ggerganov
Follow
Pinned
2,941 contributions in the last year
Less
No contributions.
Low contributions.
Medium-low contributions.
Medium-high contributions.
High contributions.
More
Contribution activity
December 2023
Created 51 commits in 6 repositories
Created 1 repository
-
ggerganov/seamless_communication
C
This contribution was made on Dec 2
Created a pull request in ggerganov/llama.cpp that received 18 comments
llama : support quantum K cache
This PR adds support for quantizing the K cache using any of the existing quantization formats. It is applied on top of #4309 which allows per-laye…
+552
−74
lines changed
•
18
comments
Opened 16 other pull requests in 4 repositories
ggerganov/llama.cpp
9
merged
-
llama.swiftui : add bench functionality
This contribution was made on Dec 15
-
sync : ggml (SD ops, tests, kernels)
This contribution was made on Dec 13
-
sync : ggml (new ops, tests, backend, etc.)
This contribution was made on Dec 7
-
server : recognize cache_prompt parameter in OAI API
This contribution was made on Dec 6
-
llama : per-layer KV cache
This contribution was made on Dec 3
-
ggml : reuse ggml_get_n_tasks() in ggml_graph_plan()
This contribution was made on Dec 3
-
ggml : fix soft max out-of-bounds access
This contribution was made on Dec 3
-
llama : fix integer overflow during quantization
This contribution was made on Dec 1
-
llama : pad KV cache size
This contribution was made on Dec 1
ggerganov/whisper.cpp
3
merged
-
sync : ggml (Metal fixes, new ops, tests)
This contribution was made on Dec 13
-
sync : ggml (new ops, new backend, etc)
This contribution was made on Dec 7
-
ggml : re-enable blas for src0 != F32
This contribution was made on Dec 1
ggerganov/ggml
3
merged
-
sync : llama (mul_mat_id + get_rows kernels, typos)
This contribution was made on Dec 13
-
sync : llama.cpp (fused soft max, gpu cpy ops, etc.)
This contribution was made on Dec 7
-
metal : check supported ops at runtime
This contribution was made on Dec 5
facebookresearch/seamless_communication
1
merged
-
ggml : update cmake to build on macOS
This contribution was made on Dec 2
Reviewed 41 pull requests in 4 repositories
ggerganov/llama.cpp
25 pull requests
-
Add ability to cancel model loading
This contribution was made on Dec 19
-
Support for Phi-2
This contribution was made on Dec 18
-
Fix try_override for bool_value which always return true ignoring ove…
This contribution was made on Dec 18
-
Nomic Vulkan backend
This contribution was made on Dec 18
-
Prompt lookup decoding
This contribution was made on Dec 17
-
finetune : keep allocs alive until all allocations are done
This contribution was made on Dec 17
-
gguf-py : fail fast on nonsensical special token IDs
This contribution was made on Dec 17
-
build : Check the ROCm installation location
This contribution was made on Dec 17
-
Link to cublas dynamically on Windows even with LLAMA_STATIC
This contribution was made on Dec 17
-
Implement optional API Key Authentication for Secure Server-Client Communication
This contribution was made on Dec 15
-
ggml : group mul_mat_id rows by matrix (cpu only)
This contribution was made on Dec 15
-
ggml : use ggml_row_size where possible
This contribution was made on Dec 14
-
ggml : remove n_dims from ggml_tensor
This contribution was made on Dec 14
-
Fix "not enough space in the context's memory pool" error when loading certain models.
This contribution was made on Dec 14
-
build : detect host compiler and cuda compiler separately
This contribution was made on Dec 13
-
llama : add Mixtral support
This contribution was made on Dec 13
-
Adding Tests for GGUFWriter Class
This contribution was made on Dec 12
-
Windows XP: support MinGW 8.1.0
This contribution was made on Dec 12
-
Use
typosto fix comments and logs.This contribution was made on Dec 12 -
Check the full vocab for grammar only if necessary
This contribution was made on Dec 6
-
Samplers order parameters
This contribution was made on Dec 5
-
Create api_like_OAI.sh
This contribution was made on Dec 5
-
feat: Allow overriding GGUF metadata when loading model
This contribution was made on Dec 5
-
llama : per-layer KV cache
This contribution was made on Dec 3
-
build : enable libstdc++ assertions for debug builds
This contribution was made on Dec 1
- Some pull request reviews not shown.
ggerganov/ggml
10 pull requests
-
ggml: new gpu kernels + extends ggml_leaky_relu + ggml_pad
This contribution was made on Dec 13
-
ggml : add support for multiple ids rows in ggml_mat_mul_id
This contribution was made on Dec 9
-
sync : llama.cpp (fused soft max, gpu cpy ops, etc.)
This contribution was made on Dec 7
-
ggml-backend : remove backend self-registration
This contribution was made on Dec 7
-
Feature: Circular Padding
This contribution was made on Dec 7
-
add more int ops
This contribution was made on Dec 7
-
test-backend-ops : add performance eval mode + improve CUDA repeat and binary broadcast ops performance
This contribution was made on Dec 6
-
test-backend-ops : initialize ggml_argsort test with unique values
This contribution was made on Dec 5
-
ggml : full broadcast in mul, add, div + ggml_mul_mat_id, ggml_argsort, ggml_top_k
This contribution was made on Dec 5
-
ggml-cuda : fix usage without CUDA devices
This contribution was made on Dec 1
ggerganov/whisper.cpp
1 pull request
-
wchess: whisper assisted chess
This contribution was made on Dec 14
ggerganov/imtui
1 pull request
-
Minor rewrite - add support for drawing other things on the terminal while ImTui is active.
This contribution was made on Dec 3
Created an issue in ggerganov/llama.cpp that received 1 comment
llama : benchmark for Apple Silicon A-series mobile chips
Recently, we did a performance benchmark of llama.cpp for Apple Silicon M-series chips: #4167
I am planning to do a similar benchmark for Apple's m…
1
comment
Started 2 discussions in 2 repositories
ggerganov/llama.cpp
ggerganov/llama.cpp
-
Performance of llama.cpp on Apple Silicon A-series
This contribution was made on Dec 17
ggerganov/whisper.cpp
ggerganov/whisper.cpp
-
v1.5.2
This contribution was made on Dec 14
Answered 1 discussion in 1 repository
ggerganov/llama.cpp
ggerganov/llama.cpp
-
Where are the architectures implemented?
This contribution was made on Dec 6