Basic script for OLMoE inference.

Add: basic script for OLMoE inference and saving expert logits.

Add: OLMoE vLLM module with logits saving.

Add: vLLM model registering.

Fix: update DataLogger module for multiprocessing.

models/log_expert/olmoe.py

@@ -17,6 +17,7 @@
 This file is origin from vllm/model_executor/models/olmoe.py
 """
 from collections.abc import Iterable
+from datetime import datetime
 from functools import partial
 from typing import Any, Optional, Union
 
@@ -51,7 +52,7 @@ from vllm.model_executor.models.utils import (AutoWeightsLoader, is_pp_missing_p
                     make_empty_intermediate_tensors_factory, make_layers,
                     maybe_prefix)
 
-from utils import DataLogger
+from utils import DataLogger as dlog
 
 logger = init_logger(__name__)
 
@@ -100,10 +101,20 @@ class OlmoeMoE(nn.Module):
         hidden_states = hidden_states.view(-1, hidden_dim)
         # router_logits: (num_tokens, n_experts)
         router_logits, _ = self.gate(hidden_states)
+
+        dlog.log({
+            "_time": datetime.now(),
+            "router_logits": router_logits.cpu().float().numpy(),
+            "layer": self.layer_idx,
+        })
+
         final_hidden_states = self.experts(hidden_states=hidden_states,
                                            router_logits=router_logits)
         return final_hidden_states.view(orig_shape)
 
+    def add_logging_metrics(self, layer_idx: int):
+        self.layer_idx = layer_idx
+
 
 class OlmoeAttention(nn.Module):
 
@@ -234,6 +245,16 @@ class OlmoeDecoderLayer(nn.Module):
             prefix=f"{prefix}.self_attn",
         )
 
+        # Extract layer_idx from prefix
+        self.layer_idx = None
+        try:
+            # Prefix format: "model.layers.7"
+            parts = prefix.split('.')
+            if len(parts) >= 3 and parts[2].isdigit():
+                self.layer_idx = int(parts[2])
+        except ValueError:
+            pass
+
         self.mlp = OlmoeMoE(
             num_experts=config.num_experts,
             top_k=config.num_experts_per_tok,
@@ -245,6 +266,8 @@ class OlmoeDecoderLayer(nn.Module):
         self.input_layernorm = RMSNorm(config.hidden_size, eps=1e-5)
         self.post_attention_layernorm = RMSNorm(config.hidden_size, eps=1e-5)
 
+        self.mlp.add_logging_metrics(layer_idx=self.layer_idx)
+
     def forward(
         self,
         positions: torch.Tensor,