Update Optional[x] -> x | None and Union[x, y] to x | y (#26633)

Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
Harry Mellor
2025-10-12 09:51:31 -07:00
committed by GitHub
parent 9bb38130cb
commit 8fcaaf6a16
944 changed files with 9490 additions and 10121 deletions
+23 -24
View File
@@ -16,7 +16,6 @@
# See the License for the specific language governing permissions and
# limitations under the License.
from collections.abc import Iterable
from typing import Optional, Union
import torch
from torch import nn
@@ -196,7 +195,7 @@ class Gemma3nLaurelBlock(nn.Module):
laurel_rank: int,
rms_norm_eps: float,
*,
quant_config: Optional[QuantizationConfig] = None,
quant_config: QuantizationConfig | None = None,
prefix: str,
) -> None:
super().__init__()
@@ -236,7 +235,7 @@ class Gemma3nMLP(nn.Module):
intermediate_size: int,
hidden_activation: str,
activation_sparsity: float = 0.0,
quant_config: Optional[QuantizationConfig] = None,
quant_config: QuantizationConfig | None = None,
prefix: str = "",
) -> None:
super().__init__()
@@ -285,8 +284,8 @@ class Gemma3nAttention(nn.Module):
num_kv_heads: int,
head_dim: int,
max_position_embeddings: int,
cache_config: Optional[CacheConfig] = None,
quant_config: Optional[QuantizationConfig] = None,
cache_config: CacheConfig | None = None,
quant_config: QuantizationConfig | None = None,
prefix: str = "",
) -> None:
super().__init__()
@@ -412,8 +411,8 @@ class Gemma3nDecoderLayer(nn.Module):
def __init__(
self,
config: Gemma3nTextConfig,
cache_config: Optional[CacheConfig] = None,
quant_config: Optional[QuantizationConfig] = None,
cache_config: CacheConfig | None = None,
quant_config: QuantizationConfig | None = None,
prefix: str = "",
) -> None:
super().__init__()
@@ -650,7 +649,7 @@ class Gemma3nSelfDecoder(nn.Module):
def get_per_layer_inputs(
self,
hidden_states_0: torch.Tensor,
per_layer_inputs: Optional[torch.Tensor],
per_layer_inputs: torch.Tensor | None,
) -> torch.Tensor:
per_layer_projection = self.per_layer_model_projection(hidden_states_0)
per_layer_projection = per_layer_projection.reshape(
@@ -687,8 +686,8 @@ class Gemma3nSelfDecoder(nn.Module):
self,
input_ids: torch.Tensor,
positions: torch.Tensor,
inputs_embeds: Optional[torch.Tensor] = None,
per_layer_inputs: Optional[torch.Tensor] = None,
inputs_embeds: torch.Tensor | None = None,
per_layer_inputs: torch.Tensor | None = None,
**kwargs,
) -> tuple[torch.Tensor, torch.Tensor]:
if inputs_embeds is not None:
@@ -870,8 +869,8 @@ class Gemma3nTextModel(nn.Module, SupportsQuant):
self,
input_ids: torch.Tensor,
positions: torch.Tensor,
inputs_embeds: Optional[torch.Tensor] = None,
per_layer_inputs: Optional[torch.Tensor] = None,
inputs_embeds: torch.Tensor | None = None,
per_layer_inputs: torch.Tensor | None = None,
**kwargs,
) -> torch.Tensor:
logits_indices_padded, num_logits_indices = None, None
@@ -947,8 +946,8 @@ class Gemma3nTextModel(nn.Module, SupportsQuant):
self,
input_ids: torch.Tensor,
positions: torch.Tensor,
inputs_embeds: Optional[torch.Tensor] = None,
per_layer_inputs: Optional[torch.Tensor] = None,
inputs_embeds: torch.Tensor | None = None,
per_layer_inputs: torch.Tensor | None = None,
**kwargs,
) -> torch.Tensor:
hidden_states, per_layer_inputs = self.self_decoder(
@@ -990,13 +989,13 @@ class Gemma3nTextModel(nn.Module, SupportsQuant):
def forward(
self,
input_ids: Optional[torch.Tensor],
input_ids: torch.Tensor | None,
positions: torch.Tensor,
per_layer_inputs: Optional[torch.Tensor] = None,
intermediate_tensors: Optional[IntermediateTensors] = None,
inputs_embeds: Optional[torch.Tensor] = None,
per_layer_inputs: torch.Tensor | None = None,
intermediate_tensors: IntermediateTensors | None = None,
inputs_embeds: torch.Tensor | None = None,
**kwargs,
) -> Union[torch.Tensor, IntermediateTensors]:
) -> torch.Tensor | IntermediateTensors:
if self.fast_prefill_enabled:
hidden_states = self.fast_prefill_forward(
input_ids,
@@ -1116,11 +1115,11 @@ class Gemma3nForCausalLM(nn.Module):
input_ids: torch.Tensor,
positions: torch.Tensor,
*,
per_layer_inputs: Optional[torch.Tensor] = None,
intermediate_tensors: Optional[IntermediateTensors] = None,
inputs_embeds: Optional[torch.Tensor] = None,
per_layer_inputs: torch.Tensor | None = None,
intermediate_tensors: IntermediateTensors | None = None,
inputs_embeds: torch.Tensor | None = None,
**kwargs,
) -> Union[torch.Tensor, IntermediateTensors]:
) -> torch.Tensor | IntermediateTensors:
hidden_states = self.model(
input_ids,
positions,
@@ -1134,7 +1133,7 @@ class Gemma3nForCausalLM(nn.Module):
def compute_logits(
self,
hidden_states: torch.Tensor,
) -> Optional[torch.Tensor]:
) -> torch.Tensor | None:
logits = self.logits_processor(self.model.embed_tokens, hidden_states)
return logits