ggerganov
Follow
Pinned
3,728 contributions in the last year
Less
No contributions.
Low contributions.
Medium-low contributions.
Medium-high contributions.
High contributions.
More
Contribution activity
February 2024
Created 141 commits in 4 repositories
Created 1 repository
-
ggerganov/bert.cpp
C++
This contribution was made on Feb 3
Created a pull request in ggerganov/llama.cpp that received 26 comments
ggml : add ALiBi support for ggml_soft_max_ext
ref: #3470 CPU Metal CUDA Vulkan (cc @0cc4m) SYCL (cc @abhilash1910) Kompute (cc @cebtenzzre) Missing backends currently generate compile w…
+344
−353
lines changed
•
26
comments
Opened 32 other pull requests in 5 repositories
ggerganov/llama.cpp
2
open
20
merged
1
closed
-
llama : switch to floating-point token positions
This contribution was made on Feb 23
-
py : minor fixes
This contribution was made on Feb 22
-
ggml : always define ggml_fp16_t as uint16_t
This contribution was made on Feb 22
-
gemma : use more bits for the token_embd.weight tensor
This contribution was made on Feb 21
-
py : add Gemma conversion from HF models
This contribution was made on Feb 21
-
sync : ggml
This contribution was made on Feb 21
-
sync : ggml
This contribution was made on Feb 19
-
cmake : pass -Werror through -Xcompiler
This contribution was made on Feb 19
-
ci : fix wikitext url + compile warnings
This contribution was made on Feb 18
-
llama : rename n_ctx to kv_size
This contribution was made on Feb 18
-
nix: update flake.lock
This contribution was made on Feb 18
-
cmake : try to fix Android build
This contribution was made on Feb 16
-
cmake : fix VULKAN and ROCm builds
This contribution was made on Feb 16
-
scripts : add helpers script for bench comparing commits
This contribution was made on Feb 16
-
scripts : add hf.sh helper script
This contribution was made on Feb 15
-
bert : add tests + fix quantization
This contribution was made on Feb 13
-
tests : multi-thread the tokenizer tests
This contribution was made on Feb 13
-
tests : disable moe test
This contribution was made on Feb 13
-
swift : package no longer use ggml dependency
This contribution was made on Feb 12
-
sync : ggml
This contribution was made on Feb 11
-
nix: update flake.lock
This contribution was made on Feb 11
-
py : handle byte tokens in
get_token_typeThis contribution was made on Feb 5 -
nix: update flake.lock
This contribution was made on Feb 4
ggerganov/whisper.cpp
3
merged
1
closed
-
ggml : 32-bit arm compat
This contribution was made on Feb 22
-
swift : package no longer use ggml dependency
This contribution was made on Feb 12
-
whisper : fix external encoder
This contribution was made on Feb 12
-
whisper : fix usage of extenral encoders (e.g. CoreML)
This contribution was made on Feb 12
ggerganov/ggml
3
merged
-
sync : llama.cpp
This contribution was made on Feb 21
-
sync : llama.cpp
This contribution was made on Feb 19
-
examples : remove old stuff
This contribution was made on Feb 10
iamlemec/bert.cpp
1
merged
-
bert : various improvements
This contribution was made on Feb 3
Pints-App/llama.cpp
1
merged
-
cuda : fix flash_attn kernel to produce same results as CPU
This contribution was made on Feb 1
Reviewed 137 pull requests in 4 repositories
ggerganov/llama.cpp
25 pull requests
-
mpt : do not duplicate token_embd.weight on disk
This contribution was made on Feb 22
-
Add Gemma chat template
This contribution was made on Feb 22
-
server: init functional tests
This contribution was made on Feb 22
-
Server: fallback to chatml, add AlphaMonarch chat template
This contribution was made on Feb 22
-
server: clarify some params in the docs
This contribution was made on Feb 22
-
[SYCL] Add support for soft_max ALiBi
This contribution was made on Feb 22
-
MPT: add optional bias parameters
This contribution was made on Feb 22
-
Add docs for llama_chat_apply_template
This contribution was made on Feb 21
-
llama : fix session save/load with quantized KV
This contribution was made on Feb 21
-
gemma : allow offloading the output tensor
This contribution was made on Feb 21
-
readme: add LocalAI to the availables UI
