpp and added switch statements for op handling * Corrected the enum missnaming mistake * BROKEN : THE FUCKING NN IS NOT LEARNING WRRRRRYYYYYYYYY FUCK * BROKEN 2 : THE FUCKING NN IS NOT LEARNING WRRRRRYYYYYYYYY FUCK * Working operations there is still problem with routing but will resolve that in some time Auto grad almost complete * FAAAAAAAAAAAAAAAAAAAAAAAA: I HAVE COMPLETED THE AUTOGRAD FAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA AUTOGRAD WORKS 4NN XOR WORKS AND LIFE IS BEST * GPU fallback is working but i feel something is wrong with hybrid path will need to see * Broken push don't try it it's super broken likely * CORRECTED: The backend switch and GPU CPU thrashing was resolved, it was some config and default return error --------- Co-authored-by: Aakarsh Kashyap <souls-syntax@gmail.com> * R * feat(python): Add flat-C Python bridge layer (sc_bridge) Introduces include/soft-cuda/python/ with five headers: - soft_cuda_python.h : master include - tensor_pool.h : sc_pool_* arena wrappers - tensor_core.h : sc_tensor_* lifecycle + SC_DTYPE_* constants - tensor_ops.h : sc_tensor_mul/add/relu/... + sc_tensor_evaluate* - tensor_graph.h : sc_graph_t opaque handle, Layer 1 + Layer 2 API - tensor_io.h : sc_save_model / sc_load_model Adds src/python/sc_bridge.cpp — single C++ TU that implements all sc_* symbols with extern C linkage, wrapping std::vector via placement-new in sc_graph_t so Python ctypes/cffi never sees any C++ types. Updates CMakeLists.txt: - Excludes src/python/ from the core GLOB (prevents duplicate symbols) - Adds soft_cuda_python SHARED target linked against soft_lib * fix: Remove stray conflict markers from assignBackend.cu * docs: Rewrite main.cpp as sc_* API tutorial, add PYTHON_BRIDGE.md main.cpp: - Complete rewrite using only sc_* bridge functions - 9 numbered sections: pools, data, weights, lazy ops, build_graph, training loop, inference, save/load, cleanup - Pure C (no iostream, no std::vector visible) - Demonstrates both Layer 2 convenience API (sc_build_graph, sc_graph_step) and data readback (sc_tensor_get_data) docs/PYTHON_BRIDGE.md: - Architecture diagram (Python → sc_bridge → soft_lib) - Build instructions (cmake targets) - Full API reference for all 5 headers with function signatures - Complete Python ctypes example (XOR training from Python) - Memory model explanation (arena/bump allocation) - Design decision rationale (why placement-new, why SC_DTYPE, etc.) * Some author defination * Some resolution as well as hoping to correct the API.h in next sitting as there is some graurd resoltion i.e. not happening i think i can do it like this * Working: Corrected the linker error --------- Co-authored-by: Aakarsh Kashyap <souls-syntax@gmail.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
We are implementing Backrpop and cpu fallback and the correction of one way transfer logic