Skip to content

xTimeCrystal disk offloading #310

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 3 commits into
base: main
Choose a base branch
from
Open

xTimeCrystal disk offloading #310

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
619a5d9
Update rwkv_v8_rc00_demo.py
xTimeCrystal May 29, 2025
0eb6b12
Update rwkv_v8_rc00_demo.py
xTimeCrystal May 29, 2025
c6df56e
Update rwkv_v8_rc00_demo.py
xTimeCrystal May 29, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
11 changes: 10 additions & 1 deletion RWKV-v7/rwkv_v8_rc00_demo.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -106,6 +106,7 @@ def __init__(self, args):
z['blocks.0.att.v0'] = z['blocks.0.att.a0'] # actually ignored
z['blocks.0.att.v1'] = z['blocks.0.att.a1'] # actually ignored
z['blocks.0.att.v2'] = z['blocks.0.att.a2'] # actually ignored
self.deepembs = {}

def forward(self, idx, state, full_output=False):
if state == None:
Expand Down Expand Up @@ -178,7 +179,15 @@ def forward_seq(self, idx:List[int], state:List[torch.Tensor], full_output:bool=

xx = F.layer_norm(x, (self.n_embd,), weight=z[bbb+'ln2.weight'], bias=z[bbb+'ln2.bias'])

xx, state[i*3+2] = RWKV_x080_CMix_seq(xx, state[i*3+2], z[ffn+'x_k'], z[ffn+'key.weight'], z[ffn+'value.weight'], z[ffn+'enn.weight'][idx])
# Disk space offloading
if not (ffn+'enn.weight' in self.deepembs):
z[ffn+'enn.weight'].numpy().tofile(ffn+'enn.weight_storage'+'.pt')
emb_size = z[ffn+'enn.weight'].numel()
emb_dtype = z[ffn+'enn.weight'].dtype
z[ffn+'enn.weight'] = torch.from_file(ffn+'enn.weight_storage'+'.pt', size=emb_size, dtype=emb_dtype, device='cpu')
self.deepembs[ffn+'enn.weight'] = True

xx, state[i*3+2] = RWKV_x080_CMix_seq(xx, state[i*3+2], z[ffn+'x_k'], z[ffn+'key.weight'], z[ffn+'value.weight'], z[ffn+'enn.weight'][idx].cuda())
x = x + xx

if not full_output: x = x[-1,:]
Expand Down