Skip to content

Commit 0723ee5

Browse files
leejetleejet
authored
refactor: optimize option printing (#900)
1 parent 90ef5f8 commit 0723ee5

File tree

2 files changed

+580
-318
lines changed

2 files changed

+580
-318
lines changed

‎examples/cli/README.md‎

Lines changed: 105 additions & 109 deletions
Original file line numberDiff line numberDiff line change
@@ -1,114 +1,110 @@
11
# Run
22

33
```
4-
usage: ./bin/sd [arguments]
4+
usage: ./bin/sd [options]
55
6-
arguments:
7-
-h, --help show this help message and exit
8-
-M, --mode [MODE] run mode, one of: [img_gen, vid_gen, upscale, convert], default: img_gen
9-
-t, --threads N number of threads to use during computation (default: -1)
10-
If threads <= 0, then threads will be set to the number of CPU physical cores
11-
--offload-to-cpu place the weights in RAM to save VRAM, and automatically load them into VRAM when needed
12-
-m, --model [MODEL] path to full model
13-
--diffusion-model path to the standalone diffusion model
14-
--high-noise-diffusion-model path to the standalone high noise diffusion model
15-
--clip_l path to the clip-l text encoder
16-
--clip_g path to the clip-g text encoder
17-
--clip_vision path to the clip-vision encoder
18-
--t5xxl path to the t5xxl text encoder
19-
--qwen2vl path to the qwen2vl text encoder
20-
--qwen2vl_vision path to the qwen2vl vit
21-
--vae [VAE] path to vae
22-
--taesd [TAESD_PATH] path to taesd. Using Tiny AutoEncoder for fast decoding (low quality)
23-
--control-net [CONTROL_PATH] path to control net model
24-
--embd-dir [EMBEDDING_PATH] path to embeddings
25-
--upscale-model [ESRGAN_PATH] path to esrgan model. For img_gen mode, upscale images after generate, just RealESRGAN_x4plus_anime_6B supported by now
26-
--upscale-repeats Run the ESRGAN upscaler this many times (default 1)
27-
--type [TYPE] weight type (examples: f32, f16, q4_0, q4_1, q5_0, q5_1, q8_0, q2_K, q3_K, q4_K)
28-
If not specified, the default is the type of the weight file
29-
--tensor-type-rules [EXPRESSION] weight type per tensor pattern (example: "^vae\.=f16,model\.=q8_0")
30-
--lora-model-dir [DIR] lora model directory
31-
-i, --init-img [IMAGE] path to the init image, required by img2img
32-
--mask [MASK] path to the mask image, required by img2img with mask
33-
-i, --end-img [IMAGE] path to the end image, required by flf2v
34-
--control-image [IMAGE] path to image condition, control net
35-
-r, --ref-image [PATH] reference image for Flux Kontext models (can be used multiple times)
36-
--control-video [PATH] path to control video frames, It must be a directory path.
37-
The video frames inside should be stored as images in lexicographical (character) order
38-
For example, if the control video path is `frames`, the directory contain images such as 00.png, 01.png, ... etc.
39-
--increase-ref-index automatically increase the indices of references images based on the order they are listed (starting with 1).
40-
--disable-auto-resize-ref-image disable auto resize of ref images
41-
-o, --output OUTPUT path to write result image to (default: ./output.png)
42-
-p, --prompt [PROMPT] the prompt to render
43-
-n, --negative-prompt PROMPT the negative prompt (default: "")
44-
--cfg-scale SCALE unconditional guidance scale: (default: 7.0)
45-
--img-cfg-scale SCALE image guidance scale for inpaint or instruct-pix2pix models: (default: same as --cfg-scale)
46-
--guidance SCALE distilled guidance scale for models with guidance input (default: 3.5)
47-
--slg-scale SCALE skip layer guidance (SLG) scale, only for DiT models: (default: 0)
48-
0 means disabled, a value of 2.5 is nice for sd3.5 medium
49-
--eta SCALE eta in DDIM, only for DDIM and TCD: (default: 0)
50-
--skip-layers LAYERS Layers to skip for SLG steps: (default: [7,8,9])
51-
--skip-layer-start START SLG enabling point: (default: 0.01)
52-
--skip-layer-end END SLG disabling point: (default: 0.2)
53-
--scheduler {discrete, karras, exponential, ays, gits, smoothstep, sgm_uniform, simple} Denoiser sigma scheduler (default: discrete)
54-
--sampling-method {euler, euler_a, heun, dpm2, dpm++2s_a, dpm++2m, dpm++2mv2, ipndm, ipndm_v, lcm, ddim_trailing, tcd}
