[Feature Request][API]: I would like to easily change the checkpoint from the API

[siriux]

Is there an existing issue for this?

• [X] I have searched the existing issues and checked the recent builds/commits

What would your feature do ?

I would like to be able to get a list of the available checkpoints from the API, and then change the current checkpoint also from the API in a simple and clear way more inline with the new /sdapi/v1/txt2img and /sdapi/v1/img2img APIs.

Currently this is possible by copying the same request the interface does (using /api/predict/) but the parameters to provide are not simple or clear.

This is an example:

{"data":[{"value":"sd-v1-5.ckpt [81761151]","__type__":"update"},"{\"samples_save\": false, \"samples_format\": \"png\", \"samples_filename_pattern\": \"\", \"save_images_add_number\": true, \"grid_save\": false, \"grid_format\": \"png\", \"grid_extended_filename\": false, \"grid_only_if_multiple\": false, \"grid_prevent_empty_spots\": false, \"n_rows\": -1, \"enable_pnginfo\": true, \"save_txt\": true, \"save_images_before_face_restoration\": true, \"jpeg_quality\": 80, \"export_for_4chan\": false, \"use_original_name_batch\": false, \"save_selected_only\": true, \"do_not_add_watermark\": true, \"outdir_samples\": \"outputs/\", \"outdir_txt2img_samples\": \"outputs/txt2img-images\", \"outdir_img2img_samples\": \"outputs/img2img-images\", \"outdir_extras_samples\": \"outputs/extras-images\", \"outdir_grids\": \"\", \"outdir_txt2img_grids\": \"outputs/txt2img-grids\", \"outdir_img2img_grids\": \"outputs/img2img-grids\", \"outdir_save\": \"outputs/\", \"save_to_dirs\": false, \"grid_save_to_dirs\": false, \"use_save_to_dirs_for_ui\": false, \"directories_filename_pattern\": \"\", \"directories_max_prompt_words\": 8, \"ESRGAN_tile\": 192, \"ESRGAN_tile_overlap\": 8, \"realesrgan_enabled_models\": [\"R-ESRGAN x4+\", \"R-ESRGAN x4+ Anime6B\"], \"SWIN_tile\": 192, \"SWIN_tile_overlap\": 8, \"ldsr_steps\": 100, \"upscaler_for_img2img\": null, \"use_scale_latent_for_hires_fix\": false, \"face_restoration_model\": \"CodeFormer\", \"code_former_weight\": 0.5, \"face_restoration_unload\": false, \"memmon_poll_rate\": 8, \"samples_log_stdout\": false, \"multiple_tqdm\": true, \"unload_models_when_training\": false, \"dataset_filename_word_regex\": \"\", \"dataset_filename_join_string\": \" \", \"training_image_repeats_per_epoch\": 1, \"training_write_csv_every\": 500.0, \"sd_model_checkpoint\": \"sd-v1-5.ckpt [81761151]\", \"sd_checkpoint_cache\": 0, \"sd_hypernetwork\": \"None\", \"sd_hypernetwork_strength\": 1, \"img2img_color_correction\": false, \"save_images_before_color_correction\": false, \"img2img_fix_steps\": false, \"enable_quantization\": false, \"enable_emphasis\": true, \"use_old_emphasis_implementation\": false, \"enable_batch_seeds\": true, \"comma_padding_backtrack\": 20, \"filter_nsfw\": false, \"CLIP_stop_at_last_layers\": 1, \"random_artist_categories\": [], \"interrogate_keep_models_in_memory\": false, \"interrogate_use_builtin_artists\": true, \"interrogate_return_ranks\": false, \"interrogate_clip_num_beams\": 1, \"interrogate_clip_min_length\": 24, \"interrogate_clip_max_length\": 48, \"interrogate_clip_dict_limit\": 1500.0, \"interrogate_deepbooru_score_threshold\": 0.5, \"deepbooru_sort_alpha\": true, \"deepbooru_use_spaces\": false, \"deepbooru_escape\": true, \"show_progressbar\": true, \"show_progress_every_n_steps\": 10, \"show_progress_grid\": true, \"return_grid\": true, \"do_not_show_images\": false, \"add_model_hash_to_info\": true, \"add_model_name_to_info\": false, \"disable_weights_auto_swap\": false, \"font\": \"\", \"js_modal_lightbox\": true, \"js_modal_lightbox_initially_zoomed\": true, \"show_progress_in_title\": true, \"quicksettings\": \"sd_model_checkpoint\", \"localization\": \"None\", \"hide_samplers\": [], \"eta_ddim\": 0, \"eta_ancestral\": 1, \"ddim_discretize\": \"uniform\", \"s_churn\": 0, \"s_tmin\": 0, \"s_noise\": 1, \"eta_noise_seed_delta\": 0}"],"is_generating":false,"duration":0.003006458282470703,"average_duration":0.003006458282470703}

