ggerganov/llama.cpp

Pull Requests Commits
ggml : update get_rows f16 and q
slaren committed 2 years ago
2e4db482
ggml : get_rows : support non-contiguos tensors with gaps, generalize up to 3D
slaren committed 2 years ago
ac3f7d8e
metal : reduce the kernel launches for ggml_mul_mat_id
ggerganov committed 2 years ago
Verified 8c5b66ee
metal : add/mul/div use general kernel when src1 not cont
ggerganov committed 2 years ago
Verified 7e2006b0
llama : add basic support for offloading moe with CUDA
slaren committed 2 years ago
06dfde3e
metal : add more general support for ggml_get_rows + tests
ggerganov committed 2 years ago
Verified 2cbcba82
ggml : fix ggml_get_rows to take into account ne02 / ne11
ggerganov committed 2 years ago
Verified 9064b1ca
ggml : add n_as argument to ggml_mul_mat_id
slaren committed 2 years ago
ee8fb399
ggml : ggml_get_rows support 2D indexing [n_tokens, n_experts] (cpu only)
ggerganov committed 2 years ago
Verified 7372b622
llama : fix expert weighting in the FFN
ggerganov committed 2 years ago
Verified 8b185b70
llama : first working version
ggerganov committed 2 years ago
Verified 7ea36953
llama : fix cur -> cur_expert
ggerganov committed 2 years ago
Verified af1a096b
llama : update graph to support MoE
ggerganov committed 2 years ago
Verified aedfad12
ggml : sync latest ggml_mul_mat_id
ggerganov committed 2 years ago
Verified 861cd678
llama : model loading
ggerganov committed 2 years ago
Verified a3eefe95
convert : fix n_ff typo
ggerganov committed 2 years ago
Verified d38e41ee
convert : support Mixtral as LLAMA arch
ggerganov committed 2 years ago
Verified dff8cbeb
sync : ggml (new ops, tests, backend, etc.) (#4359)
ggerganov committed 2 years ago
Verified fe680e3d
llama : per-layer KV cache + quantum K cache (#4309)
ggerganov committed 2 years ago
Verified bcc0eb45
train : fix #4227 (double free in examples/train-text-from-scratch/train-text-from-scratch.cpp) (#4351)
casavaca committed 2 years ago
Verified 81bc9214
server : recognize cache_prompt parameter in OAI API (#4347)
ggerganov committed 2 years ago
Verified 05cd6e50
common : fix compile warning
ggerganov committed 2 years ago
Verified caa92492
speculative : support `--color` (#4343)
stduhpf committed 2 years ago
Verified da5eaef1
grammar : pre-computed pieces + reserve mem + less string copies (#4330)
MarcusDunn committed 2 years ago
Verified 5f6e0c0d
llama : allow overriding GGUF metadata when loading model (#4092)
KerfuffleV2 committed 2 years ago
Verified 5aa365d8
sampling : custom samplers order (#4285)
MaggotHATE committed 2 years ago
Verified 52c8bc3c
swift : revert compiler checks for swift package (#4332)
kchro3 committed 2 years ago
Verified e4b76bbe
simple : update error message for KV cache check (#4324)
danbev committed 2 years ago
Verified 23b5e12e
swift : fix concatenation method to avoid invalid UTF8 stringfication (#4325)
ensan-hcl committed 2 years ago
Verified d208995c
swift : fix prompt tokenization logic (#4321)
ensan-hcl committed 2 years ago
Verified 5c9f90cb
Terms Of Use Privacy Policy Refund Policy Impressum