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

Update llama.py

Browse source
This commit is contained in:
Daniel Han-Chen 2024-02-01 23:40:36 +11:00
commit e0ea238256
1 changed files with 8 additions and 4 deletions
Download patch file Download diff file Expand all files Collapse all files

12
unsloth/models/llama.py
View file

@ -261,10 +261,10 @@ def LlamaAttention_fast_forward_inference(
# New KV cache
# Kn = torch.cat([K1, Kn], dim = 2)
# Vn = torch.cat([V1, Vn], dim = 2)
if not hasattr(self, "paged_attention_K"):
paged_attention = torch.empty((2, bsz, n_kv_heads, 2048, head_dim), dtype = dtype, device = "cuda")
self.paged_attention_K = paged_attention[0]
self.paged_attention_V = paged_attention[1]
if not hasattr(self, "paged_attention"):
self.paged_attention = torch.empty((2, bsz, n_kv_heads, 2048, head_dim), dtype = dtype, device = "cuda")
self.paged_attention_K = self.paged_attention[0]
self.paged_attention_V = self.paged_attention[1]
self.paged_attention_K[:,:,:kv_seq_len,:] = K1
self.paged_attention_V[:,:,:kv_seq_len,:] = V1
pass
@ -346,6 +346,10 @@ def LlamaAttention_fast_forward(
position_ids,
)
return A, None, past_key_value
elif hasattr(self, "paged_attention_K"):
del self.paged_attention_K
del self.paged_attention_V
del self.paged_attention
pass
bsz, q_len, _ = hidden_states.size()