Skip to content

Commit 17d70b9

Browse files
leejetleejet
committed
docs: replace example option lists with help commands
1 parent 5db680c commit 17d70b9

2 files changed

Lines changed: 9 additions & 382 deletions

File tree

‎examples/cli/README.md‎

Lines changed: 4 additions & 199 deletions
Original file line numberDiff line numberDiff line change
@@ -1,204 +1,9 @@
1-
# Run
1+
# Usage
22

3-
```
4-
usage: ./bin/sd-cli [options]
5-
6-
CLI Options:
7-
-o, --output <string> path to write result image to. you can use printf-style %d format specifiers for image
8-
sequences (default: ./output.png) (eg. output_%03d.png). Single-file video outputs
9-
support .avi, .webm, and animated .webp
10-
--image <string> path to the image to inspect (for metadata mode)
11-
--metadata-format <string> metadata output format, one of [text, json] (default: text)
12-
--preview-path <string> path to write preview image to (default: ./preview.png). Multi-frame previews support
13-
.avi, .webm, and animated .webp
14-
--preview-interval <int> interval in denoising steps between consecutive updates of the image preview file
15-
(default is 1, meaning updating at every step)
16-
--output-begin-idx <int> starting index for output image sequence, must be non-negative (default 0 if specified
17-
%d in output path, 1 otherwise)
18-
--canny apply canny preprocessor (edge detection)
19-
--convert-name convert tensor name (for convert mode)
20-
-v, --verbose print extra info
21-
--color colors the logging tags according to level
22-
--taesd-preview-only prevents usage of taesd for decoding the final image. (for use with --preview tae)
23-
--preview-noisy enables previewing noisy inputs of the models rather than the denoised outputs
24-
--metadata-raw include raw hex previews for unparsed metadata payloads
25-
--metadata-brief truncate long metadata text values in text output
26-
--metadata-all include structural/container entries such as IHDR, IDAT, and non-metadata JPEG segments
27-
-M, --mode run mode, one of [img_gen, vid_gen, upscale, convert, metadata], default: img_gen
28-
--preview preview method. must be one of the following [none, proj, tae, vae] (default is none)
29-
-h, --help show this help message and exit
3+
For detailed command-line arguments, run:
304

31-
Context Options:
32-
-m, --model <string> path to full model
33-
--clip_l <string> path to the clip-l text encoder
34-
--clip_g <string> path to the clip-g text encoder
35-
--clip_vision <string> path to the clip-vision encoder
36-
--t5xxl <string> path to the t5xxl text encoder
37-
--llm <string> path to the llm text encoder. For example: (qwenvl2.5 for qwen-image,
38-
mistral-small3.2 for flux2, ...)
39-
--llm_vision <string> path to the llm vit
40-
--qwen2vl <string> alias of --llm. Deprecated.
41-
--qwen2vl_vision <string> alias of --llm_vision. Deprecated.
42-
--diffusion-model <string> path to the standalone diffusion model
43-
--high-noise-diffusion-model <string> path to the standalone high noise diffusion model
44-
--uncond-diffusion-model <string> path to the standalone unconditional diffusion model, currently used by
45-
Ideogram4 CFG
46-
--vae <string> path to standalone vae model
47-
--taesd <string> path to taesd. Using Tiny AutoEncoder for fast decoding (low quality)
48-
--tae <string> alias of --taesd
49-
--control-net <string> path to control net model
50-
--embd-dir <string> embeddings directory
51-
--lora-model-dir <string> lora model directory
52-
--hires-upscalers-dir <string> highres fix upscaler model directory
53-