Proposed workflow

1. Use the GET method /sdapi/v1/availablecheckpoints to get the list of available checkpoints
2. Use the POST method /sdapi/v1/checkpoint to set the current checkpoint

Additional information

No response

[Kilvoctu]

I'd like this, as well. Currently, the way to get available checkpoints and swap them is convoluted at best. The new API is fresh out of the oven, so there's a ways to go in developing it. Hoping checkpoints swapping could be prioritized. It's important enough that it's the very first element on the web ui, even before the prompt 🙂

[aliencaocao]

+1

[arttukataja]

+1

[aliencaocao]

BTW you can already do it in the options API, what I am still wishing for is the override settings part

[mnpenner]

@aliencaocao This options method? http://localhost:7860/docs#/default/set_config_sdapi_v1_options_post Do you have an example? The docs don't show the schema

Edit: Nvm, the payload should be something like this

export type PostOptions = {
    samples_save:                          boolean;
    samples_format:                        string;
    samples_filename_pattern:              string;
    save_images_add_number:                boolean;
    grid_save:                             boolean;
    grid_format:                           string;
    grid_extended_filename:                boolean;
    grid_only_if_multiple:                 boolean;
    grid_prevent_empty_spots:              boolean;
    n_rows:                                number;
    enable_pnginfo:                        boolean;
    save_txt:                              boolean;
    save_images_before_face_restoration:   boolean;
    save_images_before_highres_fix:        boolean;
    save_images_before_color_correction:   boolean;
    jpeg_quality:                          number;
    export_for_4chan:                      boolean;
    use_original_name_batch:               boolean;
    use_upscaler_name_as_suffix:           boolean;
    save_selected_only:                    boolean;
    do_not_add_watermark:                  boolean;
    temp_dir:                              string;
    clean_temp_dir_at_start:               boolean;
    outdir_samples:                        string;
    outdir_txt2img_samples:                string;
    outdir_img2img_samples:                string;
    outdir_extras_samples:                 string;
    outdir_grids:                          string;
    outdir_txt2img_grids:                  string;
    outdir_img2img_grids:                  string;
    outdir_save:                           string;
    save_to_dirs:                          boolean;
    grid_save_to_dirs:                     boolean;
    use_save_to_dirs_for_ui:               boolean;
    directories_filename_pattern:          string;
    directories_max_prompt_words:          number;
    ESRGAN_tile:                           number;
    ESRGAN_tile_overlap:                   number;
    realesrgan_enabled_models:             string[];
    upscaler_for_img2img:                  null;
    use_scale_latent_for_hires_fix:        boolean;
    ldsr_steps:                            number;
    ldsr_cached:                           boolean;
    SWIN_tile:                             number;
    SWIN_tile_overlap:                     number;
    face_restoration_model:                string;
    code_former_weight:                    number;
    face_restoration_unload:               boolean;
    memmon_poll_rate:                      number;
    samples_log_stdout:                    boolean;
    multiple_tqdm:                         boolean;
    unload_models_when_training:           boolean;
    pin_memory:                            boolean;
    save_optimizer_state:                  boolean;
    dataset_filename_word_regex:           string;
    dataset_filename_join_string:          string;
    training_image_repeats_per_epoch:      number;
    training_write_csv_every:              number;
    training_xattention_optimizations:     boolean;
    sd_model_checkpoint:                   string;
    sd_checkpoint_cache:                   number;
    sd_vae:                                string;
    sd_vae_as_default:                     boolean;
    sd_hypernetwork:                       string;
    sd_hypernetwork_strength:              number;
    inpainting_mask_weight:                number;
    initial_noise_multiplier:              number;
    img2img_color_correction:              boolean;
    img2img_fix_steps:                     boolean;
    img2img_background_color:              string;
    enable_quantization:                   boolean;
    enable_emphasis:                       boolean;
    use_old_emphasis_implementation:       boolean;
    enable_batch_seeds:                    boolean;
    comma_padding_backtrack:               number;
    CLIP_stop_at_last_layers:              number;
    random_artist_categories:              any[];
    interrogate_keep_models_in_memory:     boolean;
    interrogate_use_builtin_artists:       boolean;
    interrogate_return_ranks:              boolean;
    interrogate_clip_num_beams:            number;
    interrogate_clip_min_length:           number;
    interrogate_clip_max_length:           number;
    interrogate_clip_dict_limit:           number;
    interrogate_deepbooru_score_threshold: number;
    deepbooru_sort_alpha:                  boolean;
    deepbooru_use_spaces:                  boolean;
    deepbooru_escape:                      boolean;
    deepbooru_filter_tags:                 string;
    show_progressbar:                      boolean;
    show_progress_every_n_steps:           number;
    show_progress_type:                    string;
    show_progress_grid:                    boolean;
    return_grid:                           boolean;
    do_not_show_images:                    boolean;
    add_model_hash_to_info:                boolean;
    add_model_name_to_info:                boolean;
    disable_weights_auto_swap:             boolean;
    send_seed:                             boolean;
    send_size:                             boolean;
    font:                                  string;
    js_modal_lightbox:                     boolean;
    js_modal_lightbox_initially_zoomed:    boolean;
    show_progress_in_title:                boolean;
    quicksettings:                         string;
    localization:                          string;
    hide_samplers:                         any[];
    eta_ddim:                              number;
    eta_ancestral:                         number;
    ddim_discretize:                       string;
    s_churn:                               number;
    s_tmin:                                number;
    s_noise:                               number;
    eta_noise_seed_delta:                  number;
    disabled_extensions:                   string[];
    images_history_preload:                boolean;
    images_history_page_columns:           number;
    images_history_page_rows:              number;
    images_history_pages_perload:          number;
    wildcards_same_seed:                   boolean;
}

