format

Author: giterator

python/mlc_llm/conversation_template/llama.py

@@ -21,8 +21,8 @@
         role_content_sep="<|header_end|>\n\n",
         role_empty_sep="<|header_end|>\n\n",
         stop_str=[],
-        stop_token_ids=[200001, 200007, 200008],# "<|end_of_text|>", "<|eom|>", "<|eot|>"
-        system_prefix_token_ids=[200000],# "<|begin_of_text|>"
+        stop_token_ids=[200001, 200007, 200008],  # "<|end_of_text|>", "<|eom|>", "<|eot|>"
+        system_prefix_token_ids=[200000],  # "<|begin_of_text|>"
         add_role_after_system_message=False,
     )
 )

python/mlc_llm/model/llama4/llama4_loader.py

@@ -66,7 +66,9 @@ def huggingface(model_config: Llama4Config, quantization: Quantization) -> Exter
         mlc_param = named_parameters[mlc_name]
         mapping.add_mapping(
             mlc_name,
-            [hf_name,],
+            [
+                hf_name,
+            ],
             functools.partial(
                 lambda x, dtype: x.astype(dtype),
                 dtype=mlc_param.dtype,
@@ -80,7 +82,9 @@ def huggingface(model_config: Llama4Config, quantization: Quantization) -> Exter
         mlc_param = named_parameters[mlc_name]
         mapping.add_mapping(
             mlc_name,
-            [hf_name,],
+            [
+                hf_name,
+            ],
             functools.partial(
                 lambda x, dtype: x.astype(dtype),
                 dtype=mlc_param.dtype,
@@ -90,11 +94,13 @@ def huggingface(model_config: Llama4Config, quantization: Quantization) -> Exter
         mlp = f"model.layers.{i}.feed_forward"
         mlc_name = f"{mlp}.experts.down_proj"
         hf_name = f"language_model.{mlp}.experts.down_proj"
-        
+
         mlc_param = named_parameters[mlc_name]
         mapping.add_mapping(
             mlc_name,
-            [hf_name,],
+            [
+                hf_name,
+            ],
             functools.partial(
                 lambda x, dtype: x.astype(dtype),
                 dtype=mlc_param.dtype,
@@ -113,7 +119,8 @@ def huggingface(model_config: Llama4Config, quantization: Quantization) -> Exter
             )
     return mapping
 
-#TODO: This needs to be done for llama4
+
+# TODO: This needs to be done for llama4
 def awq(model_config: Llama4Config, quantization: Quantization) -> ExternMapping:
     """Returns a parameter mapping that maps from the names of MLC LLM parameters to
     the names of AWQ parameters.