--tensor-type-rules <string> weight type per tensor pattern (example: "^vae\.=f16,model\.=q8_0")
54-
--photo-maker <string> path to PHOTOMAKER model
55-
--upscale-model <string> path to esrgan model.
56-
-t, --threads <int> number of threads to use during computation (default: -1). If threads <= 0,
57-
then threads will be set to the number of CPU physical cores
58-
--chroma-t5-mask-pad <int> t5 mask pad size of chroma
59-
--max-vram <float> maximum VRAM budget in GiB for graph-cut segmented execution. 0 disables
60-
graph splitting; a negative value auto-detects free VRAM, sparing the
61-
specified value (e.g. -0.5 will keep at least 0.5 GiB free)
62-
--force-sdxl-vae-conv-scale force use of conv scale on sdxl vae
63-
--offload-to-cpu place the weights in RAM to save VRAM, and automatically load them into VRAM
64-
when needed
65-
--mmap whether to memory-map model
66-
--control-net-cpu deprecated; use --backend controlnet=cpu
67-
--clip-on-cpu deprecated; use --backend te=cpu
68-
--vae-on-cpu deprecated; use --backend vae=cpu
69-
--fa use flash attention
70-
--diffusion-fa use flash attention in the diffusion model only
71-
--diffusion-conv-direct use ggml_conv2d_direct in the diffusion model
72-
--vae-conv-direct use ggml_conv2d_direct in the vae model
73-
--circular enable circular padding for convolutions
74-
--circularx enable circular RoPE wrapping on x-axis (width) only
75-
--circulary enable circular RoPE wrapping on y-axis (height) only
76-
--chroma-disable-dit-mask disable dit mask for chroma
77-
--qwen-image-zero-cond-t enable zero_cond_t for qwen image
78-
--chroma-enable-t5-mask enable t5 mask for chroma
79-
--type weight type (examples: f32, f16, q4_0, q4_1, q5_0, q5_1, q8_0, q2_K, q3_K,
80-
q4_K). If not specified, the default is the type of the weight file
81-
--rng RNG, one of [std_default, cuda, cpu], default: cuda(sd-webui), cpu(comfyui)
82-
--sampler-rng sampler RNG, one of [std_default, cuda, cpu]. If not specified, use --rng
83-
--prediction prediction type override, one of [eps, v, edm_v, sd3_flow, flux_flow,
84-
flux2_flow]
85-
--lora-apply-mode the way to apply LoRA, one of [auto, immediately, at_runtime], default is
86-
auto. In auto mode, if the model weights contain any quantized parameters,
87-
the at_runtime mode will be used; otherwise, immediately will be used.The
88-
immediately mode may have precision and compatibility issues with quantized
89-
parameters, but it usually offers faster inference speed and, in some cases,
90-
lower memory usage. The at_runtime mode, on the other hand, is exactly the
91-
opposite.
92-
93-
Generation Options:
94-
-p, --prompt <string> the prompt to render
95-
-n, --negative-prompt <string> the negative prompt (default: "")
96-
-i, --init-img <string> path to the init image
97-
--end-img <string> path to the end image, required by flf2v
98-
--mask <string> path to the mask image
99-
--control-image <string> path to control image, control net
100-
--control-video <string> path to control video frames, It must be a directory path. The video frames
101-
inside should be stored as images in lexicographical (character) order. For
102-
example, if the control video path is `frames`, the directory contain images
103-
such as 00.png, 01.png, ... etc.
104-
--pm-id-images-dir <string> path to PHOTOMAKER input id images dir
105-
--pm-id-embed-path <string> path to PHOTOMAKER v2 id embed
106-
--hires-upscaler <string> highres fix upscaler, Lanczos, Nearest, Latent, Latent (nearest), Latent
107-
(nearest-exact), Latent (antialiased), Latent (bicubic), Latent (bicubic
108-
