Skip to content

Activity

Showing most recent first

ggml : fix batch for ggml_conv_2d (ggml-org#528)

jploskipushed 53 commits to master • 244776a…13e2fa6 • 
on Sep 30, 2023

ggml : implementation of xPos RoPE (ggml-org#441); also extends ggml_…

jploskicreated xpos • 879a92a • 
on Aug 9, 2023

ggml : mul mat tweaks (ggml-org#439)

jploskipushed 95 commits to master • 4a7db90…244776a • 
on Aug 9, 2023

Alternative implementation using ggml_flash_attn (~16% slower than fa…

jploskicreated falcon40b-flash • fac72a2 • 
on Jul 15, 2023

Removed unnecessary reshapes when retrieving kv from cache

jploskipushed 1 commit to falcon40b • d8c51b2…d5295b4 • 
on Jun 26, 2023

Eliminated need for ggml_repeat2 by using a modified version of ggml-…

jploskicreated falcon40b-norepeat • 2e30a2b • 
on Jun 25, 2023

Merge remote-tracking branch 'origin/pr/224' into falcon40b

jploskipushed 2 commits to falcon40b • 78ccbba…d8c51b2 • 
on Jun 25, 2023

Merge branch 'master' into falcon40b

jploskipushed 40 commits to falcon40b • 8b22ea8…78ccbba • 
on Jun 25, 2023

py : add requirements.txt (ggml-org#201)

jploskipushed 39 commits to master • f52d2a0…4a7db90 • 
on Jun 25, 2023

py : add requirements.txt (ggml-org#201)

jploskipushed 39 commits to master • f52d2a0…4a7db90 • 
on Jun 25, 2023

Fixed regression because of incorrect ctx_size calculation

jploskipushed 1 commit to falcon40b • 8f2609a…8b22ea8 • 
on Jun 15, 2023

Merge pull request #2 from ochafik/falcon40b-ctx-size

Pull request merge
jploskipushed 3 commits to falcon40b • cc8ac10…8f2609a • 
on Jun 15, 2023

Merge pull request #1 from KerfuffleV2/feat-improve-falcon-convert-hf

Pull request merge
jploskipushed 2 commits to falcon40b • fe13c37…cc8ac10 • 
on Jun 14, 2023

Added Kerfuffle's magic context size fix

jploskipushed 1 commit to falcon40b • 5ee0488…fe13c37 • 
on Jun 14, 2023

Added rearrange of qkv weight memory layout to convert-hf-to-ggml.py …

jploskipushed 1 commit to falcon40b • b8bb9b8…5ee0488 • 
on Jun 14, 2023

Fixed offset calculation bug during extraction of query vectors

jploskipushed 1 commit to falcon40b • 7db8803…b8bb9b8 • 
on Jun 14, 2023

Fixed quantized version not working due to wrong data type

jploskipushed 1 commit to falcon40b • 68f6133…7db8803 • 
on Jun 13, 2023

Removed debug code

jploskipushed 1 commit to falcon40b • d332539…68f6133 • 
on Jun 13, 2023

Updated falcon-quantize to match 7B/40B format produced by convert-hf…

jploskipushed 1 commit to falcon40b • 3bc786b…d332539 • 
on Jun 13, 2023

Version which exactly reproduces outputs of the Python implementation…

jploskipushed 2 commits to falcon40b • 27cf1ad…3bc786b • 
on Jun 13, 2023

Experimental support for Falcon-40B (and Falcon-7B); breaks 7B GGML c…

jploskicreated falcon40b • 27cf1ad • 
on Jun 10, 2023

cmake : export all symbols on windows when building shared library (g…

jploskipushed 80 commits to master • ff6e03c…f52d2a0 • 
on Jun 10, 2023

cmake : export all symbols on windows when building shared library (g…

jploskipushed 80 commits to master • ff6e03c…f52d2a0 • 
on Jun 10, 2023

Added mention of missing ALiBi to README.md

jploskipushed 1 commit to mpt-experiment • 2a0b7f8…29b5732 • 
on May 7, 2023

Fixed obvious typo in layer mapping

jploskipushed 1 commit to mpt-experiment • b633277…2a0b7f8 • 
on May 7, 2023

Fixed syntax error, added comment about ALiBi

jploskipushed 1 commit to mpt-experiment • 70d7062…b633277 • 
on May 7, 2023

Added README.md

jploskipushed 6 commits to mpt-experiment • ff6e03c…70d7062 • 
on May 7, 2023

examples : add dolly-v2 sample inference (ggml-org#132)

jploskicreated mpt-experiment • ff6e03c • 
on May 7, 2023

examples : add dolly-v2 sample inference (ggml-org#132)

Force push
jploskiforce pushed to master • a650a36…ff6e03c • 
on May 7, 2023

ggml : backwards compatibility with gcc < 9 (ggml-org#99)

jploskipushed 1 commit to master • e999951…a650a36 • 
on Apr 24, 2023