ONNX models to use with avs-mlrt - Page 5

anton_foy · 24th October 2023, 08:15

Thanks I tried both of your suggestions but it is impossibly slow and I tried Vapoursynth vs-scunet on the same clip and 30 frames would take over 1 hour. So I guess this processing UHD with these models is out of the question for now.

kedautinh12 · 24th October 2023, 08:58

Waiting for cuda ver
https://github.com/Asd-g/avs-mlrt/is...ent-1771187328

anton_foy · 25th October 2023, 08:56

@kedautinh12
Thanks for the info good news.
Something I am wondering about is the num_streams param. Where to find the specification about this for my GPU? May this speed up if I put this to max for my card?

Selur · 25th October 2023, 18:51

The number of stream depends on the resolution of your input and the VRAM that is used with model X and 1 stream.
If you use a model with one stream and only a fraction of your VRAM is used, you can run multiple streams in parallel to speed things up.
=> increasing the streams will cost VRAM, but can speed things up
So if either the input resolution or the used model changes, the VRAM usage and thus the amount of streams your card can handle can change.

anton_foy · 26th October 2023, 07:46

Quote:

Originally Posted by Selur

The number of stream depends on the resolution of your input and the VRAM that is used with model X and 1 stream.
If you use a model with one stream and only a fraction of your VRAM is used, you can run multiple streams in parallel to speed things up.
=> increasing the streams will cost VRAM, but can speed things up
So if either the input resolution or the used model changes, the VRAM usage and thus the amount of streams your card can handle can change.

Thanks I will play around with this.

kedautinh12 · 22nd November 2023, 01:49

Would like to see SwinIR models:
https://github.com/JingyunLiang/SwinIR/releases
https://openmodeldb.info/?q=SwinIR

Dogway · 22nd November 2023, 02:43

Yes, me too. I tried in vapoursynth and SwinIR didn't work for me there either. Looking for a solid photo upscaler.

kedautinh12 · 22nd November 2023, 03:08

For Vapoursynth, why you don't ask HolyWu?
https://github.com/HolyWu/vs-swinir/...pen+is%3Aissue

Selur · 22nd November 2023, 05:42

https://github.com/HolyWu/vs-swinir works fine here,...

Dogway · 22nd November 2023, 20:06

Ok, I thought the SwinIR models in pth format were supposed to work. Thanks both.

kedautinh12 · 1st December 2023, 14:51

Would like to see CodeFormer models

https://github.com/sczhou/CodeFormer/releases

Selur · 2nd December 2023, 09:44

Assuming some folks here use those models on normal content, how do you deal with temporal inconsistencies?
(like this: https://forum.videohelp.com/attachme...73592/Face.mp4)
Personally, I use QTGMC with input type 3 to lessen this, but I was wondering what others do.

Cu Selur

kedautinh12 · 3rd December 2023, 03:16

Would like to see UniScale models
https://openmodeldb.info/?q=Uniscale

Selur · 3rd December 2023, 08:21

You can convert those with chaiNNer (https://github.com/chaiNNer-org/chaiNNer),... (these are also included in the Hybrid onnx models collection)

kedautinh12 · 20th December 2023, 04:27

Would like to see SRFormer models
https://openmodeldb.info/?q=SRFormer

Selur · 20th December 2023, 08:05

Quote:

Would like to see UniScale models

Those seem to work fine here with vs-mlrt. (they are rather slow)

Quote:

Would like to see SRFormer models

Best post such request over in the github issue tracker.

Cu Selur

kedautinh12 · 20th December 2023, 10:07

I can see it in other model. They use srformer to base for trained on
https://github.com/Phhofm/models

kedautinh12 · 21st December 2023, 04:52

Would like to see Cutie models
https://github.com/hkchengrex/Cutie?tab=readme-ov-file

takla · 19th February 2024, 10:38

Actually it runs "ok" as long as you're not using any of the broken models from the dropbox link (I haven't tested them all, but 4 of 4 gave an error)
Should be fine if converted manually with "opset 17"

takla · 10th March 2024, 16:42

Here are some models which I've converted to onnx (Mostly for Cartoons/Anime)
I can vouche for all of them working (tested on AMD GPU)
Original sources can be found here

2x_AniScale2S_Compact_i8_60K
Very good with BluRay sources

2x-LD-Anime-Compact
Amazing with high quality DVDs

2x_Garfieldjr_span48
Very Good for mediocre DVD quality

Example script:

Code:

LWLibavVideoSource("INPUT.mkv")
ConvertBits(32).ConvertToPlanarRGB()
mlrt_ort(network_path="C:\Program Files (x86)\AviSynth+\plugins64+\models2\2x-LD-Anime-Compact.onnx", builtin=false, provider="dml")

25th October 2023, 18:51	#84 \| Link
Selur Registered User Join Date: Oct 2001 Location: Germany Posts: 7,277	The number of stream depends on the resolution of your input and the VRAM that is used with model X and 1 stream. If you use a model with one stream and only a fraction of your VRAM is used, you can run multiple streams in parallel to speed things up. => increasing the streams will cost VRAM, but can speed things up So if either the input resolution or the used model changes, the VRAM usage and thus the amount of streams your card can handle can change. __________________ Hybrid here in the forum, homepage

22nd November 2023, 01:49	#86 \| Link
kedautinh12 Registered User Join Date: Jan 2018 Posts: 2,156	Would like to see SwinIR models: https://github.com/JingyunLiang/SwinIR/releases https://openmodeldb.info/?q=SwinIR Last edited by kedautinh12; 22nd November 2023 at 02:27.

