Skip to content

Keep tied embeddings in fp32 #10

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

Closed
LJX2017 wants to merge 1 commit into from
Closed

Keep tied embeddings in fp32 #10

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
12 changes: 8 additions & 4 deletions train_gpt.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -514,10 +514,14 @@ def forward(self, x: Tensor) -> Tensor:


def restore_low_dim_params_to_fp32(module: nn.Module) -> None:
# Keep small/control parameters in fp32 even when the model body runs in bf16.
# Keep the high-leverage tied embedding plus small/control parameters in fp32.
with torch.no_grad():
for name, param in module.named_parameters():
if (param.ndim < 2 or any(pattern in name for pattern in CONTROL_TENSOR_NAME_PATTERNS)) and param.dtype != torch.float32:
if (
name == "tok_emb.weight"
or param.ndim < 2
or any(pattern in name for pattern in CONTROL_TENSOR_NAME_PATTERNS)
) and param.dtype != torch.float32:
param.data = param.data.float()


Expand Down Expand Up @@ -698,7 +702,7 @@ def _init_weights(self) -> None:
nn.init.zeros_(module.weight)

def forward(self, input_ids: Tensor, target_ids: Tensor) -> Tensor:
x = self.tok_emb(input_ids)
x = F.embedding(input_ids, self.tok_emb.weight).to(dtype=torch.bfloat16)
x = F.rms_norm(x, (x.size(-1),))
x0 = x
skips: list[Tensor] = []
Expand All @@ -715,7 +719,7 @@ def forward(self, input_ids: Tensor, target_ids: Tensor) -> Tensor:
x = self.final_norm(x).reshape(-1, x.size(-1))
targets = target_ids.reshape(-1)
if self.tie_embeddings:
logits_proj = F.linear(x, self.tok_emb.weight)
logits_proj = F.linear(x, self.tok_emb.weight.to(dtype=x.dtype))
else:
if self.lm_head is None:
raise RuntimeError("lm_head is required when tie_embeddings=False")
Expand Down
2 changes: 1 addition & 1 deletion train_gpt_mlx.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -405,7 +405,7 @@ def __init__(self, vocab_size: int, num_layers: int, dim: int, num_heads: int, n
b.mlp.proj.weight = mx.zeros_like(b.mlp.proj.weight)
self.tok_emb.weight = (
mx.random.normal(self.tok_emb.weight.shape, dtype=mx.float32) * tied_embed_init_std
).astype(COMPUTE_DTYPE)
)

def softcap(self, logits: mx.array) -> mx.array:
c = self.logit_softcap
Expand Down