Cleanup debug log for Ring model (#18793)

python/sglang/srt/configs/mamba_utils.py

@@ -102,7 +102,7 @@ def mamba2_state_dtype(config=None) -> Mamba2StateDType:
         else:
             ssm_dtype = dtype_map[env_ssm_dtype]
 
-    logger.info(f"Mamba2 state dtype: conv_dtype={conv_dtype}, ssm_dtype={ssm_dtype}")
+    logger.debug(f"Mamba2 state dtype: conv_dtype={conv_dtype}, ssm_dtype={ssm_dtype}")
 
     return Mamba2StateDType(conv=conv_dtype, temporal=ssm_dtype)
 

python/sglang/srt/models/bailing_moe_linear.py

@@ -77,6 +77,7 @@ from sglang.srt.utils import (
     is_sm100_supported,
     make_layers,
 )
+from sglang.srt.utils.common import rank0_log
 
 _is_hip = is_hip()
 _is_cuda = is_cuda()
@@ -423,7 +424,7 @@ class BailingMoELinearAttention(nn.Module):
         # minimax / seg_la / fla
         # TODO support fla
         self.linear_backend = getattr(config, "linear_backend", "seg_la")
-        logger.info(f"linear_backend in bailing_moe_linear: {self.linear_backend}")
+        logger.debug(f"linear_backend in bailing_moe_linear: {self.linear_backend}")
         self.linear_scale = True if self.linear_backend == "minimax" else False
         self.linear_rope = getattr(config, "linear_rope", True)
         if hasattr(config, "use_linear_silu"):
@@ -740,7 +741,7 @@ class BailingMoELinearDecoderLayer(nn.Module):
                     alt_stream=alt_stream,
                 )
             else:
-                logger.info(f"==={layer_id=} use gqa")
+                logger.debug(f"layer {layer_id} use gqa")
                 self.attention = BailingMoEAttention(
                     config,
                     quant_config=quant_config,
@@ -895,8 +896,10 @@ class BailingMoELinearModel(nn.Module):
             0 if is_linear_layer(i, self.layer_group_size) else 1
             for i in range(self.num_layers)
         ]
-        logger.info(
+        num_linear = sum(1 for t in self.decoder_attention_types if t == 0)
-            f"attention type of layers:{self.decoder_attention_types}, 0 is linear layer and 1 is softmax layer!"
+        num_full = sum(1 for t in self.decoder_attention_types if t == 1)
+        rank0_log(
+            f"Layer config: {num_linear} linear attention layers, {num_full} full attention layers"
         )
 
         assert (
@@ -931,11 +934,6 @@ class BailingMoELinearModel(nn.Module):
             prefix=f"{prefix}.layers",
         )
 
-        linear_layer_nums = sum(
-            1 for i in range(self.num_layers) if self.decoder_attention_types[i] == 0
-        )
-        logger.info(f"linear_layer_nums={linear_layer_nums}")
-
         norm_kwargs = {}
         if hasattr(config, "rms_norm_eps"):
             norm_kwargs["eps"] = config.rms_norm_eps
@@ -1082,7 +1080,7 @@ class BailingMoELinearForCausalLM(nn.Module):
                             and layer_id >= self.model.start_layer
                         ):
                             layer_ids.add(layer_id)
-        logger.info(f"=====layer_ids {layer_ids}")
+        logger.debug(f"weight loading layer_ids: {layer_ids}")
 
         for layer_id in layer_ids:
             self_attn = (