Commit History
-
sc-22760: drop the false quantization marker from q4/q8 text_encoder/config.json
Verified signatureThe q4/q8 tiers ship the Qwen3 text encoder dense (399 BF16 tensors, no scales/biases); only config.json claimed Q4/Q8. The encoder contract refuses the config/surface mismatch, so no native load of these tiers was possible. Weights unchanged.bbf22de
-
Add missing scheduler/scheduler_config.json to q4/q8/bf16 tiers (fixes false-incomplete gate, GH SceneWorks/SceneWorks#1858)
Verified signature406265b
-
sc-8711: bf16/model_index.json
Verified signaturefc6579b -
sc-8711: bf16/tokenizer
Verified signature4ae5a76 -
sc-8711: bf16/vae
Verified signature1f02bc4 -
sc-8711: bf16/text_encoder
Verified signatureb360419 -
sc-8711: bf16/transformer
Verified signature480ed87 -
sc-8711: q8
Verified signaturea733c21 -
sc-8711: q4
Verified signature20cff58 -
Upload LICENSE.md with huggingface_hub
Verified signature6f9f22e -
Upload README.md with huggingface_hub
Verified signature72ab7ec -
initial commit
Verified signaturef02bf60