Update README.md

README.md

@@ -33,8 +33,19 @@ import requests
 import torch
 import diffusers
 from sdnq import SDNQConfig # import sdnq to register it into diffusers and transformers
+from sdnq.common import use_torch_compile as triton_is_available
+from sdnq.loader import apply_sdnq_options_to_model
 
 pipe = diffusers.Ideogram4Pipeline.from_pretrained("Disty0/Ideogram-4-SDNQ-4bit-dynamic-hadamard", torch_dtype=torch.bfloat16)
+
+# Enable INT8 MatMul for AMD, Intel ARC and Nvidia GPUs:
+if triton_is_available and (torch.cuda.is_available() or torch.xpu.is_available()):
+    pipe.transformer = apply_sdnq_options_to_model(pipe.transformer, use_quantized_matmul=True)
+    pipe.unconditional_transformer = apply_sdnq_options_to_model(pipe.unconditional_transformer, use_quantized_matmul=True)
+    pipe.text_encoder = apply_sdnq_options_to_model(pipe.text_encoder, use_quantized_matmul=True)
+    # pipe.transformer = torch.compile(pipe.transformer) # optional for faster speeds
+    # pipe.unconditional_transformer = torch.compile(pipe.unconditional_transformer) # optional for faster speeds
+
 pipe.enable_model_cpu_offload()
 
 # Expand the prompt into a structured JSON caption with Ideogram's free hosted magic-prompt API.