ggerganov/llama.cpp

Pull Requests Commits

llama : change starcoder2 rope type

ggerganov committed 1 year ago

Verified 9862d59c

handle `rope-theta`

pacman100 committed 1 year ago

b67b8f64

remove redundant changes

pacman100 committed 1 year ago

fdd886f7

Update llama.cpp

pacman100 committed 1 year ago

5c06625f

resolve comments

pacman100 committed 1 year ago

10aa6e92

skip rope freq and rotary embeddings from being serialized

pacman100 committed 1 year ago

d62ce1c6

handle rope type

pacman100 committed 1 year ago

6c108068

Add support for starcoder2

pacman100 committed 1 year ago

ab4eab3a

llama : constified `llama_set_state_data`'s `src` (#5774)

MarcusDunn committed 1 year ago

Verified d5ab2975

ci : reduce 3b ppl chunks to 1 to avoid timeout (#5771)

ggerganov committed 1 year ago

Verified 87c91c07

make portability_enumeration_ext apple only (#5757)

netrunnereve committed 1 year ago

Verified 317709b2

llama : remove deprecated API (#5770)

ggerganov committed 1 year ago

Verified 08c5ee87

awq-py : remove (#5768)

ggerganov committed 1 year ago

Verified 78aacf36

ggerganov committed 1 year ago

Verified 8c0e8f4e

add google magika inference example (ggml/748)

slaren committed 1 year ago

Verified 2774b0c9

Introduce backend GUIDs (ggml/743)

uextm committed 1 year ago

Verified 5f706718

server : hit Ctrl+C twice to exit (#5734)

ngxson committed 1 year ago

Verified a693bea1

llama : fix non-quantization of expert gating tensors (#5754)

compilade committed 1 year ago

Verified adcb12a9

llama : improve BERT tokenization (#5740)

iamlemec committed 1 year ago

Verified 177628bf

readme : add link to LLaVA 1.6 models (#5758)

danbev committed 1 year ago

Verified 6c441686

server : add "/chat/completions" alias for "/v1/...` (#5722)

jorgealias committed 1 year ago

Verified efc72253

ggml : make i-quants work with super-blocks of 64 (CPU,Metal) (#5760)

ikawrakow committed 1 year ago

Verified 7c4263d4

Attempt to fix android build (#5752)

ikawrakow committed 1 year ago

Verified cb49e0f8

IQ4_XS: a 4.25 bpw quantization (#5747)

ikawrakow committed 1 year ago

Verified 0becb22a

cuda : replace remaining shfl_xor with calls to warp_reduce functions (#5744)

Engininja2 committed 1 year ago

Verified c24a2a6e

ggml-quants : fix avx2 iq1_s vec_dot when compiled with gcc (#5742)

Engininja2 committed 1 year ago

Verified 1f30b7a9

llama : fix defrag bugs + add parameter (#5735)

ggerganov committed 1 year ago

Verified 9d533a77

Makefile: use variables for cublas (#5689)

lindeer committed 1 year ago

Verified cbbd1efa

fix server hangs on empty prompt (#5733)

ngxson committed 1 year ago

Verified b11a93df

Adding IQ2_S and IQ2_M to complete coverage of the 2-3 bit quantization range (#5721)

ikawrakow committed 1 year ago

Verified a33e6a0d

Older