feat(model-ports): add mamba_1_4b, chatglm3_6b - #207
Open
DarthCeltic wants to merge 3 commits into
Open
Conversation
…sive (no claim) mamba_1_4b and chatglm3_6b: full compliant shape (claim + framework registration + real host-CPU perplexity). Built a plain CPU-only (GGML_ET=OFF) llama-perplexity from the vendored source, ran against the board-pinned WikiText-2 corpus. Both load and score cleanly: mamba 11.7985 (confirms SSM_CONV/SSM_SCAN work on ggml-cpu -- the first non-attention architecture in this campaign; ET-backend support for these ops is a separate, genuinely open question), chatglm3 15.9100. flan_t5_base loads and builds its compute graph, but neither llama-perplexity nor llama-cli in this vendored submodule can drive it to produce output: perplexity asserts !add_eos (T5's tokenizer always appends EOS), and llama-cli calls llama_decode() without ever calling llama_encode() first (required for T5's cross-attention). Both are tool/harness gaps, not model defects -- no claim filed, documented as an inconclusive result rather than overstating what was shown. mamba_1_4b's license is unspecified in the upstream HF repo metadata -- flagged as "unspecified" rather than assumed.
github-actions
Bot
added
track: model-ports
DarthCeltic
added a commit
to DarthCeltic/hf-hackathon
that referenced
this pull request
Jul 25, 2026
…eal measured PPL, see PR aifoundry-org#207)
…b, mamba_1_4b 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 Mamba and ChatGLM3 model-port credit, please add committed deterministic oracle evidence and successful full-offload ET-SoC1 runs. Flan-T5 currently has only a recipe and no registered benchmark/claim; add those if it is intended to count. Maintainers will adopt reviewed identities/contracts after the evidence passes. |
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 mamba_1_4b and chatglm3_6b causal-LM ports to llama.cpp-et, with model-ports claims for each. flan_t5_base recipe included as a documented inconclusive result (loads, but neither llama-perplexity nor llama-cli in this vendored submodule can drive an encoder-decoder model); no claim filed for it.