55-
sampling method (default: "euler" for Flux/SD3/Wan, "euler_a" otherwise)
56-
--timestep-shift N shift timestep for NitroFusion models, default: 0, recommended N for NitroSD-Realism around 250 and 500 for NitroSD-Vibrant
57-
--steps STEPS number of sample steps (default: 20)
58-
--high-noise-cfg-scale SCALE (high noise) unconditional guidance scale: (default: 7.0)
59-
--high-noise-img-cfg-scale SCALE (high noise) image guidance scale for inpaint or instruct-pix2pix models: (default: same as --cfg-scale)
60-
--high-noise-guidance SCALE (high noise) distilled guidance scale for models with guidance input (default: 3.5)
61-
--high-noise-slg-scale SCALE (high noise) skip layer guidance (SLG) scale, only for DiT models: (default: 0)
62-
0 means disabled, a value of 2.5 is nice for sd3.5 medium
63-
--high-noise-eta SCALE (high noise) eta in DDIM, only for DDIM and TCD: (default: 0)
64-
--high-noise-skip-layers LAYERS (high noise) Layers to skip for SLG steps: (default: [7,8,9])
65-
--high-noise-skip-layer-start (high noise) SLG enabling point: (default: 0.01)
66-
--high-noise-skip-layer-end END (high noise) SLG disabling point: (default: 0.2)
67-
--high-noise-scheduler {discrete, karras, exponential, ays, gits, smoothstep, sgm_uniform, simple} Denoiser sigma scheduler (default: discrete)
68-
--high-noise-sampling-method {euler, euler_a, heun, dpm2, dpm++2s_a, dpm++2m, dpm++2mv2, ipndm, ipndm_v, lcm, ddim_trailing, tcd}
69-
(high noise) sampling method (default: "euler_a")
70-
--high-noise-steps STEPS (high noise) number of sample steps (default: -1 = auto)
71-
SLG will be enabled at step int([STEPS]*[START]) and disabled at int([STEPS]*[END])
72-
--strength STRENGTH strength for noising/unnoising (default: 0.75)
73-
--control-strength STRENGTH strength to apply Control Net (default: 0.9)
74-
1.0 corresponds to full destruction of information in init image
75-
-H, --height H image height, in pixel space (default: 512)
76-
-W, --width W image width, in pixel space (default: 512)
77-
--rng {std_default, cuda} RNG (default: cuda)
78-
-s SEED, --seed SEED RNG seed (default: 42, use random seed for < 0)
79-
-b, --batch-count COUNT number of images to generate
80-
--prediction {eps, v, edm_v, sd3_flow, flux_flow} Prediction type override
81-
--clip-skip N ignore last layers of CLIP network; 1 ignores none, 2 ignores one layer (default: -1)
82-
<= 0 represents unspecified, will be 1 for SD1.x, 2 for SD2.x
83-
--vae-tiling process vae in tiles to reduce memory usage
84-
--vae-tile-size [X]x[Y] tile size for vae tiling (default: 32x32)
85-
--vae-relative-tile-size [X]x[Y] relative tile size for vae tiling, in fraction of image size if < 1, in number of tiles per dim if >=1 (overrides --vae-tile-size)
86-
--vae-tile-overlap OVERLAP tile overlap for vae tiling, in fraction of tile size (default: 0.5)
87-
--force-sdxl-vae-conv-scale force use of conv scale on sdxl vae
88-
--vae-on-cpu keep vae in cpu (for low vram)
89-
--clip-on-cpu keep clip in cpu (for low vram)
90-
--diffusion-fa use flash attention in the diffusion model (for low vram)
91-
Might lower quality, since it implies converting k and v to f16.
92-
This might crash if it is not supported by the backend.
93-
--diffusion-conv-direct use Conv2d direct in the diffusion model
94-
This might crash if it is not supported by the backend.
95-
--vae-conv-direct use Conv2d direct in the vae model (should improve the performance)
96-
This might crash if it is not supported by the backend.
97-
--control-net-cpu keep controlnet in cpu (for low vram)
98-
--canny apply canny preprocessor (edge detection)
99-
--color colors the logging tags according to level
100-
--chroma-disable-dit-mask disable dit mask for chroma
101-
--chroma-enable-t5-mask enable t5 mask for chroma
102-
--chroma-t5-mask-pad PAD_SIZE t5 mask pad size of chroma
103-
--video-frames video frames (default: 1)
104-
--fps fps (default: 24)
105-
--moe-boundary BOUNDARY timestep boundary for Wan2.2 MoE model. (default: 0.875)
106-
only enabled if `--high-noise-steps` is set to -1
