ggerganov/llama.cpp

Pull Requests Commits

tokenize example: Respect normal add BOS token behavior (#4126)

KerfuffleV2 committed 2 years ago

Verified 28a2e6e7

scripts : Remove missed baichuan convert script (#4127)

Galunid committed 2 years ago

Verified 0b5c3b04

Clean up ggml-cuda.cu warnings when compiling with clang (for ROCM) (#4124)

KerfuffleV2 committed 2 years ago

Verified 2923f17f

llama : increase max nodes (#4115)

slaren committed 2 years ago

Verified bbecf3f4

build : support ppc64le build for make and CMake (#3963)

bufferoverflow committed 2 years ago

Verified 8e936108

tokenize : fix trailing whitespace

ggerganov committed 2 years ago

Verified 5ad387e9

examples : add tokenize (#4039)

zakkor committed 2 years ago

Verified 2fa02b4b

convert : use 'model' value if it exists. This allows karpathy/tinyllamas to load (#4089)

dmahurin committed 2 years ago

Verified 2ab0707a

py : Falcon HF compatibility (#4104)

cmp-nct committed 2 years ago

Verified 11173c92

common : improve yaml log escaping (#4080)

joennlae committed 2 years ago

Verified 9e87ef60

llava : fix compilation warning that fread return value is not used (#4069)

huawei-lin committed 2 years ago

Verified c7cce124

py : remove superfluous import statements (#4076)

jpodivin committed 2 years ago

Verified f7d5e975

train : move number of gpu layers argument parsing to common/train.cpp (#4074)

jpodivin committed 2 years ago

Verified ba4cf5c0

llama : add functions to get the model's metadata (#4013)

slaren committed 2 years ago

Verified e85bb1a8

finetune : speed-up ggml_compute_forward_out_prod_f32 via BLAS (#4079)

gwjr committed 2 years ago

Verified 3e916a07

finetune : zero the loraB initial vectors (#4082)

AndrewGodfrey committed 2 years ago

Verified 947f64f1

cuda : get_row_rounding F32 (#4095)

AndrewGodfrey committed 2 years ago

Verified b83e149e

llama : fix data units (#4101)

ggerganov committed 2 years ago

Verified 4f447a48

Respect tokenizer.ggml.add_bos_token value when tokenizing (#4040)

KerfuffleV2 committed 2 years ago

Verified 91f64993

gguf : fix potential infinite loops while parsing (#4100)

texmex76 committed 2 years ago

Verified 8da46278

llama : restore prefix space in llama tokenizer (#4081)

cebtenzzre committed 2 years ago

Verified a6fc554e

ggml-cuda : increase max graph size (#4084)

slaren committed 2 years ago

Verified 1cf2850d

Fix MacOS Sonoma model quantization (#4052)

TortoiseHam committed 2 years ago

Verified 6bb4908a

stablelm : StableLM support (#3586)

Galunid committed 2 years ago

Verified 36eed0c4

convert.py: also look for plain model.safetensors (#4043)

afrideva committed 2 years ago

Verified b46d12f8

llava : fix regression for square images in #3613 (#4056)

monatis committed 2 years ago

Verified bd90eca2

ggml : sync (im2col, GPU conv, 32-bit arm compat) (#4060)

ggerganov committed 2 years ago

Verified 3d68f364

readme : update hot topics

ggerganov committed 2 years ago

Verified c049b37d

sync : ggml (backend v2) (#3912)

ggerganov committed 2 years ago

Verified 4760e7cc

Add ReLU and SQR CUDA ops to (partially) fix Persimmon offloading (#4041)

KerfuffleV2 committed 2 years ago

Verified bb50a792

Newer Older