| .. |
|
models
|
a5eaa1d6a3
mla : make the V tensor a view of K (#18986)
|
6 days ago |
|
CMakeLists.txt
|
ad8d85bd94
memory : add llama_memory_hybrid_iswa (#18601)
|
1 week ago |
|
llama-adapter.cpp
|
a7e6ddb8bd
lora: make sure model keep track of associated adapters (#18490)
|
2 weeks ago |
|
llama-adapter.h
|
a7e6ddb8bd
lora: make sure model keep track of associated adapters (#18490)
|
2 weeks ago |
|
llama-arch.cpp
|
60591f01d4
model : add EXAONE MoE (#18543)
|
2 weeks ago |
|
llama-arch.h
|
60591f01d4
model : add EXAONE MoE (#18543)
|
2 weeks ago |
|
llama-batch.cpp
|
d9f8f60618
batch : fix sequence id ownership (#17915)
|
1 month ago |
|
llama-batch.h
|
d9f8f60618
batch : fix sequence id ownership (#17915)
|
1 month ago |
|
llama-chat.cpp
|
60591f01d4
model : add EXAONE MoE (#18543)
|
2 weeks ago |
|
llama-chat.h
|
60591f01d4
model : add EXAONE MoE (#18543)
|
2 weeks ago |
|
llama-context.cpp
|
be8e3d9515
context : do not reserve scheduler for warmups (#18867)
|
2 weeks ago |
|
llama-context.h
|
39173bcacb
context : reserve new scheduler when graph topology changes (#18547)
|
2 weeks ago |
|
llama-cparams.cpp
|
c311ac664d
cparams : rename LLAMA_MAX_PARALLEL_SEQUENCES to LLAMA_MAX_SEQ (#14188)
|
7 months ago |
|
llama-cparams.h
|
39173bcacb
context : reserve new scheduler when graph topology changes (#18547)
|
2 weeks ago |
|
llama-grammar.cpp
|
cef1d23c5a
common/grammar : replace problematic backtracking regex `[\s\S]*` (#18342)
|
3 weeks ago |
|
llama-grammar.h
|
cef1d23c5a
common/grammar : replace problematic backtracking regex `[\s\S]*` (#18342)
|
3 weeks ago |
|
llama-graph.cpp
|
a5eaa1d6a3
mla : make the V tensor a view of K (#18986)
|
6 days ago |
|
llama-graph.h
|
ad8d85bd94
memory : add llama_memory_hybrid_iswa (#18601)
|
1 week ago |
|
llama-hparams.cpp
|
2fbde785bc
kv-cache : optimize KQ mask construction (#18842)
|
1 week ago |
|
llama-hparams.h
|
2fbde785bc
kv-cache : optimize KQ mask construction (#18842)
|
1 week ago |
|
llama-impl.cpp
|
b1f3a6e5db
llama: automatically set parameters not set by the user in such a way that maximizes GPU utilization (#16653)
|
1 month ago |
|
llama-impl.h
|
37adc9c6ba
ggml, llama : use defaulted constructors/destructors (#17649)
|
1 month ago |
|
llama-io.cpp
|
e0dbec0bc6
llama : refactor llama_context, llama_kv_cache, llm_build_context (#12181)
|
10 months ago |
|
llama-io.h
|
e0dbec0bc6
llama : refactor llama_context, llama_kv_cache, llm_build_context (#12181)
|
10 months ago |
|
llama-kv-cache-iswa.cpp
|
16bcc1259d
kv-cache : pad the cache size to 256 for performance (#17046)
|
2 months ago |
|
llama-kv-cache-iswa.h
|
e789095502
llama: print memory breakdown on exit (#15860)
|
4 months ago |
|
llama-kv-cache.cpp
|
a5eaa1d6a3
mla : make the V tensor a view of K (#18986)
|
6 days ago |
|
llama-kv-cache.h
|
2fbde785bc
kv-cache : optimize KQ mask construction (#18842)
