Skip to content

fix: respect explicit config.head_dim in cnets attention #342

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
AchuthReddy-16 wants to merge 1 commit into
base: main
Choose a base branch
from
Open

fix: respect explicit config.head_dim in cnets attention #342

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
13 changes: 5 additions & 8 deletions eagle/model/cnets.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -196,16 +196,13 @@ def __init__(self, config):
self.config = config
self.hidden_size = config.hidden_size
self.num_heads = config.num_attention_heads
self.head_dim = self.hidden_size // self.num_heads
if hasattr(config, "head_dim") and config.head_dim is not None:
self.head_dim = config.head_dim
else:
self.head_dim = self.hidden_size // self.num_heads
self.num_key_value_heads = config.num_key_value_heads
self.num_key_value_groups = self.num_heads // self.num_key_value_heads
self.max_position_embeddings = config.max_position_embeddings

if (self.head_dim * self.num_heads) != self.hidden_size:
raise ValueError(
f"hidden_size must be divisible by num_heads (got `hidden_size`: {self.hidden_size}"
f" and `num_heads`: {self.num_heads})."
)
self.q_proj = nn.Linear(self.hidden_size * 2, self.num_heads * self.head_dim, bias=False)
self.k_proj = nn.Linear(self.hidden_size * 2, self.num_key_value_heads * self.head_dim, bias=False)
self.v_proj = nn.Linear(self.hidden_size * 2, self.num_key_value_heads * self.head_dim, bias=False)
Expand Down Expand Up @@ -318,7 +315,7 @@ def forward(
)

attn_output = attn_output.transpose(1, 2).contiguous()
attn_output = attn_output.reshape(bsz, q_len, self.hidden_size)
attn_output = attn_output.reshape(bsz, q_len, self.num_heads * self.head_dim)

if self.config.pretraining_tp > 1:
attn_output = attn_output.split(self.hidden_size // self.config.pretraining_tp, dim=2)
Expand Down