Skip to content

Activity

Showing most recent first

move matmul multithreading pragma to inner loop

andrewkchanpushed 1 commit to main • 618ed7d…7beb553 • 
on Feb 11

fix divide-by-zero for f16, f32 models

andrewkchanpushed 1 commit to main • 9731ddc…618ed7d • 
on Feb 11

optimize blockwise scale reads in matmuls

Pull request merge
andrewkchanpushed 1 commit to main • c65d68d…9731ddc • 
on Feb 11

optimize blockwise scale reads in matmuls

Force push
andrewkchanforce pushed to block-reads • 538f548…b7ef6a1 • 
on Feb 11

optimize blockwise scale reads in matmuls

andrewkchancreated block-reads • 538f548 • 
on Feb 11

add note about using swap

andrewkchanpushed 1 commit to main • 677dcec…c65d68d • 
on Feb 8

readme

Force push
andrewkchanforce pushed to main • 5ad1359…677dcec • 
on Feb 6

readme

Force push
andrewkchanforce pushed to main • c31e780…5ad1359 • 
on Feb 6

readme

Force push
andrewkchanforce pushed to main • e7df4f1…c31e780 • 
on Feb 6

readme

andrewkchanpushed 1 commit to main • 53041de…e7df4f1 • 
on Feb 6

readme

Force push
andrewkchanforce pushed to main • d6f1ff2…53041de • 
on Feb 6

readme

andrewkchanpushed 1 commit to main • d2320e7…d6f1ff2 • 
on Feb 6

readme

andrewkchanpushed 1 commit to main • 96ce8c7…d2320e7 • 
on Feb 6

support table

andrewkchanpushed 1 commit to main • d5feb9f…96ce8c7 • 
on Feb 6

Fix signed arithmetic overflow

Pull request merge
andrewkchanpushed 18 commits to main • 3123a10…d5feb9f • 
on Feb 6

Fix signed arithmetic overflow

Force push
andrewkchanforce pushed to quant • 54a4d37…0637010 • 
on Feb 6

Fix signed arithmetic overflow

andrewkchanpushed 1 commit to quant • 36ae450…54a4d37 • 
on Feb 6

fixes

andrewkchanpushed 1 commit to quant • 30ad021…36ae450 • 
on Feb 5

.

andrewkchanpushed 1 commit to quant • 397017a…30ad021 • 
on Feb 5

.

andrewkchanpushed 1 commit to quant • 7acba15…397017a • 
on Feb 5

use generator to avoid OOM

Force push
andrewkchanforce pushed to quant • 8b976b3…7acba15 • 
on Feb 5

use generator to avoid OOM

andrewkchanpushed 1 commit to quant • 1bbd0f8…8b976b3 • 
on Feb 5

write to multiple shards in output_dir

andrewkchanpushed 1 commit to quant • 50d44e4…1bbd0f8 • 
on Feb 5

support reading from multiple files

andrewkchanpushed 1 commit to quant • b90552b…50d44e4 • 
on Feb 5

Apply bias after sigmoid rather than before

andrewkchanpushed 1 commit to quant • 5fe20e8…b90552b • 
on Feb 5

Add rope_v3

andrewkchanpushed 1 commit to quant • c84e31f…5fe20e8 • 
on Feb 5

fixes

andrewkchanpushed 1 commit to quant • b753e25…c84e31f • 
on Feb 4

dont require bias for dsv2

andrewkchanpushed 1 commit to main • 393613a…3123a10 • 
on Feb 4

.

Force push
andrewkchanforce pushed to quant • f5db332…b753e25 • 
on Feb 4

allow specifying num of layers to convert

Force push
andrewkchanforce pushed to main • f5686ac…393613a • 
on Feb 4