wangbo/ComfyUI
1
0
Fork 0
mirror of https://github.com/comfyanonymous/ComfyUI.git synced 2026-04-24 09:22:32 +08:00
Code Issues Actions 31 Packages Projects Releases Wiki Activity

llama: use a more efficient rope implementation (#12434)

Browse Source 
Get rid of the cat and unary negation and inplace add-cmul the two
halves of the rope. Precompute -sin once at the start of the model
rather than every transformer block.

This is slightly faster on both GPU and CPU bound setups.
This commit is contained in:
rattus 2026-02-12 16:56:42 -08:00 committed by GitHub
parent 117e214354
commit ae79e33345
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 14 additions and 11 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

25
comfy/text_encoders/llama.py
View File

@ -355,13 +355,6 @@ class RMSNorm(nn.Module):
def rotate_half(x):
"""Rotates half the hidden dims of the input."""
x1 = x[..., : x.shape[-1] // 2]
x2 = x[..., x.shape[-1] // 2 :]
return torch.cat((-x2, x1), dim=-1)
def precompute_freqs_cis(head_dim, position_ids, theta, rope_scale=None, rope_dims=None, device=None): def precompute_freqs_cis(head_dim, position_ids, theta, rope_scale=None, rope_dims=None, device=None):
if not isinstance(theta, list): if not isinstance(theta, list):
theta = [theta] theta = [theta]
@ -390,20 +383,30 @@ def precompute_freqs_cis(head_dim, position_ids, theta, rope_scale=None, rope_di
else: else:
cos = cos.unsqueeze(1) cos = cos.unsqueeze(1)
sin = sin.unsqueeze(1) sin = sin.unsqueeze(1)
out.append((cos, sin)) sin_split = sin.shape[-1] // 2
out.append((cos, sin[..., : sin_split], -sin[..., sin_split :]))
if len(out) == 1: if len(out) == 1:
return out[0] return out[0]
return out return out
def apply_rope(xq, xk, freqs_cis): def apply_rope(xq, xk, freqs_cis):
org_dtype = xq.dtype org_dtype = xq.dtype
cos = freqs_cis[0] cos = freqs_cis[0]
sin = freqs_cis[1] sin = freqs_cis[1]
q_embed = (xq * cos) + (rotate_half(xq) * sin) nsin = freqs_cis[2]
k_embed = (xk * cos) + (rotate_half(xk) * sin)
q_embed = (xq * cos)
q_split = q_embed.shape[-1] // 2
q_embed[..., : q_split].addcmul_(xq[..., q_split :], nsin)
q_embed[..., q_split :].addcmul_(xq[..., : q_split], sin)
k_embed = (xk * cos)
k_split = k_embed.shape[-1] // 2
k_embed[..., : k_split].addcmul_(xk[..., k_split :], nsin)
k_embed[..., k_split :].addcmul_(xk[..., : k_split], sin)
return q_embed.to(org_dtype), k_embed.to(org_dtype) return q_embed.to(org_dtype), k_embed.to(org_dtype)