Skip to content

Support GLM 4.6 #34

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
gmlwns2000 wants to merge 1 commit into
base: deepauto/dev
Choose a base branch
from
Open

Support GLM 4.6 #34

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
1 change: 1 addition & 0 deletions python/sglang/srt/layers/attention/hip_attention.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -290,6 +290,7 @@ def forward_extend(
# For multi-head latent attention
q_rope=q_rope,
k_rope=k_rope,
sinks=sinks,
)
else:
if not self.is_kv_cache_offload_enabled:
Expand Down
8 changes: 4 additions & 4 deletions python/sglang/srt/models/glm4_moe.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -12,7 +12,7 @@
# limitations under the License.
# ==============================================================================

"""Inference-only GLM-4.5 model compatible with HuggingFace weights"""
"""Inference-only GLM-4.5, GLM-4.6 model compatible with HuggingFace weights"""

import logging
from typing import Any, Dict, Iterable, Optional, Tuple
Expand Down Expand Up @@ -812,9 +812,9 @@ def determine_num_fused_shared_experts(
or self.config.architectures[0] != architecture
or self.config.n_shared_experts != 1
):
disable_reason = "Only GLM-4.5 on NV-platform with capability >= 80 can use shared experts fusion optimization."
disable_reason = "Only GLM-4.5 or GLM-4.6 on NV-platform with capability >= 80 can use shared experts fusion optimization."
elif get_moe_expert_parallel_world_size() > 1:
disable_reason = "Deepseek and GLM-4.5 can not use shared experts fusion optimization under expert parallelism."
disable_reason = "Deepseek and GLM-4.5 or GLM-4.6 can not use shared experts fusion optimization under expert parallelism."

if disable_reason is not None:
global_server_args_dict["disable_shared_experts_fusion"] = True
Expand Down Expand Up @@ -1107,4 +1107,4 @@ def load_weights(self, weights: Iterable[Tuple[str, torch.Tensor]], is_nextn=Fal
weight_loader(param, loaded_weight)


EntryClass = [Glm4MoeForCausalLM]
EntryClass = [Glm4MoeForCausalLM]
Loading