Repository Issue Activity (beta)

#1320 Investigate: nanovllm backend may work with Modal GPU snapshots

#1294 Acemusic.ai: Remix not working at all

#1295 follow-up: design bounded diffusion diagnostics

#1239 start_gradio_ui.sh fails with "LANGUAGE: unbound variable"

#1274 Fix NaN diffusion on pre-Ampere NVIDIA GPUs (Pascal/Turing) by using float32 fallback instead of float16

#1253 Issues with remix mode

#1317 LoRA training output silently lost: relative ./lora_output default is never resolved or shown, and isn't mounted in docker-compose.yml

#614 in training, path to folder error: Rejected unsafe directory path

#1277 [Feature req.] Passing chord progression(s) in Complete mode

#1271 Cover-mode cover_noise_strength silently persists into text2music, producing full-duration static

#1280 Dead project

#1299 Why does the SFT XL sounds like garble noise?

#1306 Commercial use of ACEMusic.ai API outputs in a paid application

#1288 sft checkpoints (2B and XL) output noise via the Python pipeline. Identical weights work in ComfyUI

#1227 Windows ROCm: vector_quantize_pytorch imports torch.distributed which is unavailable in ROCm Windows builds

#1286 complete/lego tasks ignore src_audio — LM-generated codes silently pre-empt real source-audio conditioning

#1291 bug(ui): Repaint does not pass explicit chunk mask mode

#1281 First lazy-init request blocks the entire API server (health checks included) until model loading finishes

#1259 Non-distilled models (xl-base / xl-sft) produce garbled audio on Apple Silicon — MLX and PyTorch-MPS; distilled turbo models fine

#1289 bug: chunk_mask_mode="auto" is coerced to a boolean mask

#1248 ACESTEP XL 1.5 Full Inference Tutorial - Working on LoRA Training Tutorial As Well - Fully edited 75 minutes and + written tutorial

#1144 Feature Request: AutoPlay

#704 For Google colab FREE plan (T4, 12.7GB RAM, 15GB VRAM)

#1114 [Important] Technical Feedback on ACE-Step 1.5 XL (Music Quality & Expressiveness)

#1261 Default Gradio generation format/bitrate should probably be different than MP3/128kbps