From 40c244a7bc13202a56aff2ffbf3a429170d24880 Mon Sep 17 00:00:00 2001 From: Daniel Han-Chen Date: Sun, 25 Feb 2024 03:53:40 +1100 Subject: [PATCH] Update gemma.py --- unsloth/models/gemma.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/unsloth/models/gemma.py b/unsloth/models/gemma.py index 0072c8c47b..6e8b1c1a37 100644 --- a/unsloth/models/gemma.py +++ b/unsloth/models/gemma.py @@ -105,7 +105,7 @@ class FastGemmaRotaryEmbedding(torch.nn.Module): ) t = torch.arange(self.max_position_embeddings, device="cpu", dtype=torch.int64).float().to("cuda").unsqueeze(0) - inv_freq_expanded = self.inv_freq[None, :, None].float().expand(position_ids.shape[0], -1, 1).to("cuda") + inv_freq_expanded = self.inv_freq[None, :, None].float().expand(1, -1, 1).to("cuda") position_ids_expanded = t[:, None, :].float() freqs = (inv_freq_expanded @ position_ids_expanded).transpose(1, 2) print(freqs.to(torch.int64))