japanese-learning-avatar / requirements.txt
WolfDavid's picture
chore(02-01): arm LFS for .spm and pin the language stack
8e5161f
Raw History Blame
1.19 kB
gradio==6.22.0
# NOTE (plan 01-05, first deploy): `spaces` is deliberately NOT pinned here, and must
# not be re-added. The Hugging Face Space builder appends its own pin to the install
# command it generates:
#
# pip install -r /tmp/requirements.txt "torch<=2.11.0" gradio[oauth,mcp]==6.22.0 \
# "uvicorn>=0.14.0" "websockets>=10.4" spaces==0.51.3
#
# so pinning spaces==0.51.1 here made pip see two exact pins for one package and the
# build died with `ResolutionImpossible: Cannot install spaces==0.51.1 and
# spaces==0.51.3` (BUILD_ERROR on commit ad85cc8). The platform owns this version on a
# ZeroGPU Space; whatever it injects is what @spaces.GPU will run against. The local
# dev pin still lives in pyproject.toml, where it is ours to choose.
huggingface_hub==1.28.0
# --- Phase 2 language core (plan 02-01): tokenizer, dictionary, CPU translation runtime. No transformers, no torch - the MT model is pre-converted to CTranslate2 int8 in data/mt/.
SudachiPy==0.6.11
SudachiDict-core==20260723
jaconv==0.5.0
ctranslate2==4.8.2
sentencepiece==0.2.2
https://github.com/VOICEVOX/voicevox_core/releases/download/0.17.0/voicevox_core-0.17.0-cp310-abi3-manylinux_2_34_x86_64.whl