Skip to content

added "tie_word_emebeddings" setting necessary for Llama 3.2 #1895

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 1 commit into from
Sep 30, 2024
Merged

added "tie_word_emebeddings" setting necessary for Llama 3.2 #1895

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
4 changes: 3 additions & 1 deletion keras_hub/src/models/llama/llama_backbone.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -90,13 +90,14 @@ def __init__(
layer_norm_epsilon=1e-6,
dropout=0,
dtype=None,
tie_word_embeddings=False,
**kwargs,
):
# === Layers ===
self.token_embedding = ReversibleEmbedding(
input_dim=vocabulary_size,
output_dim=hidden_dim,
tie_weights=False,
tie_weights=tie_word_embeddings,
embeddings_initializer=_llama_kernel_initializer(stddev=0.01),
dtype=dtype,
name="token_embedding",
Expand Down Expand Up @@ -155,6 +156,7 @@ def __init__(
self.rope_scaling_factor = rope_scaling_factor
self.layer_norm_epsilon = layer_norm_epsilon
self.dropout = dropout
self.tie_word_embeddings = tie_word_embeddings

def get_config(self):
config = super().get_config()
Expand Down
16 changes: 10 additions & 6 deletions keras_hub/src/utils/transformers/convert_llama3.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -14,6 +14,7 @@ def convert_backbone_config(transformers_config):
"hidden_dim": transformers_config["hidden_size"],
"intermediate_dim": transformers_config["intermediate_size"],
"num_key_value_heads": transformers_config["num_key_value_heads"],
"tie_word_embeddings": transformers_config["tie_word_embeddings"],
}


Expand All @@ -22,12 +23,15 @@ def convert_weights(backbone, loader, transformers_config):
keras_variable=backbone.get_layer("token_embedding").embeddings,
hf_weight_key="model.embed_tokens.weight",
)
loader.port_weight(
keras_variable=backbone.get_layer("token_embedding").reverse_embeddings,
hf_weight_key="lm_head.weight",
# rearrange_pattern="b a -> a b",
hook_fn=lambda hf_tensor, _: np.transpose(hf_tensor, axes=(1, 0)),
)
if not backbone.tie_word_embeddings:
loader.port_weight(
keras_variable=backbone.get_layer(
"token_embedding"
).reverse_embeddings,
hf_weight_key="lm_head.weight",
# rearrange_pattern="b a -> a b",
hook_fn=lambda hf_tensor, _: np.transpose(hf_tensor, axes=(1, 0)),
)

def transpose_and_reshape(x, shape):
return np.reshape(np.transpose(x), shape)
Expand Down