Repository navigation
Bind STDP to graph Affine/Linear weights; refresh README training scope - #50
Merged
Merged
Conversation
…pe (#48, #49) Add nir_stdp_graph_bind/step/free (stdp_graph.h): snapshots a named node's [out][in] weights into an owned STDP instance, transposing across the pre-major/out-major layout gap, and writes each step back through nir_graph_set_weights. Conv2d/Conv1d fail with a named error pending a flattening decision; the header documents the on-chip local-learning flow. test_stdp_graph covers the Affine round-trip (weight delta plus changed checkpoint bytes), bit-exact equivalence with standalone nir_stdp_step, Linear binding, rejects, and non-finite spike handling. README Scope now marks training IO (rng/encode/data/checkpoint) done and Out of scope links the remaining training follow-ups (#39-#47).
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Closes #48. Closes #49.
#48: new
stdp_graph.hadapter (nir_stdp_graph_bind/step/free). It snapshots a named Affine/Linear node's[out][in]weights into an owned STDP instance, transposing across the pre-major/out-major layout gap that makes manual get/set wiring silently wrong, and writes every step back throughnir_graph_set_weights(biases untouched). Conv2d/Conv1d fail with a named error pending a flattening decision. The header documents the intended on-chip local-learning flow (runtime run -> record pre/post spikes -> step -> checkpoint save).New
test_stdp_graphcovers: Affine round-trip with closed-form weight delta plus changed checkpoint bytes, bit-exact equivalence with standalonenir_stdp_step(weights and traces) over 6 spike steps, Linear binding, rejects (unknown/unweighted/Conv2d/NULL/bad-hyperparam), and non-finite spike refusal with unchanged weights.#49: README Scope marks training IO done (
rng.h,encode.h,data.h,checkpoint.h) and Out of scope enumerates the actual remaining gaps with links (#39-#47).Local verification: full ctest 36/36 in Debug and under ASan/UBSan, warning-clean (
-Wall -Wextra -Werror).