diff --git a/tunix/models/gemma4/attention.py b/tunix/models/gemma4/attention.py index 42a1d87c0..cc6b36f43 100644 --- a/tunix/models/gemma4/attention.py +++ b/tunix/models/gemma4/attention.py @@ -879,7 +879,7 @@ def __call__( def _checkpointed_block(state, *args, **kwargs): module = nnx.merge(graphdef, state) - return module.block(*args, **kwargs) + return module.block(*args, **kwargs, force_eager=force_eager) return jax.checkpoint(_checkpointed_block)( state, @@ -889,7 +889,6 @@ def _checkpointed_block(state, *args, **kwargs): attn_mask, kv_shared_cache, segment_ids, - force_eager, ) else: return self.block(