bf16+tiled (m=128) GEMMs at M=4096 — 1072 GOPS, default serving kernels
Browse files- kernels/minilm-gemms-bf16/M4096/ffn1.insts.txt +0 -0
- kernels/minilm-gemms-bf16/M4096/ffn1.xclbin +0 -0
- kernels/minilm-gemms-bf16/M4096/ffn2.insts.txt +0 -0
- kernels/minilm-gemms-bf16/M4096/ffn2.xclbin +0 -0
- kernels/minilm-gemms-bf16/M4096/qkv.insts.txt +0 -0
- kernels/minilm-gemms-bf16/M4096/qkv.xclbin +0 -0
- kernels/minilm-gemms-bf16/M4096/qkvfused.insts.txt +0 -0
- kernels/minilm-gemms-bf16/M4096/qkvfused.xclbin +0 -0
kernels/minilm-gemms-bf16/M4096/ffn1.insts.txt
CHANGED
|
Binary files a/kernels/minilm-gemms-bf16/M4096/ffn1.insts.txt and b/kernels/minilm-gemms-bf16/M4096/ffn1.insts.txt differ
|
|
|
kernels/minilm-gemms-bf16/M4096/ffn1.xclbin
CHANGED
|
Binary files a/kernels/minilm-gemms-bf16/M4096/ffn1.xclbin and b/kernels/minilm-gemms-bf16/M4096/ffn1.xclbin differ
|
|
|
kernels/minilm-gemms-bf16/M4096/ffn2.insts.txt
CHANGED
|
Binary files a/kernels/minilm-gemms-bf16/M4096/ffn2.insts.txt and b/kernels/minilm-gemms-bf16/M4096/ffn2.insts.txt differ
|
|
|
kernels/minilm-gemms-bf16/M4096/ffn2.xclbin
CHANGED
|
Binary files a/kernels/minilm-gemms-bf16/M4096/ffn2.xclbin and b/kernels/minilm-gemms-bf16/M4096/ffn2.xclbin differ
|
|
|
kernels/minilm-gemms-bf16/M4096/qkv.insts.txt
CHANGED
|
Binary files a/kernels/minilm-gemms-bf16/M4096/qkv.insts.txt and b/kernels/minilm-gemms-bf16/M4096/qkv.insts.txt differ
|
|
|
kernels/minilm-gemms-bf16/M4096/qkv.xclbin
CHANGED
|
Binary files a/kernels/minilm-gemms-bf16/M4096/qkv.xclbin and b/kernels/minilm-gemms-bf16/M4096/qkv.xclbin differ
|
|
|
kernels/minilm-gemms-bf16/M4096/qkvfused.insts.txt
CHANGED
|
Binary files a/kernels/minilm-gemms-bf16/M4096/qkvfused.insts.txt and b/kernels/minilm-gemms-bf16/M4096/qkvfused.insts.txt differ
|
|
|
kernels/minilm-gemms-bf16/M4096/qkvfused.xclbin
CHANGED
|
Binary files a/kernels/minilm-gemms-bf16/M4096/qkvfused.xclbin and b/kernels/minilm-gemms-bf16/M4096/qkvfused.xclbin differ
|
|
|