Sebastian Krüger
d74a7cb7cb
All checks were successful
Build and Push RunPod Docker Image / build-and-push (push) Successful in 14s
- Remove custom PivoineDiffRhythmRun wrapper node - Add git patch file for ComfyUI_DiffRhythm __init__.py - Patch adds LlamaConfig fix at import time - Add arty script 'fix/diffrhythm-patch' to apply patch - Revert all workflows to use original DiffRhythmRun - Remove startup_patch.py and revert start.sh This approach is cleaner and more maintainable than wrapping the node. The patch directly fixes the tensor dimension mismatch (32 vs 64) in DiffRhythm's rotary position embeddings by ensuring num_attention_heads and num_key_value_heads are properly set based on hidden_size. References: - https://github.com/billwuhao/ComfyUI_DiffRhythm/issues/44 - https://github.com/billwuhao/ComfyUI_DiffRhythm/issues/48
144 lines
4.5 KiB
JSON
144 lines
4.5 KiB
JSON
{
|
|
"last_node_id": 3,
|
|
"last_link_id": 2,
|
|
"nodes": [
|
|
{
|
|
"id": 1,
|
|
"type": "DiffRhythmRun",
|
|
"pos": [100, 100],
|
|
"size": [400, 400],
|
|
"flags": {},
|
|
"order": 0,
|
|
"mode": 0,
|
|
"outputs": [
|
|
{
|
|
"name": "AUDIO",
|
|
"type": "AUDIO",
|
|
"links": [1, 2]
|
|
}
|
|
],
|
|
"properties": {
|
|
"Node name for S&R": "DiffRhythmRun"
|
|
},
|
|
"widgets_values": [
|
|
"cfm_model_v1_2.pt",
|
|
"",
|
|
true,
|
|
"euler",
|
|
30,
|
|
1,
|
|
"speed",
|
|
-1,
|
|
"randomize",
|
|
false,
|
|
"[-1, 20], [60, -1]"
|
|
],
|
|
"title": "DiffRhythm Random Generation (No Prompt)"
|
|
},
|
|
{
|
|
"id": 2,
|
|
"type": "PreviewAudio",
|
|
"pos": [600, 100],
|
|
"size": [300, 100],
|
|
"flags": {},
|
|
"order": 1,
|
|
"mode": 0,
|
|
"inputs": [
|
|
{
|
|
"name": "audio",
|
|
"type": "AUDIO",
|
|
"link": 1
|
|
}
|
|
],
|
|
"properties": {
|
|
"Node name for S&R": "PreviewAudio"
|
|
},
|
|
"title": "Preview Audio"
|
|
},
|
|
{
|
|
"id": 3,
|
|
"type": "SaveAudio",
|
|
"pos": [600, 250],
|
|
"size": [300, 100],
|
|
"flags": {},
|
|
"order": 2,
|
|
"mode": 0,
|
|
"inputs": [
|
|
{
|
|
"name": "audio",
|
|
"type": "AUDIO",
|
|
"link": 2
|
|
}
|
|
],
|
|
"properties": {
|
|
"Node name for S&R": "SaveAudio"
|
|
},
|
|
"widgets_values": [
|
|
"diffrhythm_random_output"
|
|
],
|
|
"title": "Save Audio"
|
|
}
|
|
],
|
|
"links": [
|
|
[1, 1, 0, 2, 0, "AUDIO"],
|
|
[2, 1, 0, 3, 0, "AUDIO"]
|
|
],
|
|
"groups": [],
|
|
"config": {},
|
|
"extra": {
|
|
"workflow_info": {
|
|
"name": "DiffRhythm Random Generation v1",
|
|
"description": "Random music generation without any prompt or guidance - pure AI creativity",
|
|
"version": "1.0.0",
|
|
"author": "valknar@pivoine.art",
|
|
"category": "text-to-music",
|
|
"tags": ["diffrhythm", "music-generation", "random", "no-prompt", "discovery"],
|
|
"requirements": {
|
|
"custom_nodes": ["ComfyUI_DiffRhythm"],
|
|
"models": ["ASLP-lab/DiffRhythm-1_2", "ASLP-lab/DiffRhythm-vae", "OpenMuQ/MuQ-MuLan-large", "OpenMuQ/MuQ-large-msd-iter", "FacebookAI/xlm-roberta-base"],
|
|
"vram_min": "12GB",
|
|
"vram_recommended": "16GB",
|
|
"system_deps": ["espeak-ng"]
|
|
},
|
|
"usage": {
|
|
"model": "cfm_model_v1_2.pt (DiffRhythm 1.2 - 95s generation)",
|
|
"style_prompt": "Empty string for random generation (no guidance)",
|
|
"unload_model": "Boolean to unload model after generation (default: true)",
|
|
"odeint_method": "ODE solver: euler (default)",
|
|
"steps": "Number of diffusion steps: 30 (default)",
|
|
"cfg": "Classifier-free guidance: 1 (minimal guidance for random output)",
|
|
"quality_or_speed": "Generation mode: speed (default)",
|
|
"seed": "-1 for random seed each generation, or specific number for reproducibility",
|
|
"edit": "false (no editing)",
|
|
"note": "NO prompt, NO guidance, NO reference audio - pure random generation"
|
|
},
|
|
"use_cases": [
|
|
"Discovery: Explore what the model can create without constraints",
|
|
"Inspiration: Generate unexpected musical ideas and styles",
|
|
"Testing: Quick way to verify model is working correctly",
|
|
"Ambient music: Create random background music for various uses",
|
|
"Sample generation: Generate large batches of diverse music samples"
|
|
],
|
|
"workflow_tips": [
|
|
"Run multiple times to discover different musical styles",
|
|
"Use seed=-1 for completely random output each time",
|
|
"Use fixed seed to reproduce interesting random results",
|
|
"Batch process: Run 10-20 times to find interesting compositions",
|
|
"Save any interesting results with their seed numbers",
|
|
"Empty style_prompt with cfg=1 produces truly random output"
|
|
],
|
|
"notes": [
|
|
"This workflow demonstrates DiffRhythm's ability to generate music without any input",
|
|
"All DiffRhythm parameters except model are at their most permissive settings",
|
|
"Empty string for style_prompt means no guidance from text",
|
|
"cfg=1 provides minimal guidance, maximizing randomness",
|
|
"Results can range from ambient to energetic, classical to electronic",
|
|
"Each generation is unique (with seed=-1)",
|
|
"Generation time: ~30-60 seconds on RTX 4090",
|
|
"Perfect for discovering unexpected musical combinations"
|
|
]
|
|
}
|
|
},
|
|
"version": 0.4
|
|
}
|