2-v1.4

Full Changelog: 2-v1.3.2...2-v1.4
I'm having a hard time uploading large files... It was seeding eaten up my bandwidth...

Python 3.12.3
VapourSynth R68+10
FFmpeg 7.0
VapourSynth-RIFE-ncnn-Vulkan r9_mod_v18
vs-mlrt v14.test (TensorRT 8.6) (keep using old version because new versions seem to have some problems)
onnxruntime-directml (for new scd method: sudo)

jumbo: rifenvk v2-v4.17 models + vs-mlrt v4.0-v4.17 models (2 implementations) note: the 7z archive uses 3840M dictionary, it takes at least 3842M RAM to decompress.
jumbo-lp2-lc4: so I tried many combinations, instead of making some lighter packs, and found that -mlp=2 -mlc=4 with 7-zip gives better compression, it saves 35.996 MiB! [canned applause and cheer] (what have I done with my life)
basic: rifenvk v2-v4.17 models
slim: rifenvk v4.7-v4.17 models

(basic is compressed with -mlp=1 -mlc=1 and slim is with -mlp=1 -mlc=0 holy [beep] I need to stop doing this)

2-v1.3.2

base: rife-nvk, model ver up to 4.8
slim: rife-nvk with only v4+ models
fat: base + vs-mlrt (trt backend only), model ver up to 4.10

2-v1.3.1

相比上一release，仅更新了ddfi2.py。

2-v1.3.0

基于vs-mlrt v13的新功能做出的改动，其余部分没有变化
这次的full只包含了vstrt

2-v1.2.1

相比上一个release，只有ddfi2.py的区别

2-v1.2.0

vapoursynth r60 -> r61
full包含了vs-mlrt和它附带的rife模型
另一个仅包含vapoursynth-rife-ncnn-vulkan和它的模型
单独使用rife进行2倍帧率的话，在我的3080ti上trt可以提供高达80%的速度提升，然而在ddfi里面只能提速10%，不知道怎么回事，我太菜了吧

2-v1.1.0

改用fork https://github.com/styler00dollar/VapourSynth-RIFE-ncnn-Vulkan

2-v1.0.1 Components Update

change for vs-rife-nvk r9

2-v1.0

完善部分细节

2-v0.9

重新设计流程，使用ssim与planestatsmax来判断重复帧，使用预处理的列表与std.deleteframes来去除重复帧。