antialiased), or a model name under --hires-upscalers-dir (default: Latent)
109-
--extra-sample-args <string> extra sampler/scheduler/guidance args, key=value list. APG supports apg_eta,
110-
apg_momentum, apg_norm_threshold, apg_norm_threshold_smoothing; SLG supports
111-
slg_uncond; lcm supports noise_clip_std, noise_scale_start, noise_scale_end;
112-
ltx2 supports max_shift, base_shift, stretch, terminal; euler_ge supports gamma
113-
--extra-tiling-args <string> extra VAE tiling args, key=value list. LTX video VAE supports
114-
temporal_tile_frames (default: 4), temporal_tile_overlap (default: 1)
115-
-H, --height <int> image height, in pixel space (default: 512)
116-
-W, --width <int> image width, in pixel space (default: 512)
117-
--steps <int> number of sample steps (default: 20)
118-
--high-noise-steps <int> (high noise) number of sample steps (default: -1 = auto)
119-
--clip-skip <int> ignore last layers of CLIP network; 1 ignores none, 2 ignores one layer
120-
(default: -1). <= 0 represents unspecified, will be 1 for SD1.x, 2 for SD2.x
121-
-b, --batch-count <int> batch count
122-
--video-frames <int> video frames (default: 1)
123-
--fps <int> fps (default: 24)
124-
--timestep-shift <int> shift timestep for NitroFusion models (default: 0). recommended N for
125-
NitroSD-Realism around 250 and 500 for NitroSD-Vibrant
126-
--upscale-repeats <int> Run the ESRGAN upscaler this many times (default: 1)
127-
--upscale-tile-size <int> tile size for ESRGAN upscaling (default: 128)
128-
--hires-width <int> highres fix target width, 0 to use --hires-scale (default: 0)
129-
--hires-height <int> highres fix target height, 0 to use --hires-scale (default: 0)
130-
--hires-steps <int> highres fix second pass sample steps, 0 to reuse --steps (default: 0)
131-
--hires-upscale-tile-size <int> highres fix upscaler tile size, reserved for model-backed upscalers (default:
132-
128)
133-
--cfg-scale <float> unconditional guidance scale: (default: 7.0)
134-
--img-cfg-scale <float> image guidance scale for inpaint or image edit models: (default: same as
135-
--cfg-scale)
136-
--guidance <float> distilled guidance scale for models with guidance input (default: 3.5)
137-
--slg-scale <float> skip layer guidance (SLG) scale, only for DiT models: (default: 0). 0 means
138-
disabled, a value of 2.5 is nice for sd3.5 medium
139-
--skip-layer-start <float> SLG enabling point (default: 0.01)
140-
--skip-layer-end <float> SLG disabling point (default: 0.2)
141-
--eta <float> noise multiplier (default: 0 for ddim_trailing, tcd, res_multistep and
142-
res_2s; 1 for euler_a, er_sde and dpm++2s_a)
143-
--flow-shift <float> shift value for Flow models like SD3.x or WAN (default: auto)
144-
--high-noise-cfg-scale <float> (high noise) unconditional guidance scale: (default: 7.0)
145-
--high-noise-img-cfg-scale <float> (high noise) image guidance scale for inpaint or image edit models (default:
146-
same as --cfg-scale)
147-
--high-noise-guidance <float> (high noise) distilled guidance scale for models with guidance input
148-
(default: 3.5)
149-
--high-noise-slg-scale <float> (high noise) skip layer guidance (SLG) scale, only for DiT models: (default:
150-
0)
151-
--high-noise-skip-layer-start <float> (high noise) SLG enabling point (default: 0.01)
152-
--high-noise-skip-layer-end <float> (high noise) SLG disabling point (default: 0.2)
153-
--high-noise-eta <float> (high noise) noise multiplier (default: 0 for ddim_trailing, tcd,
154-
res_multistep and res_2s; 1 for euler_a, er_sde and dpm++2s_a)
155-
