MORPH
SIGN IN SIGN UP
abetlen / llama-cpp-python UNCLAIMED

Python bindings for llama.cpp

0 0 1 Python

COMMITS

/ llama_cpp/llama.py
422ebc89ce45af63c5f919da74ea68188ef70c1c
November 21, 2023
A
Fix: Add logit_bias to all completion api methods
Andrei Betlen committed
422ebc8
A
Add support for logit_bias outside of server api. Closes #827
Andrei Betlen committed
07e47f5
T
Added support for min_p (#921)
TK-Master committed
b8438f7
A
Fix #929
Andrei Betlen committed
a34d480
November 10, 2023
A
Fix sampling bug when logits_all=False
Andrei Betlen committed
6f0b0b1
A
Potential bugfix for eval
Andrei Betlen committed
d9b38e3
A
Fix: default max_tokens matches openai api (16 for completion, max length for chat completion)
Andrei Betlen committed
e7962d2
November 8, 2023
A
Add set_seed to Llama class
Andrei Betlen committed
fd41ed3
A
Fix destructor NoneType is not callable error
Andrei Betlen committed
ca4cb88
A
Add JSON mode support. Closes #881
Andrei Betlen committed
b30b9c3
A
Add seed parameter support for completion and chat_completion requests. Closes #884
Andrei Betlen committed
86aeb9f
D
Multimodal Support (Llava 1.5) (#821)
Damian Stewart committed
aab74f0
November 6, 2023
A
Fix type bug
Andrei Betlen committed
be0add1
A
Refactor Llama class internals
Andrei Betlen committed
e214a58
November 3, 2023
A
Clean up stdout / stderr suppression
Andrei Betlen committed
2ec043a
A
Rename internal only module utils to _utils
Andrei Betlen committed
4ea7027
A
Update llama.cpp
Andrei Betlen committed
df9362e
A
Add functionary support (#784)
Andrei committed
3af7b21
A
Migrate inference to llama_batch and llama_decode api (#795)
Andrei committed
ab028cb
November 2, 2023
A
Update llama.cpp
Andrei Betlen committed
fa83cc5
A
fix: tokenization of special characters: (#850)
Antoine Lizee committed
4d4e0f1
November 1, 2023
C
llama: fix exception in Llama.__del__ (#846)
cebtenzzre committed
eefd76f
M
fixed Llama._create_completion suffix check, it can be either None or str instance (#854)
Marko Tasic committed
9c8f4dc
October 24, 2023
A
Update llama.cpp
Andrei Betlen committed
53861c9
October 19, 2023
G
Fix streaming doesn't return finish reason (#798)
gmcgoldr committed
09a8406
A
Update llama.cpp
Andrei Betlen committed
ff58003
October 15, 2023
P
Make use of suppress_stdout_stderr when freeing model (#803)
Pierre Alexandre SCHEMBRI committed
10304d7
E
Add validation for tensor_split size exceeding LLAMA_MAX_DEVICES (#820)
Eric Liu committed
b501665
September 30, 2023
A
Fix logits_all bug
Andrei Betlen committed
d696251
A
Fix bug in embedding
Andrei Betlen committed
42bb721
September 29, 2023
A
Configurable Chat Formats (#711)
Andrei committed
3bca770
J
Fix rope scaling defaults (#767)
Josh XT committed
a945404
A
Update llama.cpp
Andrei Betlen committed
1a1c3dc
September 18, 2023
A
Update llama.cpp
Andrei Betlen committed
38e34c9
September 14, 2023
A
Add numa support, low level api users must now explicitly call llama_backend_init at the start of their programs.
Andrei Betlen committed
f4090a0
A
Reorder init params to match llama.cpp order
Andrei Betlen committed
6a20293
A
Explicitly make all init params other than model_path into keyword only params
Andrei Betlen committed
c8f9b8a
A
Add kwargs to init to catch extra params
Andrei Betlen committed
a68f9e2
A
remove print
Andrei Betlen committed
9e345a4
A
Convert missed llama.cpp constants into standard python types
Andrei Betlen committed
517f9ed
A
Fix tensor_split cli option
Andrei Betlen committed
c4c440b
September 12, 2023
A
Merge branch 'main' into v0.2-wip
Andrei Betlen committed
1910793
August 29, 2023
A
cjk pr minor cleanup
Andrei Betlen committed
3f76e1d
A
Merge pull request #309 from MeouSker77/fix-CJK
Andrei committed
bae44ec
August 27, 2023
A
Update llama.cpp
Andrei Betlen committed
4887973
A
Update llama.cpp
Andrei Betlen committed
3a29d65
August 25, 2023
A
Merge branch 'main' into v0.2-wip
Andrei Betlen committed
ac47d55
A
Use _with_model variants for tokenization
Andrei Betlen committed
48cf43b
A
Strip leading space when de-tokenizing.
Andrei Betlen committed
8ac5946
August 24, 2023
A
Remove deprecated params
Andrei Betlen committed
4ed632c