You're looking at a specific version of this model. Jump to the model overview.
simpletuner /minimax-music3:9740bd64
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 matching .wav/.mp3/.flac, .txt style captions, and .lyrics files
|
|
| regularisation_data |
string
|
Optional audio/.txt/.lyrics archive for frozen-base prior preservation
|
|
| trigger_word |
string
|
|
Optional instance prompt replacing per-track captions
|
| max_train_steps |
integer
|
750
Min: 1 Max: 50000 |
Optimizer updates
|
| checkpoint_interval |
integer
|
250
Min: 1 Max: 5000 |
Save every N updates
|
| lora_rank |
integer
|
64
Min: 1 Max: 256 |
LoRA rank
|
| lora_dropout |
number
|
0
Max: 0.5 |
LoRA dropout
|
| learning_rate |
number
|
0.00008
Min: 0.0000001 Max: 0.001 |
Learning rate
|
| window_mode |
None
|
continuation
|
Continuation supervises the end of a context window; prefix and random supervise the entire selected window
|
| continuation_crop_mode |
None
|
full
|
Full keeps the song opening; random may start later while retaining causal context
|
| caption_dropout |
number
|
0
Max: 1 |
Probability of omitting style captions while retaining lyrics
|
| max_audio_duration |
number
|
30
Min: 5.12 Max: 300 |
Maximum visible training window in seconds
|
| enable_regional_compile |
boolean
|
True
|
Compile repeated transformer blocks
|
| attention_mechanism |
None
|
kohaku-fa-auto
|
Attention backend
|
| gradient_checkpointing |
boolean
|
True
|
Recompute language-model activations to reduce memory use
|
| 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
|
Output schema
The shape of the response you’ll get when you run this model with an API.
{'format': 'uri', 'title': 'Output', 'type': 'string'}