--strength <float> strength for noising/unnoising (default: 0.75)
156-
--pm-style-strength <float>
157-
--control-strength <float> strength to apply Control Net (default: 0.9). 1.0 corresponds to full
158-
destruction of information in init image
159-
--moe-boundary <float> timestep boundary for Wan2.2 MoE model. (default: 0.875). Only enabled if
160-
`--high-noise-steps` is set to -1
161-
--vace-strength <float> wan vace strength
162-
--vae-tile-overlap <float> tile overlap for vae tiling, in fraction of tile size (default: 0.5)
163-
--hires-scale <float> highres fix scale when target size is not set (default: 2.0)
164-
--hires-denoising-strength <float> highres fix second pass denoising strength (default: 0.7)
165-
--increase-ref-index automatically increase the indices of references images based on the order
166-
they are listed (starting with 1).
167-
--disable-auto-resize-ref-image disable auto resize of ref images
168-
--disable-image-metadata do not embed generation metadata on image files
169-
--vae-tiling process vae in tiles to reduce memory usage
170-
--temporal-tiling enable temporal tiling for LTX video VAE decode
171-
--hires enable highres fix
172-
-s, --seed RNG seed (default: 42, use random seed for < 0)
173-
--sampling-method sampling method, one of [euler, euler_a, heun, dpm2, dpm++2s_a, dpm++2m,
174-
dpm++2mv2, ipndm, ipndm_v, lcm, ddim_trailing, tcd, res_multistep, res_2s,
175-
er_sde, euler_cfg_pp, euler_a_cfg_pp] (default: euler for Flux/SD3/Wan, euler_a otherwise)
176-
--high-noise-sampling-method (high noise) sampling method, one of [euler, euler_a, heun, dpm2, dpm++2s_a,
177-
dpm++2m, dpm++2mv2, ipndm, ipndm_v, lcm, ddim_trailing, tcd, res_multistep,
178-
res_2s, er_sde, euler_cfg_pp, euler_a_cfg_pp] default: euler for Flux/SD3/Wan, euler_a otherwise
179-
--scheduler denoiser sigma scheduler, one of [discrete, karras, exponential, ays, gits,
180-
smoothstep, sgm_uniform, simple, kl_optimal, lcm, bong_tangent, ltx2], default:
181-
model-specific
182-
--sigmas custom sigma values for the sampler, comma-separated (e.g.,
183-
"14.61,7.8,3.5,0.0").
184-
--hires-sigmas custom sigma values for the highres fix second pass, comma-separated (e.g.,
185-
"0.85,0.725,0.421875,0.0").
186-
--skip-layers layers to skip for SLG steps (default: [7,8,9])
187-
--high-noise-skip-layers (high noise) layers to skip for SLG steps (default: [7,8,9])
188-
-r, --ref-image reference image for Flux Kontext models (can be used multiple times)
189-
--cache-mode caching method: 'easycache' (DiT), 'ucache' (UNET),
190-
'dbcache'/'taylorseer'/'cache-dit' (DiT block-level), 'spectrum' (UNET/DiT
191-
Chebyshev+Taylor forecasting)
192-
--cache-option named cache params (key=value format, comma-separated). easycache/ucache:
193-
threshold=,start=,end=,decay=,relative=,reset=; dbcache/taylorseer/cache-dit:
194-
Fn=,Bn=,threshold=,warmup=; spectrum: w=,m=,lam=,window=,flex=,warmup=,stop=.
195-
Examples: "threshold=0.25" or "threshold=1.5,reset=0"
196-
--scm-mask SCM steps mask for cache-dit: comma-separated 0/1 (e.g.,
197-
"1,1,1,0,0,1,0,0,1,0") - 1=compute, 0=can cache
198-
--scm-policy SCM policy: 'dynamic' (default) or 'static'
199-
--vae-tile-size tile size for vae tiling, format [X]x[Y] (default: 32x32)
200-
--vae-relative-tile-size relative tile size for vae tiling, format [X]x[Y], in fraction of image size
201-
if < 1, in number of tiles per dim if >=1 (overrides --vae-tile-size)
5+
```bash
6+
./bin/sd-cli -h
2027
```
2038

2049
Metadata mode inspects PNG/JPEG container metadata without loading any model:

0 commit comments

Comments
 (0)