107-
--flow-shift SHIFT shift value for Flow models like SD3.x or WAN (default: auto)
108-
--vace-strength wan vace strength
109-
--photo-maker path to PHOTOMAKER model
110-
--pm-id-images-dir [DIR] path to PHOTOMAKER input id images dir
111-
--pm-id-embed-path [PATH] path to PHOTOMAKER v2 id embed
112-
--pm-style-strength strength for keeping PHOTOMAKER input identity (default: 20)
113-
-v, --verbose print extra info
6+
Options:
7+
-m, --model <string> path to full model
8+
--clip_l <string> path to the clip-l text encoder
9+
--clip_g <string> path to the clip-g text encoder
10+
--clip_vision <string> path to the clip-vision encoder
11+
--t5xxl <string> path to the t5xxl text encoder
12+
--qwen2vl <string> path to the qwen2vl text encoder
13+
--qwen2vl_vision <string> path to the qwen2vl vit
14+
--diffusion-model <string> path to the standalone diffusion model
15+
--high-noise-diffusion-model <string> path to the standalone high noise diffusion model
16+
--vae <string> path to standalone vae model
17+
--taesd <string> path to taesd. Using Tiny AutoEncoder for fast decoding (low quality)
18+
--control-net <string> path to control net model
19+
--embd-dir <string> embeddings directory
20+
--lora-model-dir <string> lora model directory
21+
-i, --init-img <string> path to the init image
22+
--end-img <string> path to the end image, required by flf2v
23+
--tensor-type-rules <string> weight type per tensor pattern (example: "^vae\.=f16,model\.=q8_0")
24+
--photo-maker <string> path to PHOTOMAKER model
25+
--pm-id-images-dir <string> path to PHOTOMAKER input id images dir
26+
--pm-id-embed-path <string> path to PHOTOMAKER v2 id embed
27+
--mask <string> path to the mask image
28+
--control-image <string> path to control image, control net
29+
--control-video <string> path to control video frames, It must be a directory path. The video frames inside should be stored as images in
30+
lexicographical (character) order. For example, if the control video path is
31+
`frames`, the directory contain images such as 00.png, 01.png, ... etc.
32+
-o, --output <string> path to write result image to (default: ./output.png)
33+
-p, --prompt <string> the prompt to render
34+
-n, --negative-prompt <string> the negative prompt (default: "")
35+
--upscale-model <string> path to esrgan model.
36+
-t, --threads <int> number of threads to use during computation (default: -1). If threads <= 0, then threads will be set to the number of
37+
CPU physical cores
38+
--upscale-repeats <int> Run the ESRGAN upscaler this many times (default: 1)
39+
-H, --height <int> image height, in pixel space (default: 512)
40+
-W, --width <int> image width, in pixel space (default: 512)
41+
--steps <int> number of sample steps (default: 20)
42+
--high-noise-steps <int> (high noise) number of sample steps (default: -1 = auto)
43+
--clip-skip <int> ignore last layers of CLIP network; 1 ignores none, 2 ignores one layer (default: -1). <= 0 represents unspecified,
44+
will be 1 for SD1.x, 2 for SD2.x
45+
-b, --batch-count <int> batch count
46+
--chroma-t5-mask-pad <int> t5 mask pad size of chroma
47+
--video-frames <int> video frames (default: 1)
48+
--fps <int> fps (default: 24)
49+
--timestep-shift <int> shift timestep for NitroFusion models (default: 0). recommended N for NitroSD-Realism around 250 and 500 for
50+
NitroSD-Vibrant
51+
--cfg-scale <float> unconditional guidance scale: (default: 7.0)
52+
--img-cfg-scale <float> image guidance scale for inpaint or instruct-pix2pix models: (default: same as --cfg-scale)
53+
--guidance <float> distilled guidance scale for models with guidance input (default: 3.5)
54+
--slg-scale <float> skip layer guidance (SLG) scale, only for DiT models: (default: 0). 0 means disabled, a value of 2.5 is nice for sd3.5
55+
medium
56+
--skip-layer-start <float> SLG enabling point (default: 0.01)
57+
--skip-layer-end <float> SLG disabling point (default: 0.2)
58+
--eta <float> eta in DDIM, only for DDIM and TCD (default: 0)
59+
--high-noise-cfg-scale <float> (high noise) unconditional guidance scale: (default: 7.0)
60+
