HF 4.31

2025-06-05 21:59:24 +02:00 · 2023-07-22 15:25:14 +02:00
parent 7823da564e
commit fa9d17b3d3
4 changed files with 4 additions and 3 deletions
--- a/modeling/inference_models/generic_hf_torch/class.py
+++ b/modeling/inference_models/generic_hf_torch/class.py
@@ -81,6 +81,7 @@ class model_backend(HFTorchInferenceModel):

        tf_kwargs = {
            "low_cpu_mem_usage": True,
+            "pretraining_tp": 1,
        }
        
        if self.quantization == "8bit":