chore: fix the load quantization model

2025-09-01 12:49:50 +08:00 · 2023-12-29 23:01:20 +11:00
parent 31ddbd7806
commit a7da4ad7a3
1 changed files with 1 additions and 1 deletions
--- a/llms/deepseek-coder/deepseek_coder.py
+++ b/llms/deepseek-coder/deepseek_coder.py
@@ -244,7 +244,7 @@ def load_model(model_path: str):

    model = DeepseekCoder(model_args)
    weights = mx.load(str(model_path / "weights.npz"))
-    if quantization := config.get("quantization", False):
+    if quantization:
        nn.QuantizedLinear.quantize_module(model, **quantization)
    model.update(tree_unflatten(list(weights.items())))