Skip to content

Replace RMSNorm by nn.RMSNorm #1464

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

Merged
merged 2 commits into from
Jan 24, 2025
Merged

Replace RMSNorm by nn.RMSNorm #1464

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
ba9aba1
Replace RMSNorm by nn.RMSNorm
manuelcandales Jan 16, 2025
b8801f5
pin bump
manuelcandales Jan 24, 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
6 changes: 3 additions & 3 deletions install/install_requirements.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -51,13 +51,13 @@ echo "Using pip executable: $PIP_EXECUTABLE"
# NOTE: If a newly-fetched version of the executorch repo changes the value of
# PYTORCH_NIGHTLY_VERSION, you should re-run this script to install the necessary
# package versions.
PYTORCH_NIGHTLY_VERSION=dev20250119
PYTORCH_NIGHTLY_VERSION=dev20250124

# Nightly version for torchvision
VISION_NIGHTLY_VERSION=dev20250119
VISION_NIGHTLY_VERSION=dev20250124

# Nightly version for torchtune
TUNE_NIGHTLY_VERSION=dev20250119
TUNE_NIGHTLY_VERSION=dev20250124

# The pip repository that hosts nightly torch packages. cpu by default.
# If cuda is available, based on presence of nvidia-smi, install the pytorch nightly
Expand Down
20 changes: 3 additions & 17 deletions torchchat/model.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -657,7 +657,7 @@ def __init__(self, config: TransformerArgs) -> None:
self.layers[str(layer_id)] = TransformerBlock(config)

if config.stage_idx == config.n_stages - 1:
self.norm = RMSNorm(config.dim, eps=config.norm_eps)
self.norm = nn.RMSNorm(config.dim, eps=config.norm_eps)
self.output = nn.Linear(config.dim, config.vocab_size, bias=False)
if config.tie_word_embeddings:
self.output.weight = self.tok_embeddings.weight
Expand Down Expand Up @@ -751,8 +751,8 @@ def __init__(self, config: TransformerArgs) -> None:
super().__init__()
self.attention = Attention(config)
self.feed_forward = FeedForward(config)
self.ffn_norm = RMSNorm(config.dim, config.norm_eps)
self.attention_norm = RMSNorm(config.dim, config.norm_eps)
self.ffn_norm = nn.RMSNorm(config.dim, config.norm_eps)
self.attention_norm = nn.RMSNorm(config.dim, config.norm_eps)
# None for llama architecture, set for granite architectures
self.residual_multiplier = (
config.residual_multiplier
Expand Down Expand Up @@ -928,20 +928,6 @@ def forward(self, x: Tensor) -> Tensor:
return self.w2(F.silu(self.w1(x)) * self.w3(x))


class RMSNorm(nn.Module):
def __init__(self, dim: int, eps: float = 1e-5):
super().__init__()
self.eps = eps
self.weight = nn.Parameter(torch.ones(dim))

def _norm(self, x):
return x * torch.rsqrt(torch.mean(x * x, dim=-1, keepdim=True) + self.eps)

def forward(self, x: Tensor) -> Tensor:
output = self._norm(x.float()).type_as(x)
return output * self.weight


def apply_scaling(freqs: torch.Tensor, rope_scaling: Dict[str, Any]):
# Check for the presence of the required keys
required_keys = {
Expand Down
Loading