|
1 week ago |
|
llama-kv-cells.h
|
e3af5563bd
llama: store mrope data in KV cell (#16825)
|
3 months ago |
|
llama-memory-hybrid-iswa.cpp
|
ad8d85bd94
memory : add llama_memory_hybrid_iswa (#18601)
|
1 week ago |
|
llama-memory-hybrid-iswa.h
|
ad8d85bd94
memory : add llama_memory_hybrid_iswa (#18601)
|
1 week ago |
|
llama-memory-hybrid.cpp
|
c560316440
graph : reuse SSM graphs (#16490)
|
1 month ago |
|
llama-memory-hybrid.h
|
e789095502
llama: print memory breakdown on exit (#15860)
|
4 months ago |
|
llama-memory-recurrent.cpp
|
0c74f32632
memory: Hybrid context shift (#17009)
|
2 months ago |
|
llama-memory-recurrent.h
|
7a0e900e36
llama: consistent ctx <-> buf order for KV cache (#16746)
|
3 months ago |
|
llama-memory.cpp
|
745f11fed0
memory : correctly handle failure in apply() (#14438)
|
7 months ago |
|
llama-memory.h
|
e789095502
llama: print memory breakdown on exit (#15860)
|
4 months ago |
|
llama-mmap.cpp
|
287a33017b
llama : Extend fallback, fix fileno for dio file, exclude case that mmap uses dio file (#18887)
|
1 week ago |
|
llama-mmap.h
|
2038101bd9
llama : add `use_direct_io` flag for model loading (#18166)
|
3 weeks ago |
|
llama-model-loader.cpp
|
287a33017b
llama : Extend fallback, fix fileno for dio file, exclude case that mmap uses dio file (#18887)
|
1 week ago |
|
llama-model-loader.h
|
2038101bd9
llama : add `use_direct_io` flag for model loading (#18166)
|
3 weeks ago |
|
llama-model-saver.cpp
|
73d284a250
model : add LFM2-ColBert-350M (#18607)
|
3 weeks ago |
|
llama-model-saver.h
|
10d2af0eaa
llama/ggml: add LLM training support (#10544)
|
8 months ago |
|
llama-model.cpp
|
ad8d85bd94
memory : add llama_memory_hybrid_iswa (#18601)
|
1 week ago |
|
llama-model.h
|
a7e6ddb8bd
lora: make sure model keep track of associated adapters (#18490)
|
2 weeks ago |
|
llama-quant.cpp
|
0e4ebeb057
quant : manual overrides of tensor types take precedence (#18952)
|
1 week ago |
|
llama-quant.h
|
f66f582927
llama : refactor `src/llama.cpp` (#10902)
|
1 year ago |
|
llama-sampling.cpp
|
13f1e4a9ca
llama : add adaptive-p sampler (#17927)
|
2 weeks ago |
|
llama-sampling.h
|
d3dce4e0a5
sampling : add support for backend sampling (#17004)
|
3 weeks ago |
|
llama-vocab.cpp
|
6e36299b47
llama : print_info alignment fix (#18708)
|
2 weeks ago |
|
llama-vocab.h
|
60591f01d4
model : add EXAONE MoE (#18543)
|
2 weeks ago |
|
llama.cpp
|
046d5fd44e
llama: use host memory if device reports 0 memory (#18587)
|
2 weeks ago |
|
unicode-data.cpp
|
458367a906
server : better security control for public deployments (#9776)
|
1 year ago |
|
unicode-data.h
|
a39ab216aa
llama : reduce compile time and binary size (#9712)
|
1 year ago |
|
unicode.cpp
|
ced765be44
model: support youtu-vl model (#18479)
|
4 weeks ago |
|
unicode.h
|
624207e676
devops: add s390x & ppc64le CI (#15925)
|
4 months ago |
Georgi Gerganov