22nd November 2023, 02:43	#87 \| Link
Dogway Registered User Join Date: Nov 2009 Posts: 2,361	Yes, me too. I tried in vapoursynth and SwinIR didn't work for me there either. Looking for a solid photo upscaler. __________________ i7-4790K@Stock::GTX 1070] AviSynth+ filters and mods on GitHub + Discussion thread

22nd November 2023, 05:42	#89 \| Link
Selur Registered User Join Date: Oct 2001 Location: Germany Posts: 7,277	https://github.com/HolyWu/vs-swinir works fine here,... __________________ Hybrid here in the forum, homepage

22nd November 2023, 20:06	#90 \| Link
Dogway Registered User Join Date: Nov 2009 Posts: 2,361	Ok, I thought the SwinIR models in pth format were supposed to work. Thanks both. __________________ i7-4790K@Stock::GTX 1070] AviSynth+ filters and mods on GitHub + Discussion thread

24th October 2023, 08:15	#81 \| Link
anton_foy Registered User Join Date: Dec 2005 Location: Sweden Posts: 703	Thanks I tried both of your suggestions but it is impossibly slow and I tried Vapoursynth vs-scunet on the same clip and 30 frames would take over 1 hour. So I guess this processing UHD with these models is out of the question for now.

24th October 2023, 08:58	#82 \| Link
kedautinh12 Registered User Join Date: Jan 2018 Posts: 2,156	Waiting for cuda ver https://github.com/Asd-g/avs-mlrt/is...ent-1771187328

25th October 2023, 08:56	#83 \| Link
anton_foy Registered User Join Date: Dec 2005 Location: Sweden Posts: 703	@kedautinh12 Thanks for the info good news. Something I am wondering about is the num_streams param. Where to find the specification about this for my GPU? May this speed up if I put this to max for my card?

22nd November 2023, 03:08	#88 \| Link
kedautinh12 Registered User Join Date: Jan 2018 Posts: 2,156	For Vapoursynth, why you don't ask HolyWu? https://github.com/HolyWu/vs-swinir/...pen+is%3Aissue

1st December 2023, 14:51	#91 \| Link
kedautinh12 Registered User Join Date: Jan 2018 Posts: 2,156	Would like to see CodeFormer models https://github.com/sczhou/CodeFormer/releases

2nd December 2023, 09:44	#92 \| Link
Selur Registered User Join Date: Oct 2001 Location: Germany Posts: 7,277	Assuming some folks here use those models on normal content, how do you deal with temporal inconsistencies? (like this: https://forum.videohelp.com/attachme...73592/Face.mp4) Personally, I use QTGMC with input type 3 to lessen this, but I was wondering what others do. Cu Selur __________________ Hybrid here in the forum, homepage

3rd December 2023, 03:16	#93 \| Link
kedautinh12 Registered User Join Date: Jan 2018 Posts: 2,156	Would like to see UniScale models https://openmodeldb.info/?q=Uniscale Last edited by kedautinh12; 3rd December 2023 at 05:04.

3rd December 2023, 08:21	#94 \| Link
Selur Registered User Join Date: Oct 2001 Location: Germany Posts: 7,277	You can convert those with chaiNNer (https://github.com/chaiNNer-org/chaiNNer),... (these are also included in the Hybrid onnx models collection) __________________ Hybrid here in the forum, homepage

20th December 2023, 04:27	#95 \| Link
kedautinh12 Registered User Join Date: Jan 2018 Posts: 2,156	Would like to see SRFormer models https://openmodeldb.info/?q=SRFormer

20th December 2023, 10:07	#97 \| Link
kedautinh12 Registered User Join Date: Jan 2018 Posts: 2,156	I can see it in other model. They use srformer to base for trained on https://github.com/Phhofm/models Last edited by kedautinh12; 20th December 2023 at 13:00.

21st December 2023, 04:52	#98 \| Link
kedautinh12 Registered User Join Date: Jan 2018 Posts: 2,156	Would like to see Cutie models https://github.com/hkchengrex/Cutie?tab=readme-ov-file

19th February 2024, 10:38	#99 \| Link
takla Registered User Join Date: May 2018 Posts: 184	Actually it runs "ok" as long as you're not using any of the broken models from the dropbox link (I haven't tested them all, but 4 of 4 gave an error) Should be fine if converted manually with "opset 17" Last edited by takla; 27th February 2024 at 03:48.

10th March 2024, 16:42	#100 \| Link
takla Registered User Join Date: May 2018 Posts: 184	Here are some models which I've converted to onnx (Mostly for Cartoons/Anime) I can vouche for all of them working (tested on AMD GPU) Original sources can be found here 2x_AniScale2S_Compact_i8_60K Very good with BluRay sources 2x-LD-Anime-Compact Amazing with high quality DVDs 2x_Garfieldjr_span48 Very Good for mediocre DVD quality Example script: Code: LWLibavVideoSource("INPUT.mkv") ConvertBits(32).ConvertToPlanarRGB() mlrt_ort(network_path="C:\Program Files (x86)\AviSynth+\plugins64+\models2\2x-LD-Anime-Compact.onnx", builtin=false, provider="dml")