This contribution was made on Feb 21
-
server: health: fix race condition on slots data using tasks queue
This contribution was made on Feb 21
-
llava: add --skip-unknown to 1.6 convert.py
This contribution was made on Feb 21
-
Add
gemmamodelThis contribution was made on Feb 21 -
Attempt to fix pre-tokenizer
This contribution was made on Feb 21
-
examples : do not assume BOS when shifting context
This contribution was made on Feb 21
-
support llava 1.6 image embedding dimension in server
This contribution was made on Feb 20
-
llava: add explicit instructions for llava-1.6
This contribution was made on Feb 20
-
IQ4_NL: 4-bit non-linear quants with blocks of 32
This contribution was made on Feb 20
-
Server: use llama_chat_apply_template
This contribution was made on Feb 20
-
Add maid to UI list
This contribution was made on Feb 20
-
metal : add build system support for embedded metal library
This contribution was made on Feb 20
-
server: health endpoint configurable failure on no slot
This contribution was made on Feb 20
-
examples : support minLength and maxLength in JSON schema grammar converter
This contribution was made on Feb 19
-
Rm obsolete warning build options sycl cmake
This contribution was made on Feb 19
- Some pull request reviews not shown.
ggerganov/whisper.cpp
17 pull requests
-
Add SYCL logic in whisper
This contribution was made on Feb 23
-
openvino : fix convert-whisper-to-openvino.py for v2023.0.0 (#1870)
This contribution was made on Feb 22
-
main : fix file existence check in main.cpp
This contribution was made on Feb 22
-
examples: refine the Android sample code
This contribution was made on Feb 22
-
talk and talk-llama: Pass text_to_speak as a file
This contribution was made on Feb 19
-
main : check if input files exist before proceeding
This contribution was made on Feb 19
-
clean up common code in examples
This contribution was made on Feb 19
-
Fix the decoding issues series: BPE Tokenizer
This contribution was made on Feb 19
-
whisper : fix usage of extenral encoders (e.g. CoreML)
This contribution was made on Feb 12
-
added audio_ctx argument to main and server examples
This contribution was made on Feb 11
-
Embed Metal library source into compiled binary
This contribution was made on Feb 11
-
server: Allow CORS request with authorization headers
This contribution was made on Feb 9
-
Expose CUDA device setting in public API
This contribution was made on Feb 9
-
Add macOS deployment target option to Makefile
This contribution was made on Feb 9
-
Fix the decoding issues
This contribution was made on Feb 6
-
WIP Very rough cut of streaming from stdin.
This contribution was made on Feb 5
-
whisper.android: How to build with CLBlast
This contribution was made on Feb 5
ggerganov/ggml
7 pull requests
-
Introduce backend GUIDs
This contribution was made on Feb 22
-
refactored compute forward to not pass in the src tensors each time
This contribution was made on Feb 21
-
fix the conv_2d batch mode
This contribution was made on Feb 20
-
ggml-alloc : allocate all leafs as if they were inputs
This contribution was made on Feb 12
-
ggml-alloc v3
This contribution was made on Feb 11
-
A way to use abort_callback with the cpu backend
This contribution was made on Feb 9
-
[WIP, don't merge] unity.cpp -> ggml master
This contribution was made on Feb 2
Pints-App/llama.cpp
1 pull request
-
WIP: Flash Attention implementation (forward + backward)
This contribution was made on Feb 11
Created an issue in ggerganov/ggml that received 2 comments
ggml : simplify the ggml_compute_forward_ calls
There is no need to explicitly list all the source tensors - they can be accessed through dst when needed:
ggml/src/ggml.c Lines 14783 to 1481…
2
comments
Opened 2 other issues in 2 repositories
ggerganov/llama.cpp
1
open
-
llama : update the convert-llama2c-to-ggml example
This contribution was made on Feb 20
ggerganov/ggml
1
open
-
ggml : add Magika inference
This contribution was made on Feb 16
Answered 1 discussion in 1 repository
ggerganov/llama.cpp
ggerganov/llama.cpp
-
Saving state
This contribution was made on Feb 11