simpletuner/minimax-h3 | API reference

Public

10 runs

Run simpletuner/minimax-h3 with an API

Use one of our client libraries to get started quickly. Clicking on a library will take you to the Playground tab where you can tweak different inputs, see the results, and copy the corresponding code to use in your own project.

Input schema

The fields you can use to run this model with an API. If you don't give a value for a field its default value will be used.

Field	Type	Default value	Description
train_data	string		ZIP or tar of audio/.txt/.lyrics triplets OR videos with .txt captions
trigger_word	string		Optional instance prompt replacing per-track captions
max_train_steps	integer	1000 Min: 1 Max: 50000	Optimizer updates
checkpoint_interval	integer	250 Min: 1 Max: 5000	Save every N updates
lora_rank	integer	16 Min: 1 Max: 256	LoRA rank
lora_dropout	number	0 Max: 0.5	LoRA dropout
learning_rate	number	0.00006 Min: 0.0000001 Max: 0.001	Learning rate
resolution	None	480	Video pixel-area resolution
video_frames	None	39	Video training frames; optional previews use 124 frames
validation_prompt	string		Optional prompt rendered at checkpoint saves
validation_lyrics	string	[instrumental]	Lyrics for optional validation audio
max_audio_duration	number	30 Min: 3 Max: 300	Maximum training audio duration in seconds
enable_regional_compile	boolean	True	Compile repeated transformer blocks
attention_mechanism	None	kohaku-fa-auto	Attention backend
gradient_checkpointing_interval	integer	2 Min: 1 Max: 16	Checkpoint interval in transformer blocks
gradient_checkpointing_segment_stride	integer	4 Min: 1 Max: 64	Spacing between checkpointed segments; at least the checkpoint interval
dataloader_prefetch	boolean	True	Prepare upcoming training batches in the background
seed	integer	42	Training seed
hub_model_id	string		Optional Hugging Face repository for the trained adapter
hf_token	string		Write-capable Hugging Face token for uploading the adapter
push_checkpoints_to_hub	boolean	True	Upload intermediate checkpoints to the Hub in the background

{
  "type": "object",
  "title": "Input",
  "required": [
    "train_data"
  ],
  "properties": {
    "seed": {
      "type": "integer",
      "title": "Seed",
      "default": 42,
      "minimum": 0,
      "x-order": 17,
      "description": "Training seed"
    },
    "hf_token": {
      "type": "string",
      "title": "Hf Token",
      "format": "password",
      "x-order": 19,
      "nullable": true,
      "writeOnly": true,
      "description": "Write-capable Hugging Face token for uploading the adapter",
      "x-cog-secret": true
    },
    "lora_rank": {
      "type": "integer",
      "title": "Lora Rank",
      "default": 16,
      "maximum": 256,
      "minimum": 1,
      "x-order": 4,
      "description": "LoRA rank"
    },
    "resolution": {
      "enum": [
        256,
        384,
        480
      ],
      "type": "integer",
      "title": "resolution",
      "description": "Video pixel-area resolution",
      "default": 480,
      "x-order": 7
    },
    "train_data": {
      "type": "string",
      "title": "Train Data",
      "format": "uri",
      "x-order": 0,
      "description": "ZIP or tar of audio/.txt/.lyrics triplets OR videos with .txt captions"
    },
    "hub_model_id": {
      "type": "string",
      "title": "Hub Model Id",
      "x-order": 18,
      "nullable": true,
      "description": "Optional Hugging Face repository for the trained adapter"
    },
    "lora_dropout": {
      "type": "number",
      "title": "Lora Dropout",
      "default": 0,
      "maximum": 0.5,
      "minimum": 0,
      "x-order": 5,
      "description": "LoRA dropout"
    },
    "trigger_word": {
      "type": "string",
      "title": "Trigger Word",
      "default": "",
      "x-order": 1,
      "description": "Optional instance prompt replacing per-track captions"
    },
    "video_frames": {
      "enum": [
        5,
        22,
        39
      ],
      "type": "integer",
      "title": "video_frames",
      "description": "Video training frames; optional previews use 124 frames",
      "default": 39,
      "x-order": 8
    },
    "learning_rate": {
      "type": "number",
      "title": "Learning Rate",
      "default": 6e-05,
      "maximum": 0.001,
      "minimum": 1e-07,
      "x-order": 6,
      "description": "Learning rate"
    },
    "max_train_steps": {
      "type": "integer",
      "title": "Max Train Steps",
      "default": 1000,
      "maximum": 50000,
      "minimum": 1,
      "x-order": 2,
      "description": "Optimizer updates"
    },
    "validation_lyrics": {
      "type": "string",
      "title": "Validation Lyrics",
      "default": "[instrumental]",
      "x-order": 10,
      "description": "Lyrics for optional validation audio"
    },
    "validation_prompt": {
      "type": "string",
      "title": "Validation Prompt",
      "default": "",
      "x-order": 9,
      "description": "Optional prompt rendered at checkpoint saves"
    },
    "max_audio_duration": {
      "type": "number",
      "title": "Max Audio Duration",
      "default": 30,
      "maximum": 300,
      "minimum": 3,
      "x-order": 11,
      "description": "Maximum training audio duration in seconds"
    },
    "attention_mechanism": {
      "enum": [
        "kohaku-fa-auto",
        "native-efficient",
        "diffusers"
      ],
      "type": "string",
      "title": "attention_mechanism",
      "description": "Attention backend",
      "default": "kohaku-fa-auto",
      "x-order": 13
    },
    "checkpoint_interval": {
      "type": "integer",
      "title": "Checkpoint Interval",
      "default": 250,
      "maximum": 5000,
      "minimum": 1,
      "x-order": 3,
      "description": "Save every N updates"
    },
    "dataloader_prefetch": {
      "type": "boolean",
      "title": "Dataloader Prefetch",
      "default": true,
      "x-order": 16,
      "description": "Prepare upcoming training batches in the background"
    },
    "enable_regional_compile": {
      "type": "boolean",
      "title": "Enable Regional Compile",
      "default": true,
      "x-order": 12,
      "description": "Compile repeated transformer blocks"
    },
    "push_checkpoints_to_hub": {
      "type": "boolean",
      "title": "Push Checkpoints To Hub",
      "default": true,
      "x-order": 20,
      "description": "Upload intermediate checkpoints to the Hub in the background"
    },
    "gradient_checkpointing_interval": {
      "type": "integer",
      "title": "Gradient Checkpointing Interval",
      "default": 2,
      "maximum": 16,
      "minimum": 1,
      "x-order": 14,
      "description": "Checkpoint interval in transformer blocks"
    },
    "gradient_checkpointing_segment_stride": {
      "type": "integer",
      "title": "Gradient Checkpointing Segment Stride",
      "default": 4,
      "maximum": 64,
      "minimum": 1,
      "x-order": 15,
      "description": "Spacing between checkpointed segments; at least the checkpoint interval"
    }
  }
}

Output schema

The shape of the response you’ll get when you run this model with an API.

Schema

{
  "type": "string",
  "title": "Output",
  "format": "uri"
}