Skip to content

Tmp #1

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Open
wants to merge 109 commits into
base: master
Choose a base branch
from
Open

Tmp #1

wants to merge 109 commits into from

Conversation

orca-zhang
Copy link
Owner

Make sure to read the contributing guidelines before submitting a PR

sszymczy and others added 30 commits January 22, 2025 15:19
@sszymczy
llama : experimental DeepSeek2 MLA implementation that caches latent …
93864cd 
…kv representations
@sszymczy
Merge remote-tracking branch 'origin/master' into deepseek2-mla-exp
f0ce53f
@slaren
llama : add option to override tensor buffers
f07c2ec
@sszymczy
llama : optimize DeepSeek MLA implementation
de538aa
@sszymczy
llama : Update tensor names in DeepSeek2 MLA implementation.
ce73063
@sszymczy
llama : add a second copy of c^KV cache in DeepSeek2 MLA to avoid tra…
202f323 
…nsposing the cache during inference
@sszymczy
llama : modified tensor permutations to multiply larger matrices duri…
93c5937 
…ng inference
@sszymczy
llama : removed unnecessary code in DeepSeek V2 implementation
1eee98f
@sszymczy
convert : make lint happy
8ff0991
@sszymczy
llama : prompt processing optimizations in DeepSeek V2
8a887de
@sszymczy
llama : avoid ggml_cont() is possible in DeepSeek V2 implementation
7654331
@sszymczy
llama : use all experts during warmup
83a473a
@sszymczy
llama : increased max_nodes as large MoE models use massive amounts o…
c8bc6e4 
…f nodes during warmup
@slaren
Merge remote-tracking branch 'origin/master' into sl/custom-tensor-of…
bb6b97e 
…fload
@slaren
add regex support
6c8d01a
@slaren
ggml : fix possible underflow in ggml_nbytes
538f609
@slaren
rebuild buft list on every call
8770ffa
@slaren
Merge remote-tracking branch 'origin/master' into sl/custom-tensor-of…
2e54433 
…fload
@orca-zhang
Merge remote-tracking branch 'origin' into sl/custom-tensor-offload
c44de8a
@orca-zhang
can shift
0d4ff95
@orca-zhang
Merge remote-tracking branch 'fairydreaming/deepseek2-mla-exp' into tmp
e162e47
@orca-zhang
Merge remote-tracking branch 'fairydreaming/experts-warmup' into tmp
7e4cae5
@orca-zhang
tmp
d256aa0
@orca-zhang
Merge branch 'ggml-org:master' into tmp
8c88371
@orca-zhang
Merge branch 'ggml-org:master' into tmp
42c0aa2
@orca-zhang
support dynamic wkv
d13d6ff
@orca-zhang
ignore missing
822807b
@orca-zhang
fix core dump
9f75d93
@orca-zhang
add debug log
dafd46a
@orca-zhang
debug
6277a0e
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

documentation Improvements or additions to documentation examples ggml Nvidia GPU python server

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

3 participants

@orca-zhang @sszymczy @slaren