No official torrent yet — requesting it tells the flock what to host next. You cannot seed a catalog-only page.
Cataloged from lightx2v/Z-Image-Turbo-Quantized
All models by lightx2v →Quantized weights for Z-Image-Turbo optimized for 8GB VRAM GPUs.
z_image_turbo_scaled_fp8_e4m3fn.safetensors (6.17 GB) - FP8 E4M3FN quantized weightsz_image_turbo_int8.safetensors (6.17 GB) - INT8 quantized weightsTo run Z-Image-Turbo on 8GB VRAM GPUs, you need to:
FP8 (Recommended) - Better quality and speed:
INT8 - Alternative option:
"model" (Recommended for 8GB): Offload entire model to CPU, load to GPU only during inference. Maximum memory savings."block": Offload individual transformer blocks. More fine-grained control.Order matters: All enable_quantize() and enable_offload() calls must be made before create_generator(), otherwise they will not take effect.
Text encoder quantization: Using int4 Qwen3 text encoder is highly recommended for 8GB VRAM GPUs to ensure stable operation.
Memory optimization: The combination of FP8/INT8 transformer + int4 Qwen3 + model-level offloading is optimized for 8GB VRAM.
If you find this project helpful, please give us a ⭐ on GitHub
git clone https://github.com/ModelTC/LightX2V.gitcd LightX2Vpip install .from lightx2v import LightX2VPipeline
# Initialize pipelinepipe = LightX2VPipeline( model_path="Tongyi-MAI/Z-Image-Turbo", model_cls="z_image", task="t2i",)
# Step 1: Enable quantization (FP8 transformer + INT4 text encoder)pipe.enable_quantize( dit_quantized=True, dit_quantized_ckpt="lightx2v/Z-Image-Turbo-Quantized/z_image_turbo_scaled_fp8_e4m3fn.safetensors", quant_scheme="fp8-sgl", # IMPORTANT: Use int4 Qwen3 for 8GB VRAM text_encoder_quantized=True, text_encoder_quantized_ckpt="JunHowie/Qwen3-4B-GPTQ-Int4", text_encoder_quant_scheme="int4")
# Step 2: Enable CPU offloadingpipe.enable_offload( cpu_offload=True, offload_granularity="model", # Use "model" for maximum memory savings)
# Step 3: Create generatorpipe.create_generator( attn_mode="flash_attn3", aspect_ratio="16:9", infer_steps=9, guidance_scale=1,)
# Step 4: Generate imagepipe.generate( seed=42, prompt="A beautiful landscape with mountains and lakes, ultra HD, 4K", negative_prompt="", save_result_path="output.png",)dit_quantized_ckpt="lightx2v/Z-Image-Turbo-Quantized/z_image_turbo_scaled_fp8_e4m3fn.safetensors",quant_scheme="fp8-sgl",dit_quantized_ckpt="lightx2v/Z-Image-Turbo-Quantized/z_image_turbo_int8.safetensors",quant_scheme="int8-sgl",