feat(model-ports): add openelm_1_1b, olmo_1b, granite_3_1b_a400m - #203
Open
DarthCeltic wants to merge 4 commits into
Open
feat(model-ports): add openelm_1_1b, olmo_1b, granite_3_1b_a400m#203DarthCeltic wants to merge 4 commits into
DarthCeltic wants to merge 4 commits into
Conversation
New batch, opened directly in the compliant claim shape (new standalone ported_models/<model> root + submissions/model_ports/<model>.json, no shared-file edits) per docs/MODEL_PORTS_TRACK_PLAN.md. All 3 confirmed as genuinely distinct GGUF-reported execution families via local metadata inspection (gguf.GGUFReader): arch=openelm, arch=olmo (distinct from olmo2, already claimed elsewhere), and arch=granitemoe (distinct from dense granite, already claimed elsewhere) -- none are seed families, none collide with each other. Verification bar for this batch is GGUF-metadata-level only (file downloaded, sha256 verified, architecture + tensor count confirmed) -- not a full ET sysemu load/offload test, given the scale of this porting round. Each recipe says so explicitly. granite_3_1b_a400m's recipe additionally flags MoE routing (MUL_MAT_ID) as genuinely unconfirmed against the real ET backend, not just a formality.
github-actions
Bot
added
track: model-ports
… in the shared framework; add real host PPL Backfills PR Checklist item 2 (existing-framework registration: artifacts.json, benchmarks/<model>.json, benchmark_config.json entry) that the earlier claim-only commit skipped -- these models are now board-testable through ordinary CI, independent of model-ports track approval. Also replaces "GGUF-metadata-level only" verification with real host reference: built a plain CPU-only (GGML_ET=OFF) llama-perplexity from the same vendored source (the ET-linked build crashes at startup outside the real hackathon runner -- hard dependency on /opt/et/lib/esperanto-fw sysemu firmware files that don't exist here), ran it against the board-pinned WikiText-2 corpus with the board's own parameters. All 3 load and score cleanly: openelm 14.4998, olmo 11.5101, granitemoe 5.7635. granitemoe additionally confirms MoE routing (MUL_MAT_ID) works on ggml-cpu -- first MoE model actually exercised this campaign, though ET-backend support specifically remains unconfirmed.
…1b_a400m, olmo_1b, openelm_1_1b Each claim's benchmark_config must exactly match the model's "config" pointer in .github/ci/benchmark_config.json per effective_model_config() in model_port_claim.py, not the top-level benchmark_config.json path itself. Fixes a mismatch that would fail eligibility validation.
Member
|
For OpenELM, OLMo and Granite model-port credit, please add committed deterministic oracle evidence and successful full-offload ET-SoC1 runs for all three. Host loading/PPL alone is insufficient. Granite overlaps #197, so only the first final qualifying merge for that execution family can count. Maintainers will adopt approved identities/contracts after review. |
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.
Adds openelm_1_1b, olmo_1b, and granite_3_1b_a400m model-ports claims and framework registration.