(reverse-engineered from the GET response)

The sd_model_checkpoint needs to be set to the title from /sdapi/v1/sd-models

[aliencaocao]

No You can use options endpoint with key=option name and value=option value. Option name can be get from inspect element on the settings page Or if you use my PR, its override settings, not options. Key and values are same

[jdc4429]

Could someone explain better how to call the option from python to change the checkpoint via the API?

I got the key=sd_model_checkpoint and I believe the value would be the name of the checkpoint but how do you actually call the API to execute the command?

I tried calling this way but the sd_model_checkpoint is not recognized...

api = webuiapi.WebUIApi(host='127.0.0.1', port=7860, sampler='Euler a', steps=20)

wprompt = sys.argv[1] nprompt = sys.argv[2] selected_model = sys.argv[3]

result1 = api.txt2img(prompt=wprompt, negative_prompt=nprompt, sd_model_checkpoint=selected_model, seed=-1, cfg_scale=8,

[itsidleboy]

Here you go guys -

import json
import requests
import io
import base64
from PIL import Image, PngImagePlugin

url = "http://127.0.0.1:7860"

payload = {
    "sd_model_checkpoint": "realisticVisionV51_v51VAE.safetensors [15012c538f]",
    "prompt": "A young girl with long hair and pink background",
    "steps": 30,
}

response = requests.post(url=f'{url}/sdapi/v1/txt2img', json=payload)

r = response.json()

for i in r['images']:
    image = Image.open(io.BytesIO(base64.b64decode(i.split(",",1)[0])))

    png_payload = {
        "image": "data:image/png;base64," + i
    }
    response2 = requests.post(url=f'{url}/sdapi/v1/png-info', json=png_payload)

    pnginfo = PngImagePlugin.PngInfo()
    pnginfo.add_text("parameters", response2.json().get("info"))
    image.save('output.png', pnginfo=pnginfo)
    ```