Hugging Face's logo

ehabnegm
/
s2pro-egy

Text-to-Speech
TensorBoard
Safetensors
Arabic
fish_qwen3_omni
egyptian-arabic
arz
fish-speech
lora
Model card Files
xet
 Metrics Community
s2pro-egy / scripts /patches_notes.md
ehabnegm's picture
ehabnegm
Phase 1: merged fast-AR LoRA step-1200 + full toolkit + phase-2 handoff
5c2beba verified
preview code
|
Raw
History Blame Contribute Delete
973 Bytes

Required patches (fish-speech main e5e2926 / sglang-omni main, 2026-07-18)

fish-speech (training)

  1. fish_speech/models/text2semantic/llama.py: both checkpoint(...) calls -> use_reentrant=False
  2. fish_speech/models/text2semantic/lit_module.py: add strict_loading = False to TextToSemantic
  3. (optional) lora_slow_patch.py adds slow_* target names that do NOT imply fast_*

sglang-omni (serving on Blackwell/sm_120)

  1. rm -r site-packages/deep_gemm (import-time assert without CUDA toolkit)
  2. apt: cuda-nvcc-13-0 libcublas-dev-13-0 libcusparse-dev-13-0 libcusolver-dev-13-0 libcurand-dev-13-0 ninja-build (torch cu130)
  3. sglang_omni/models/fishaudio_s2_pro/engine_builder.py: attention_backend "fa3" -> "triton"; disable_cuda_graph True; mem_fraction_static 0.55
  4. run patch_audio_decoder.py (adds FISH_FORCE_SDPA env gate replacing sgl_kernel flash_attn_with_kvcache) Launch env: CUDA_HOME=/usr/local/cuda-13.0 SGLANG_ENABLE_JIT_DEEPGEMM=0 FISH_FORCE_SDPA=1