Repository navigation

[Do Not Merge] model : LFM2.5-Audio-1.5B - #18641

Closed

tdakhran wants to merge 54 commits into

ggml-org:masterfrom

tdakhran:tarek/feat/os-lfm2.5-audio-1.5b-upstream

Closed

[Do Not Merge] model : LFM2.5-Audio-1.5B#18641
tdakhran wants to merge 54 commits into
ggml-org:masterfrom
tdakhran:tarek/feat/os-lfm2.5-audio-1.5b-upstream

Commits on Jan 6, 2026
- model : add LFM2-ColBert-350M
  tdakhran
  committed
  3b53b5f
- Use n_cls_out for pooling rank
  tdakhran
  committed
  af74ec1
- memory : add llama_memory_hybrid_iswa
  tdakhran
  committed
  888dd47
- Add istft audio utils
  tdakhran
  committed
  40e7a15
- model : add Lfm25AudioTokenizer
  tdakhran
  committed
  4017949
- LFM2.5-Audio-1.5B
  tdakhran
  committed
  8784689
- Small fixes
  tdakhran
  committed
  e1a8fd1
- Remove pimpl from mtmd-audio
  tdakhran
  committed
  ec2890d
- mtmd: mtmd_audio_streaming_istft
  tdakhran
  committed
  b48b93e
- Merge remote-tracking branch 'tdakhran/tarek/dev/istft-upstream' into tarek/feat/os-lfm2.5-audio-1.5b-upstream
  tdakhran
  committed
  07bf242
Commits on Jan 7, 2026
- Read n_layer from gguf [no ci]
  tdakhran
  committed
  b0d4293
- Move save_wav implementation to mtmd [no ci]
  tdakhran
  committed
  4a2f68a
Commits on Jan 14, 2026
- Move implementation to mtmd-audio, tts works
  tdakhran
  committed
  9adc060
- Support interleaved mode
  tdakhran
  committed
  5e81b61
- Fix examples and chat mode
  tdakhran
  committed
  b7ab09f
Commits on Jan 16, 2026
- memory : add llama_memory_hybrid_iswa
  tdakhran
  committed
  88c90a7
- Update src/llama-memory-hybrid-iswa.cpp
  
  tdakhran
  and
  ggerganov
  committed
  99e3092
- Merge branch 'tarek/feat/memory-hybrid-iswa' into tarek/feat/os-lfm2.5-audio-1.5b-upstream
  tdakhran
  committed
  a3e8f98
- mtmd : Fix ASR for LFM2.5-Audio-1.5B
  tdakhran
  committed
  fc3525a
- reset decoder for each chat
  tdakhran
  committed
  6706849
- Move temperature selection to decoder
  tdakhran
  committed
  b5848e1
- audio output works with llama-server
  tdakhran
  committed
  6bb696d
- chat history works
  tdakhran
  committed
  a147d82
- small fixes and refactors
  tdakhran
  committed
  3391aa8
Commits on Jan 21, 2026
- memory : add llama_memory_hybrid_iswa
  tdakhran
  committed
  0c3c1d0
- Update src/llama-memory-hybrid-iswa.cpp
  
  tdakhran
  and
  ggerganov
  committed
  d6a45a4
- Update server-context.cpp
  tdakhran
  committed
  660d88e
- Merge remote-tracking branch 'tdakhran/tarek/feat/memory-hybrid-iswa' into tarek/feat/os-lfm2.5-audio-1.5b-upstream [no ci]
  tdakhran
  committed
  d64c8d4
- Fix cli [no ci]
  tdakhran
  committed
  9960b91
- Merge remote-tracking branch 'upstream/master' into tarek/feat/os-lfm2.5-audio-1.5b-upstream
  tdakhran
  committed
  f270033
- Do not output special tokens
  tdakhran
  committed
  982a33e
Commits on Jan 23, 2026
- Read n_layer from gguf [no ci]
  tdakhran
  committed
  c71f178
Commits on Jan 29, 2026
- Send reset_context for compatibility with llama-liquid-audio-server
  tdakhran
  committed
  e30495d
Commits on Feb 17, 2026
- Rework llama-liquid-audio-server to have single queue
  tdakhran
  committed
  d763051
- Single thread for server to avoid arena growth
  tdakhran
  committed
  ac98513
- Make vocoder and audiotokenizer optional [no ci]
  tdakhran
  committed
  39ff210
- model : Add tokenizer from LFM2.5-Audio-1.5B
  tdakhran
  committed
  79962f6
- Update convert_hf_to_gguf.py
  
  tdakhran
  and
  CISC
  authored
  46a9d9a
- Formatting
  tdakhran
  committed
  755d0ea
- Rework check for attention layers
  tdakhran
  committed
  1ae63be
- Add LFM2 SWA model support
  tdakhran
  committed
  e3ae935
- Merge branch 'tarek/feat/upstream-audio-tokenizer' into tarek/feat/os-lfm2.5-audio-1.5b-upstream
  tdakhran
  committed
  615bd28
- Remove script
  tdakhran
  committed
  9862353
- Address PR feedback
  tdakhran
  committed
  6c41320
- Set vocab to none
  tdakhran
  committed
  7f1114b
- Merge branch 'tarek/feat/upstream-audio-tokenizer' into tarek/feat/os-lfm2.5-audio-1.5b-upstream
  tdakhran
  committed
  51e2e0e
Commits on Feb 18, 2026
- Move helper function definitions to cpp file
  tdakhran
  committed
  567be9c
- Merge branch 'tarek/feat/upstream-audio-tokenizer' into tarek/feat/os-lfm2.5-audio-1.5b-upstream
  tdakhran
  committed
  9c9f4e3
Commits on Mar 3, 2026
- Merge remote-tracking branch 'upstream/master' into tarek/feat/os-lfm2.5-audio-1.5b-upstream [no ci]
  tdakhran
  committed
  006639c
Commits on May 25, 2026
- Read interleaved_n_text/interleaved_n_audio from config [noci]
  tdakhran
  committed
  a5a8737
Commits on May 29, 2026
- Buffer incomplete UTF-8 in server text_cb to avoid JSON crash [noci]
  
  ruke1ire
  authored and
  tdakhran
  committed
  79c5ded
Commits on Jun 24, 2026
- ggml-cpu: fix SVE leftover path in ggml_vec_dot_f32 [no-ci]
  tdakhran
  committed
  a726a57
Commits on Jul 16, 2026
- liquid-audio: commit the previous turn's EOG at the next continuation prefill [no-ci]
  
  authored and
  tdakhran
  committed
  53b3c54
Commits on Sep 1, 2026
- liquid-audio: dump JSON with the replace error handler [no-ci]
  
  ashraf-guitouni
  authored and
  tdakhran
  committed
  ec9d1fd

Commits

Commits on Jan 6, 2026

Commits on Jan 7, 2026

Commits on Jan 14, 2026

Commits on Jan 16, 2026

Commits on Jan 21, 2026

Commits on Jan 23, 2026

Commits on Jan 29, 2026

Commits on Feb 17, 2026

Commits on Feb 18, 2026

Commits on Mar 3, 2026

Commits on May 25, 2026

Commits on May 29, 2026

Commits on Jun 24, 2026

Commits on Jul 16, 2026

Commits on Sep 1, 2026