Skip to content

Fix the Rotary Position Embedding (RoPE) parameter passing in the GLM5 mode #1650

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
hanxdmech-ship-it wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Fix the Rotary Position Embedding (RoPE) parameter passing in the GLM5 mode #1650

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
3 changes: 2 additions & 1 deletion slime_plugins/models/glm5/glm5.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -421,14 +421,15 @@ def get_absorb_query_key_value_tensors(
assert hidden_states.ndim == 3, f"hidden_states should be 3D, [s, b, n*h], got {hidden_states.ndim}D"
assert packed_seq_params is not None

packed_seq = packed_seq_params is not None and packed_seq_params.qkv_format == 'thd'
# =========================================
# Prepare RoPE and seqlen related params
# =========================================
rotary_seq_len = self.rotary_pos_emb.get_rotary_seq_len(
inference_context, None, hidden_states, self.config, packed_seq_params
)
# TODO: support apply_rope_fusion
rotary_pos_emb, mscale = self.rotary_pos_emb(rotary_seq_len, packed_seq_params=packed_seq_params)
rotary_pos_emb, mscale = self.rotary_pos_emb(rotary_seq_len, packed_seq=packed_seq)

cu_seqlens_q = packed_seq_params.cu_seqlens_q
cu_seqlens_kv = packed_seq_params.cu_seqlens_kv
Expand Down