update to best iteration on unet weights

nod-ai · Jul 25, 2024 · c910c1a · c910c1a
1 parent 3fd954b
commit c910c1a
Showing 1 changed file with 3 additions and 3 deletions.
diff --git a/models/turbine_models/custom_models/sdxl_inference/unet.py b/models/turbine_models/custom_models/sdxl_inference/unet.py
@@ -90,9 +90,9 @@ def get_punet_model(hf_model_name, external_weight_path, quant_paths, precision=
     from sharktank.utils import cli
 
     if precision == "i8":
-        repo_id = "amd-shark/sdxl-quant-models"
-        subfolder = "unet/int8"
-        revision = "42e94070478ed0599c0225a4879b69b253206eb6"
+        repo_id = "amd-shark/sdxl-quant-int8"
+        subfolder = "mi300_all_sym_8_step14_fp32"
+        revision = "2e416a4205c519f5e62ba707ddf4f5022b6276c8"
     elif precision in ["fp16", "fp32"]:
         repo_id = hf_model_name
         subfolder = "unet"