update latency calcuation in allgather

cli99 · Nov 13, 2024 · c3bb7eb · c3bb7eb
1 parent 168d9a3
commit c3bb7eb
Showing 1 changed file with 10 additions and 7 deletions.
diff --git a/llm_analysis/analysis.py b/llm_analysis/analysis.py
@@ -1230,6 +1230,7 @@ def get_latency_fwd_per_tp_comm(self, batch_size: int, seq_len: int,
     def get_latency_fwd_per_layer_shared_dp_comm(self) -> float:
         dp_size = self.parallelism_config.dp_size
         ep_size = self.parallelism_config.ep_size
+        tp_size = self.parallelism_config.tp_size
 
         def time_allgather(S, n, B):
             # https://github.com/NVIDIA/nccl-tests/blob/master/doc/PERFORMANCE.md#allgather
@@ -1243,15 +1244,17 @@ def time_allgather(S, n, B):
             self.get_num_params_per_layer_layernorm()
         ) * self.dtype_config.weight_bits / BITS_PER_BYTE
 
-        latency_allgather_params_mlp = time_allgather(
-            params_bytes_mlp, dp_size / ep_size,
-            (self.get_intra_node_bandwidth()
-             if dp_size <= 8 else self.get_inter_node_bandwidth()) * 10**9)
+        # assuming tp and dp are preferred when sharding intra node, pp is only applied across nodes
+        # when (dp_size * tp_size) <= 8, the data parallel processes are within a node
+        bandwidth = self.get_intra_node_bandwidth() if (
+            dp_size * tp_size) <= 8 else self.get_inter_node_bandwidth()
+
+        latency_allgather_params_mlp = time_allgather(params_bytes_mlp,
+                                                      dp_size / ep_size,
+                                                      bandwidth * 10**9)
 
         latency_allgather_params_non_mlp = time_allgather(
-            params_bytes_non_mlp, dp_size,
-            (self.get_intra_node_bandwidth()
-             if dp_size <= 8 else self.get_inter_node_bandwidth()) * 10**9)
+            params_bytes_non_mlp, dp_size, bandwidth * 10**9)
 
         latency_fwd_per_layer_shared_dp_comm = latency_allgather_params_mlp + latency_allgather_params_non_mlp