Skip to content

converters: write hidden_activation=relu2 for bitnet models (fixes #602) #603

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
msrovani wants to merge 1 commit into
base: main
Choose a base branch
from
+5 −0
Open

converters: write hidden_activation=relu2 for bitnet models (fixes #602) #603

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: 4 additions & 0 deletions utils/convert-hf-to-gguf-bitnet.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1043,6 +1043,10 @@ def set_vocab(self):
def set_gguf_parameters(self):
super().set_gguf_parameters()

# bitnet-b1.58-2B-4T uses relu^2 in the FFN; the runtime reads it from
# <arch>.hidden_activation (see llama_model_bitnet::load_arch_hparams).
self.gguf_writer.add_hidden_act("relu2")

self.gguf_writer.add_vocab_size(self.hparams["vocab_size"])

# rope dimension count (required for correct positional encoding)
Expand Down
1 change: 1 addition & 0 deletions utils/convert-ms-to-gguf-bitnet.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1157,6 +1157,7 @@ def add_meta_arch(self, params: Params) -> None:
name = str(params.path_model.parent).split('/')[-1]

self.gguf.add_name (name)
self.gguf.add_hidden_act ("relu2")
self.gguf.add_vocab_size (params.n_vocab)
self.gguf.add_context_length (params.n_ctx)
self.gguf.add_embedding_length (params.n_embd)
Expand Down