sozai / requirements.txt
Shereen Lee
fix(img2img): add peft (load_lora_weights needs it) + isolate LoRA load so a failure falls back to base; UI: How-This-Works guide + upload label, drop corner video
8314a51
Raw
History Blame Contribute Delete
1.09 kB
gradio
websockets>=12.0
transformers
# Flux2KleinPipeline (img2img watercolour develop) is not in a released diffusers
# yet — it must come from git main. See backend/hf/3_infer_img2img.py.
diffusers @ git+https://github.com/huggingface/diffusers.git
accelerate
peft # required by diffusers load_lora_weights (the watercolour scene LoRA)
pillow-heif # HEIC/HEIF source photos for the img2img develop step
hf_xet
huggingface_hub
soundfile>=0.12
librosa>=0.10
langdetect>=1.0.9
arize-phoenix>=7.0 # the local Phoenix server + UI (run: phoenix serve)
arize-phoenix-otel>=0.16.0 # register() helper + OpenInference semconv re-exports
opentelemetry-sdk>=1.24
opentelemetry-exporter-otlp-proto-http>=1.24 # raw OTLP/HTTP fallback exporter
openinference-semantic-conventions>=0.1.9
#llama cpp
--extra-index-url https://abetlen.github.io/llama-cpp-python/whl/cu122
spaces
nvidia-cuda-runtime-cu12
nvidia-cublas-cu12
nvidia-cuda-nvrtc-cu12
llama-cpp-python
# nemotron
cython
packaging
nemo_toolkit[asr] @ git+https://github.com/NVIDIA/NeMo.git@main
#for local only
python-dotenv