fix: use F.silu in test script for old corex torch
This commit is contained in:
@@ -66,7 +66,7 @@ for k in range(top_k):
|
|||||||
eid = expert_ids[k].item()
|
eid = expert_ids[k].item()
|
||||||
w = expert_weights[k].item()
|
w = expert_weights[k].item()
|
||||||
gate_up = F.linear(hidden, w13[eid])
|
gate_up = F.linear(hidden, w13[eid])
|
||||||
gate = torch.silu(gate_up[:, :I])
|
gate = F.silu(gate_up[:, :I])
|
||||||
up = gate_up[:, I:]
|
up = gate_up[:, I:]
|
||||||
act = gate * up
|
act = gate * up
|
||||||
expert_out = F.linear(act, w2[eid])
|
expert_out = F.linear(act, w2[eid])
|
||||||
@@ -131,7 +131,7 @@ for _ in range(3):
|
|||||||
eid = expert_ids[k].item()
|
eid = expert_ids[k].item()
|
||||||
w = expert_weights[k].item()
|
w = expert_weights[k].item()
|
||||||
gate_up = F.linear(hidden, w13[eid])
|
gate_up = F.linear(hidden, w13[eid])
|
||||||
gate = torch.silu(gate_up[:, :I])
|
gate = F.silu(gate_up[:, :I])
|
||||||
up = gate_up[:, I:]
|
up = gate_up[:, I:]
|
||||||
act = gate * up
|
act = gate * up
|
||||||
out_py += w * F.linear(act, w2[eid])
|
out_py += w * F.linear(act, w2[eid])
|
||||||
@@ -144,7 +144,7 @@ for _ in range(100):
|
|||||||
eid = expert_ids[k].item()
|
eid = expert_ids[k].item()
|
||||||
w = expert_weights[k].item()
|
w = expert_weights[k].item()
|
||||||
gate_up = F.linear(hidden, w13[eid])
|
gate_up = F.linear(hidden, w13[eid])
|
||||||
gate = torch.silu(gate_up[:, :I])
|
gate = F.silu(gate_up[:, :I])
|
||||||
up = gate_up[:, I:]
|
up = gate_up[:, I:]
|
||||||
act = gate * up
|
act = gate * up
|
||||||
out_py += w * F.linear(act, w2[eid])
|
out_py += w * F.linear(act, w2[eid])
|
||||||
|
|||||||
Reference in New Issue
Block a user