diff --git a/nemo/collections/common/metrics/perf_metrics.py b/nemo/collections/common/metrics/perf_metrics.py index 285bf7b4f9d6..57b62bce4675 100644 --- a/nemo/collections/common/metrics/perf_metrics.py +++ b/nemo/collections/common/metrics/perf_metrics.py @@ -254,10 +254,10 @@ def _grok(self): vocab_size = LLM_VOCAB_SIZE_MAP["grok"] return ( self.layers - * self.gbs + * self.gbs * self.enc_seq_len - * self.hs - * self.hs + * self.hs + * self.hs * ( 12 + (12 * self.query_groups / self.attention_heads) @@ -265,4 +265,4 @@ def _grok(self): + (12 * self.enc_seq_len / self.hs) + (6 * vocab_size / (self.layers * self.hs)) ) - ) \ No newline at end of file + )