--high-noise-img-cfg-scale <float> (high noise) image guidance scale for inpaint or instruct-pix2pix models (default: same as --cfg-scale)
61+
--high-noise-guidance <float> (high noise) distilled guidance scale for models with guidance input (default: 3.5)
62+
--high-noise-slg-scale <float> (high noise) skip layer guidance (SLG) scale, only for DiT models: (default: 0)
63+
--high-noise-skip-layer-start <float> (high noise) SLG enabling point (default: 0.01)
64+
--high-noise-skip-layer-end <float> (high noise) SLG disabling point (default: 0.2)
65+
--high-noise-eta <float> (high noise) eta in DDIM, only for DDIM and TCD (default: 0)
66+
--strength <float> strength for noising/unnoising (default: 0.75)
67+
--pm-style-strength <float>
68+
--control-strength <float> strength to apply Control Net (default: 0.9). 1.0 corresponds to full destruction of information in init image
69+
--moe-boundary <float> timestep boundary for Wan2.2 MoE model. (default: 0.875). Only enabled if `--high-noise-steps` is set to -1
70+
--flow-shift <float> shift value for Flow models like SD3.x or WAN (default: auto)
71+
--vace-strength <float> wan vace strength
72+
--vae-tile-overlap <float> tile overlap for vae tiling, in fraction of tile size (default: 0.5)
73+
--vae-tiling process vae in tiles to reduce memory usage
74+
--force-sdxl-vae-conv-scale force use of conv scale on sdxl vae
75+
--offload-to-cpu place the weights in RAM to save VRAM, and automatically load them into VRAM when needed
76+
--control-net-cpu keep controlnet in cpu (for low vram)
77+
--clip-on-cpu keep clip in cpu (for low vram)
78+
--vae-on-cpu keep vae in cpu (for low vram)
79+
--diffusion-fa use flash attention in the diffusion model
80+
--diffusion-conv-direct use ggml_conv2d_direct in the diffusion model
81+
--vae-conv-direct use ggml_conv2d_direct in the vae model
82+
--canny apply canny preprocessor (edge detection)
83+
-v, --verbose print extra info
84+
--color colors the logging tags according to level
85+
--chroma-disable-dit-mask disable dit mask for chroma
86+
--chroma-enable-t5-mask enable t5 mask for chroma
87+
--increase-ref-index automatically increase the indices of references images based on the order they are listed (starting with 1).
88+
--disable-auto-resize-ref-image disable auto resize of ref images
89+
-M, --mode run mode, one of [img_gen, vid_gen, upscale, convert], default: img_gen
90+
--type weight type (examples: f32, f16, q4_0, q4_1, q5_0, q5_1, q8_0, q2_K, q3_K, q4_K). If not specified, the default is the
91+
type of the weight file
92+
--rng RNG, one of [std_default, cuda], default: cuda
93+
-s, --seed RNG seed (default: 42, use random seed for < 0)
94+
--sampling-method sampling method, one of [euler, euler_a, heun, dpm2, dpm++2s_a, dpm++2m, dpm++2mv2, ipndm, ipndm_v, lcm, ddim_trailing,
95+
tcd] (default: euler for Flux/SD3/Wan, euler_a otherwise)
96+
--prediction prediction type override, one of [eps, v, edm_v, sd3_flow, flux_flow]
97+
--scheduler denoiser sigma scheduler, one of [discrete, karras, exponential, ays, gits, smoothstep, sgm_uniform, simple], default:
98+
discrete
99+
--skip-layers layers to skip for SLG steps (default: [7,8,9])
100+
--high-noise-sampling-method (high noise) sampling method, one of [euler, euler_a, heun, dpm2, dpm++2s_a, dpm++2m, dpm++2mv2, ipndm, ipndm_v, lcm,
101+
ddim_trailing, tcd] default: euler for Flux/SD3/Wan, euler_a otherwise
102+
--high-noise-scheduler (high noise) denoiser sigma scheduler, one of [discrete, karras, exponential, ays, gits, smoothstep, sgm_uniform,
103+
simple], default: discrete
104+
--high-noise-skip-layers (high noise) layers to skip for SLG steps (default: [7,8,9])
105+
-r, --ref-image reference image for Flux Kontext models (can be used multiple times)
106+
-h, --help show this help message and exit
107+
--vae-tile-size tile size for vae tiling, format [X]x[Y] (default: 32x32)
108+
--vae-relative-tile-size relative tile size for vae tiling, format [X]x[Y], in fraction of image size if < 1, in number of tiles per dim if >=1
109+
(overrides --vae-tile-size)
114110
```

0 commit comments

Comments
 (0)