happyz
/
llama.cpp
mirror of https://github.com/ggerganov/llama.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
llama.cpp/src
History
EliteGPT AI 7bab4a3065 model : add Qwen3-Omni multimodal architecture support 
Adds support for Qwen3-Omni, Alibaba's multimodal LLM that handles
text and vision. This enables the main LLM architecture and vision
encoder support.

Main LLM changes:
- Add LLM_ARCH_QWEN3OMNI enum and architecture registration
- Add hparams loading for MoE-based architecture (48 layers, 128 experts)
- Reuse llm_build_qwen3moe graph builder
- Add IMROPE type for multimodal position encoding

Vision encoder changes (via mtmd):
- Add PROJECTOR_TYPE_QWEN3O with auto-conversion to QWEN3VL for vision
- Support different embedding dimensions (vision=8192, audio=2048)
- Add separate Q/K/V tensor support in qwen3vl graph builder

Tested with Qwen3-Omni-30B-Q8_0.gguf on distributed 5-GPU setup:
- 41-44 tokens/sec inference speed
- Text and vision inference working

Note: Audio encoder support is WIP and will follow in a separate PR.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
 		2025-12-31 20:25:55 +10:00
..
models model : Plamo3 support (#17304) 2025-12-28 17:28:31 +01:00
CMakeLists.txt model : Plamo3 support (#17304) 2025-12-28 17:28:31 +01:00
llama-adapter.cpp lora: count lora nodes in graph_max_nodes (#18469) 2025-12-30 15:53:12 +01:00
llama-adapter.h lora: count lora nodes in graph_max_nodes (#18469) 2025-12-30 15:53:12 +01:00
llama-arch.cpp model : add Qwen3-Omni multimodal architecture support 2025-12-31 20:25:55 +10:00
llama-arch.h model : add Qwen3-Omni multimodal architecture support 2025-12-31 20:25:55 +10:00
llama-batch.cpp batch : fix sequence id ownership (#17915) 2025-12-11 14:29:47 +02:00
llama-batch.h batch : fix sequence id ownership (#17915) 2025-12-11 14:29:47 +02:00
llama-chat.cpp model : add openPangu-Embedded (#16941) 2025-11-05 10:28:58 +01:00
llama-chat.h model : add openPangu-Embedded (#16941) 2025-11-05 10:28:58 +01:00
llama-context.cpp lora: count lora nodes in graph_max_nodes (#18469) 2025-12-30 15:53:12 +01:00
llama-context.h llama: automatically set parameters not set by the user in such a way that maximizes GPU utilization (#16653) 2025-12-15 09:24:59 +01:00
llama-cparams.cpp cparams : rename LLAMA_MAX_PARALLEL_SEQUENCES to LLAMA_MAX_SEQ (#14188) 2025-06-15 10:08:58 +03:00
llama-cparams.h server : support unified cache across slots (#16736) 2025-11-02 18:14:04 +02:00
llama-grammar.cpp llama : add token matching support to llama-grammar (#17816) 2025-12-09 00:32:57 -06:00
llama-grammar.h llama : add token matching support to llama-grammar (#17816) 2025-12-09 00:32:57 -06:00
llama-graph.cpp graph : reuse SSM graphs (#16490) 2025-12-16 09:36:21 +02:00
llama-graph.h graph : reuse SSM graphs (#16490) 2025-12-16 09:36:21 +02:00
llama-hparams.cpp model: support GLM4V vision encoder (#18042) 2025-12-16 11:25:26 +01:00
llama-hparams.h model: support MiMo-V2-Flash (#18328) 2025-12-24 23:07:08 +01:00
llama-impl.cpp llama: automatically set parameters not set by the user in such a way that maximizes GPU utilization (#16653) 2025-12-15 09:24:59 +01:00
llama-impl.h ggml, llama : use defaulted constructors/destructors (#17649) 2025-12-03 07:12:18 +01:00
