lshqqytiger
error at generate and terminated
see this message.
Loading Model: {'checkpoint_info': {'filename': 'F:\AI\stable-diffusion-webui-amdgpu-forge\models\Stable-diffusion\hanimix_flux_v16B_nsfw-nf4.safetensors', 'hash': 'e9f34423'}, 'additional_modules': ['F:\AI\stable-diffusion-webui-amdgpu-forge\models\VAE\ae.safetensors', 'F:\AI\stable-diffusion-webui-amdgpu-forge\models\text_encoder\clip_l.safetensors', 'F:\AI\stable-diffusion-webui-amdgpu-forge\models\text_encoder\t5xxl_fp8_e4m3fn.safetensors'], 'unet_storage_dtype': None} [Unload] Trying to free all memory for cuda:0 with 0 models keep loaded ... Done. StateDict Keys: {'transformer': 1722, 'vae': 244, 'text_encoder': 196, 'text_encoder_2': 220, 'ignore': 0} Using Detected T5 Data Type: torch.float8_e4m3fn Using Detected UNet Type: nf4 Using pre-quant state dict! Working with z of shape (1, 16, 32, 32) = 16384 dimensions. K-Model Created: {'storage_dtype': 'nf4', 'computation_dtype': torch.bfloat16} Model loaded in 0.6s (unload existing model: 0.2s, forge model load: 0.4s). Skipping unconditional conditioning when CFG = 1. Negative Prompts are ignored. [Unload] Trying to free 7723.54 MB for cuda:0 with 0 models keep loaded ... Done. [Memory Management] Target: JointTextEncoder, Free GPU: 21428.60 MB, Model Require: 5153.49 MB, Previously Loaded: 0.00 MB, Inference Require: 1024.00 MB, Remaining: 15251.11 MB, All loaded to GPU. Moving model(s) has taken 2.13 seconds Distilled CFG Scale: 3.5 [Unload] Trying to free 9411.13 MB for cuda:0 with 0 models keep loaded ... Current free memory is 16145.95 MB ... Done. [Memory Management] Target: KModel, Free GPU: 16145.95 MB, Model Require: 6246.84 MB, Previously Loaded: 0.00 MB, Inference Require: 1024.00 MB, Remaining: 8875.11 MB, All loaded to GPU. Moving model(s) has taken 3.00 seconds 0%| | 0/20 [00:00<?, ?it/s]Error named symbol not found at line 90 in file D:\a\bitsandbytes\bitsandbytes\csrc\ops.cu
It work well at "flux1-dev-fp8.safetensors" but not work with "hanimix_flux_v16B_nsfw-nf4.safetensors" custom checkpoint. Problem with "nf4" reading?
NF4 requires BitsandBytes which is currently in development for AMD on Windows but not ready for Zluda. The best you can go for are GGUF models. You can find different sizes like Q4 or Q8.
