ceos/unsloth
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Update gemma.py

Browse source
This commit is contained in:
Daniel Han-Chen 2024-02-26 04:37:12 +11:00
commit 0b693df646
1 changed files with 1 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

2
unsloth/models/gemma.py
View file

@ -92,7 +92,7 @@ class FastGemmaRotaryEmbedding(torch.nn.Module):
position_ids = torch.arange(self.max_position_embeddings, device=x.device, dtype=torch.int64).unsqueeze(0)
inv_freq_expanded = self.inv_freq[None, :, None].float().expand(1, -1, 1)
position_ids_expanded = position_ids[:, None, :].float()
with torch.autocast(enabled = False):
with torch.autocast(device_type="cuda", enabled = False):
freqs = (inv_freq_expanded @ position_ids_expanded).transpose(1, 2)
emb = torch.cat((freqs, freqs), dim=-1)
self.cos_cached = emb.cos().to(dtype=x.dtype)