llama-io.cpp llama : refactor llama_context, llama_kv_cache, llm_build_context (#12181) 2025-03-13 12:35:44 +02:00
llama-io.h llama : refactor llama_context, llama_kv_cache, llm_build_context (#12181) 2025-03-13 12:35:44 +02:00
llama-kv-cache-iswa.cpp kv-cache : pad the cache size to 256 for performance (#17046) 2025-11-07 20:03:25 +02:00
llama-kv-cache-iswa.h llama: print memory breakdown on exit (#15860) 2025-09-24 16:53:48 +02:00
llama-kv-cache.cpp kv-cache: Fix state restore fragmented cache (#17982) 2025-12-15 19:28:35 +02:00
llama-kv-cache.h llama : fix typo in comment in llama-kv-cache.h [no ci] (#18489) 2025-12-30 17:20:14 +01:00
llama-kv-cells.h llama: store mrope data in KV cell (#16825) 2025-10-29 18:09:18 +01:00
llama-memory-hybrid.cpp graph : reuse SSM graphs (#16490) 2025-12-16 09:36:21 +02:00
llama-memory-hybrid.h llama: print memory breakdown on exit (#15860) 2025-09-24 16:53:48 +02:00
llama-memory-recurrent.cpp memory: Hybrid context shift (#17009) 2025-11-10 17:14:23 +02:00
llama-memory-recurrent.h llama: consistent ctx <-> buf order for KV cache (#16746) 2025-10-28 11:23:54 +01:00
llama-memory.cpp memory : correctly handle failure in apply() (#14438) 2025-06-30 18:03:03 +03:00
llama-memory.h llama: print memory breakdown on exit (#15860) 2025-09-24 16:53:48 +02:00
llama-mmap.cpp llama : Async DirectIO model loading on Linux (#18012) 2025-12-18 08:27:19 +02:00
llama-mmap.h llama : Async DirectIO model loading on Linux (#18012) 2025-12-18 08:27:19 +02:00
llama-model-loader.cpp model : Granite Embedding support (#15641) 2025-12-23 00:28:19 +01:00
llama-model-loader.h model : Granite Embedding support (#15641) 2025-12-23 00:28:19 +01:00
llama-model-saver.cpp llama : improve sep token handling (#14272) 2025-06-20 14:04:09 +02:00
llama-model-saver.h llama/ggml: add LLM training support (#10544) 2025-05-12 14:44:49 +02:00
llama-model.cpp model : add Qwen3-Omni multimodal architecture support 2025-12-31 20:25:55 +10:00
llama-model.h lora: count lora nodes in graph_max_nodes (#18469) 2025-12-30 15:53:12 +01:00
llama-quant.cpp llama: automatically set parameters not set by the user in such a way that maximizes GPU utilization (#16653) 2025-12-15 09:24:59 +01:00
llama-quant.h llama : refactor `src/llama.cpp` (#10902) 2025-01-03 10:18:53 +02:00
llama-sampling.cpp sampling: reuse token data buffer in llama_sampler_sample (#18365) 2025-12-30 16:27:49 +02:00
llama-sampling.h sampling: reuse token data buffer in llama_sampler_sample (#18365) 2025-12-30 16:27:49 +02:00
llama-vocab.cpp model : Granite Embedding support (#15641) 2025-12-23 00:28:19 +01:00
llama-vocab.h model : add AfmoeForCausalLM support (#16477) 2025-11-14 13:54:10 +01:00
llama.cpp llama-fit-params: fix step size for last device (#18415) 2025-12-28 10:52:09 +01:00
unicode-data.cpp server : better security control for public deployments (#9776) 2024-10-08 13:27:04 +02:00
unicode-data.h llama : reduce compile time and binary size (#9712) 2024-10-02 15:49:55 +02:00
unicode.cpp fix: prevent segfault in tokenizer on highly repetitive input (#17786) 2025-12-05 13:52:23 +02:00
unicode.h devops: add s390x & ppc64le CI (#15925) 2025-09-27 02:03:33 +08:00