p
/
gallery
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Add benchmarks against WDMassTagger

This commit is contained in:
Přemysl Eric Janouch 2024-01-19 20:00:05 +01:00
parent fd5e3bb166
commit 4131bc5d31
Signed by: p
GPG Key ID: A0420B94F92B9493
1 changed files with 18 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

20
deeptagger/README.adoc
View File

@ -142,7 +142,7 @@ The kernel panic was repeatable.
GPU inference GPU inference
~~~~~~~~~~~~~ ~~~~~~~~~~~~~
[cols="<,>,>", options=header] [cols="<2,>1,>1", options=header]
|=== |===
|Model|Batch size|Time |Model|Batch size|Time
|DeepDanbooru|1|24 s |DeepDanbooru|1|24 s
@ -177,7 +177,7 @@ GPU inference
CPU inference CPU inference
~~~~~~~~~~~~~ ~~~~~~~~~~~~~
[cols="<,>,>", options=header] [cols="<2,>1,>1", options=header]
|=== |===
|Model|Batch size|Time |Model|Batch size|Time
|DeepDanbooru|8|54 s |DeepDanbooru|8|54 s
@ -210,3 +210,19 @@ CPU inference
|ML-Danbooru Caformer dec-5-97527|8|241 s |ML-Danbooru Caformer dec-5-97527|8|241 s
|ML-Danbooru Caformer dec-5-97527|1|262 s |ML-Danbooru Caformer dec-5-97527|1|262 s
|=== |===
Comparison with WDMassTagger
----------------------------
Using CUDA, on the same Linux computer as above, on a sample of 6352 images.
We're a bit slower, depending on the model.
Batch sizes of 16 and 32 give practically equivalent results for both.
[cols="<,>,>,>", options="header,autowidth"]
|===
|Model|WDMassTagger|deeptagger (batch)|Ratio
|wd-v1-4-convnext-tagger-v2 |1:18 |1:55 |68 %
|wd-v1-4-convnextv2-tagger-v2 |1:20 |2:10 |62 %
|wd-v1-4-moat-tagger-v2 |1:22 |1:52 |73 %
|wd-v1-4-swinv2-tagger-v2 |1:28 |1:34 |94 %
|wd-v1-4-vit-tagger-v2 |1:16 |1:22 |93 %
|===