Modern Bert Support#15641

Open

ryan-mangeno wants to merge 76 commits intoggml-org:masterfrom

ryan-mangeno:modern-bert-support

Commits on Aug 21, 2025

constants and tensor mappings for modern bert support, model not supported yet but working on getting conversion to work for encoder only
ryan-mangeno
committed
conversion now working, hf -> gguf
ryan-mangeno
committed

Commits on Aug 25, 2025

Commits on Aug 26, 2025

Commits on Aug 27, 2025

fixed tensor mappings and working on buildin graph
ryan-mangeno
committed

Commits on Aug 29, 2025

added cls token per previous modern bert attempt, still working on checking out the rest
ryan-mangeno
committed

Commits on Sep 2, 2025

fixed pre tokenizer and still working through previous pr
ryan-mangeno
committed

Commits on Sep 3, 2025

Commits on Sep 8, 2025

Commits on Sep 11, 2025

Commits on Sep 12, 2025

Commits on Sep 13, 2025

Commits on Sep 14, 2025

merge fixes
ryan-mangeno
committed

Commits on Sep 15, 2025

Merge branch 'master' into modern-bert-support
ryan-mangeno
authored

Commits on Sep 17, 2025

Merge branch 'master' into modern-bert-support
ryan-mangeno
authored

Commits on Sep 26, 2025

standard swa method can be used instead of a new enum being LLAMA_SWA_TYPE_LOCAL
ryan-mangeno
committed

Commits on Oct 4, 2025

replaced attn out norm with ffn_norm and cosine similarity between hf embds and llama.cpp embds went way up, from 0.05 to 0.24, replaced the cacheless kv with swa todo per the previous conversion
ryan-mangeno
committed

Commits on Oct 5, 2025

merge
ryan-mangeno
committed

Commits on Oct 10, 2025

Commits on Oct 13, 2025

fixed ordering of enum for freq_base_swa
ryan-mangeno
committed

Commits on Oct 30, 2025

removing whitespace
ryan-mangeno
committed

Commits on Nov 29, 2025