Skip to content

Disable static cache on certain MoE models #39108

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

Closed
ivarflakstad wants to merge 4 commits into from
Closed

Disable static cache on certain MoE models #39108

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
a9703d0
These MoEs do not support static cache (atm)
ivarflakstad Jun 29, 2025
3b20325
Merge branch 'main' into set-supports-static-cache-false-on-moes
ivarflakstad Jun 29, 2025
ee85f48
Merge branch 'main' into set-supports-static-cache-false-on-moes
ivarflakstad Jul 28, 2025
ccedf5b
Remove setting that should be inherited. Formatting.
ivarflakstad Jul 28, 2025
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: 1 addition & 2 deletions src/transformers/models/deepseek_v3/modeling_deepseek_v3.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -497,8 +497,7 @@ class DeepseekV3PreTrainedModel(PreTrainedModel):
_supports_flash_attn = True
_supports_sdpa = True
_supports_flex_attn = True

_can_compile_fullgraph = True
_can_compile_fullgraph = False
_supports_attention_backend = True
_can_record_outputs = {
"hidden_states": DeepseekV3DecoderLayer,
Expand Down
2 changes: 2 additions & 0 deletions src/transformers/models/deepseek_v3/modular_deepseek_v3.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -338,6 +338,8 @@ def __init__(self, config: DeepseekV3Config, layer_idx: int):


class DeepseekV3PreTrainedModel(LlamaPreTrainedModel):
_can_compile_fullgraph = False

def _init_weights(self, module):
LlamaPreTrainedModel._init_weights(module)
if isinstance(module, DeepseekV3TopkRouter):
Expand Down
3 changes: 1 addition & 2 deletions src/transformers/models/dots1/modeling_dots1.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -417,8 +417,7 @@ class Dots1PreTrainedModel(PreTrainedModel):
_supports_flash_attn = True
_supports_sdpa = True
_supports_flex_attn = True

_can_compile_fullgraph = True
_can_compile_fullgraph = False
_supports_attention_backend = True
_can_record_outputs = {
"hidden_states": Dots1DecoderLayer,
Expand Down