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

Repatch

Browse source
This commit is contained in:
Daniel Han-Chen 2024-01-26 03:15:23 +11:00
commit 46ec8bbc3d
2 changed files with 10 additions and 10 deletions
Download patch file Download diff file Expand all files Collapse all files

16
unsloth/kernels/fast_lora.py
View file

@ -128,20 +128,20 @@ class LoRA_MLP(torch.autograd.Function):
h, DW_f, DW_dfg = DW, e, g
# Down projection LoRA weights
d_downA = h.t() @ (dY @ downB.t())
d_downB = (downA.t() @ h.t()) @ dY
d_downA = (h.t() @ dY) @ downB.t()
d_downB = downA.t() @ (h.t() @ dY)
d_downA *= downS
d_downB *= downS
# Up projection LoRA weights
d_upA = X.t() @ (DW_f @ upB.t())
d_upB = (upA.t() @ X.t()) @ DW_f
d_upA = (X.t() @ DW_f) @ upB.t()
d_upB = upA.t() @ (X.t() @ DW_f)
d_upA *= upS
d_upB *= upS
# Gate projection LoRA weights
d_gateA = X.t() @ (DW_dfg @ gateB.t())
d_gateB = (gateA.t() @ X.t() @ DW_dfg)
d_gateA = (X.t() @ DW_dfg) @ gateB.t()
d_gateB = gateA.t() @ (X.t() @ DW_dfg)
d_gateA *= gateS
d_gateB *= gateS
@ -152,13 +152,13 @@ class LoRA_MLP(torch.autograd.Function):
# (D @ W.T * f) @ (U.T + B.T @ A.T)
dX = torch.matmul(DW_f, upW.t(), out = X)
del upW
dX += DW_f @ upB.to(dtype).t() @ (upS * upA.to(dtype).t())
dX += DW_f @ upB.t() @ (upS * upA.t())
# And add the derivative for the gate projection
gateW = fast_dequantize(gateW.t(), gateW_quant)
dX += DW_dfg @ gateW.t()
del gateW
dX += DW_dfg @ gateB.to(dtype).t() @ (gateS * gateA.to(dtype).t())
dX += DW_dfg @ gateB.t() @ (gateS * gateA.t())
# gateW, gateW_quant, gateA, gateB, gateS,
# upW, upW_quant, upA, upB, upS,

4
unsloth/kernels/swiglu.py
View file

@ -24,11 +24,11 @@ def _fg_kernel(e, g, h, n_elements, BLOCK_SIZE : tl.constexpr,):
offsets = block_idx*BLOCK_SIZE + tl.arange(0, BLOCK_SIZE)
mask = offsets < n_elements
e_row = tl.load(e + offsets, mask = mask, other = 0)#.to(tl.float32)
e_row = tl.load(e + offsets, mask = mask, other = 0).to(tl.float32)
g_row = tl.load(g + offsets, mask = mask, other = 0)#.to(tl.float32)
# f = e * sigmoid(e)
f_row = e_row / (1 + tl.exp(-e_row.to(tl.float32)).to(g_row.dtype))
f_row = e_row / (1 + tl.exp(-e_row))
f_row = f_row.to(g_row.dtype) # Exact copy from HF
# h = f * g
h_row = f_row * g_row