COMMITS
/ examples/speculative/speculative.cpp November 25, 2024
D
llama : accept a list of devices to use to offload a model (#10497)
Diego Devesa committed
G
speculative : refactor and add a simpler example (#10362)
Georgi Gerganov committed
November 14, 2024
G
speculative : fix out-of-bounds access (#10289)
Georgi Gerganov committed
October 21, 2024
G
llama : default sampling changes + greedy update (#9897)
Georgi Gerganov committed
G
speculative : fix handling of some input params (#9963)
Georgi Gerganov committed
October 18, 2024
X
llama : remove all_pos_0, all_pos_1, all_seq_id from llama_batch (#9745)
Xuan Son Nguyen committed
October 10, 2024
D
common : use common_ prefix for common library functions (#9805)
Diego Devesa committed
September 24, 2024
G
sampling : avoid expensive softmax during greedy sampling (#9605)
Georgi Gerganov committed
September 15, 2024
G
common : reimplement logging (#9418)
Georgi Gerganov committed
September 13, 2024
G
llama : llama_perf + option to disable timings during decode (#9355)
Georgi Gerganov committed
September 9, 2024
X
common : move arg parser code to `arg.cpp` (#9388)
Xuan Son Nguyen committed
September 7, 2024
X
common : refactor arg parser (#9308)
Xuan Son Nguyen committed
G
llama : refactor sampling v2 (#9294)
Georgi Gerganov committed
August 29, 2024
F
Threadpool: take 2 (#8672)
Faisal Zaghloul committed
August 5, 2024
L
common : Changed tuple to struct (TODO fix) (#8823)
Liu Jia committed
June 4, 2024
G
common : refactor cli arg parsing (#7675)
Georgi Gerganov committed
April 21, 2024
P
llama : support Llama 3 HF conversion (#6745)
Pedro Cuenca committed
April 9, 2024
J
BERT tokenizer fixes (#6498)
Jared Van Bortel committed
March 26, 2024
C
llama : greatly reduce output buffer memory usage (#6122)
compilade committed
March 24, 2024
M
sampling : deduplicated code for probability distribution access (#6240)
Minsoo Cheong committed
March 5, 2024
J
fix speculative decoding build on windows (#5874)
Jeffrey Quesnelle committed
March 4, 2024
M
speculative : implement stochastic speculative sampling (#5625)
Minsoo Cheong committed
February 16, 2024
B
ggml : add numa options (#5377)
bmwl committed
January 16, 2024
S
speculative : threading options (#4959)
stduhpf committed
December 12, 2023
R
english : use `typos` to fix comments and logs (#4354)
Richard Kiss committed
December 6, 2023
S
speculative : support `--color` (#4343)
stduhpf committed
November 20, 2023
B
speculative : fix prompt tokenization in speculative example (#4025)
Branden Butler committed
November 3, 2023
G
speculative : change default p_accept to 0.5 + CLI args (#3919)
Georgi Gerganov committed
November 2, 2023
C
build : link against build info instead of compiling against it (#3879)
cebtenzzre committed
October 28, 2023
G
llama : add option for greedy sampling with probs (#3813)
Georgi Gerganov committed
October 27, 2023
K
speculative : ensure draft and target model vocab matches (#3812)
Kerfuffle committed
October 23, 2023
M
llama : remove token functions with `context` args in favor of `model` (#3720)
Marcus Dunn committed
October 20, 2023
G
sampling : refactor init to use llama_sampling_params (#3696)
Georgi Gerganov committed
October 18, 2023
G
speculative : bug fixes
Georgi Gerganov committed
G
speculative : add tree-based sampling example (#3624)
Georgi Gerganov committed
October 11, 2023
K
common : fix mirostat state when using multiple sequences (#3543)
Kerfuffle committed
October 3, 2023
G
llama : fix session saving/loading (#3400)
Georgi Gerganov committed
September 28, 2023
S
G
llama : custom attention mask + parallel decoding + no context swaps (#3228)
Georgi Gerganov committed
September 14, 2023
L
speculative : add heuristic algorithm (#3006)
Leng Yue committed
September 13, 2023
F
speculative: add --n-gpu-layers-draft option (#3063)
FK committed
September 8, 2023
P
build : do not use _GNU_SOURCE gratuitously (#2035)
Przemysław Pawełczyk committed
September 5, 2023
G
speculative : add grammar support (#2991)
Georgi Gerganov committed
September 3, 2023
G
speculative : PoC for speeding-up inference via speculative sampling (#2926)
Georgi Gerganov committed