tpu-inference 0.12.0.dev20251213__py3-none-any.whl → 0.13.2.dev20251230__py3-none-any.whl

tpu_inference/layers/jax/sample/sampling_metadata.py

@@ -1,3 +1,17 @@
+# Copyright 2025 Google LLC
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
 import functools
 from dataclasses import dataclass
 from typing import Optional

tpu_inference/layers/jax/transformer_block.py

@@ -1,3 +1,17 @@
+# Copyright 2025 Google LLC
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
 from dataclasses import dataclass
 from typing import Any, Optional, Tuple
 

tpu_inference/layers/vllm/__init__.py

@@ -0,0 +1,13 @@
+# Copyright 2025 Google LLC
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.

tpu_inference/layers/vllm/attention.py

@@ -15,6 +15,7 @@ from vllm.attention.backends.abstract import (AttentionBackend, AttentionImpl,
 from tpu_inference import utils
 from tpu_inference.layers.common.attention_interface import attention
 from tpu_inference.layers.common.attention_metadata import AttentionMetadata
+from tpu_inference.layers.common.quantization import quantize_kv
 from tpu_inference.logger import init_logger
 from tpu_inference.models.vllm.vllm_model_wrapper_context import \
     get_vllm_model_wrapper_context
@@ -117,10 +118,9 @@ class PallasAttentionBackendImpl(AttentionImpl):
         query, key, value = jax_view(query), jax_view(key), jax_view(value)
         q_scale = k_scale = v_scale = None
         if self.kv_cache_quantized_dtype:
-            key, value = utils.quantize_kv(key, value,
-                                           self.kv_cache_quantized_dtype,
-                                           layer._k_scale_float,
-                                           layer._v_scale_float)
+            key, value = quantize_kv(self.kv_cache_quantized_dtype, key, value,
+                                     layer._k_scale_float,
+                                     layer._v_scale_float)
             # TODO(kyuyeunk): Enable w8a8 when VREG spill issue is resolved.
             # q_scale = layer._q_scale_float
             k_scale = layer._k_scale_float

tpu_inference/layers/vllm/fused_moe.py

@@ -1,469 +1,114 @@
-import functools
+# Copyright 2025 Google LLC
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from enum import Enum
 
 import jax
-from jax import numpy as jnp
-from jax import shard_map
-from jax.experimental.pallas.ops.tpu.megablox.gmm import gmm
+import jax.numpy as jnp
+import torch
 from jax.sharding import Mesh
-from jax.sharding import PartitionSpec as P
+from torchax.interop import jax_view, torch_view
+from vllm.model_executor.layers.fused_moe.config import FusedMoEConfig
+from vllm.model_executor.layers.fused_moe.layer import FusedMoE
 
-from tpu_inference.layers.vllm.linear_common import \
-    slice_sharded_tensor_for_concatenation
+from tpu_inference import envs
+from tpu_inference.kernels.fused_moe.v1.kernel import fused_ep_moe
+from tpu_inference.layers.common.fused_moe_gmm import fused_moe_func
+from tpu_inference.logger import init_logger
 
+logger = init_logger(__name__)
 
-def activation_fn(activation: str, x1: jax.Array, x2: jax.Array) -> jax.Array:
-    match activation:
-        case "silu":
-            return jax.nn.silu(x1) * x2
-        case "swigluoai":
-            return _swigluoai(x1, x2)
-        case _:
-            raise NotImplementedError(
-                f"FusedMoE does not support {activation} activation")
 
+class FusedMoEBackend(Enum):
+    FUSED_MOE = "fused_moe"
+    GMM_EP = "gmm_ep"
+    GMM_TP = "gmm_tp"
 
-def _swigluoai(x1: jax.Array,
-               x2: jax.Array,
-               alpha=1.702,
-               limit=7.0) -> jax.Array:
-    x1 = jnp.clip(x1, a_max=limit)
-    x2 = jnp.clip(x2, a_min=-limit, a_max=limit)
 
-    gated_activation = x1 * jax.nn.sigmoid(alpha * x1)
+def select_moe_backend(moe: FusedMoEConfig):
+    if envs.USE_MOE_EP_KERNEL:
+        if moe.use_ep:
+            return FusedMoEBackend.FUSED_MOE
+        logger.warning_once(
+            "USE_MOE_EP_KERNEL=1 but expert parallelism is not "
+            "enabled. Falling back to gmm implementation.")
 
-    return gated_activation * (x2 + 1)
+    if moe.use_ep:
+        return FusedMoEBackend.GMM_EP
 
+    # Use default implementation.
+    return FusedMoEBackend.GMM_TP
 
-def _round_up_to_multiple_of_128_within_limit(x: int, limit: int) -> int:
-    """
-    Rounds the given integer `x` up to the nearest multiple of 128, without
-    exceeding the specified `limit`.
 
-    If `x` is less than or equal to 128, returns 128.
-    If `x` is less than `limit`, returns the smallest multiple of 128 greater
-    than or equal to `x`.
-    If `x` is greater than or equal to `limit`, searches for the largest
-    multiple of 128 less than or equal to `limit` (down to 512) that divides `x`
-    evenly, and returns it.
-    If no such candidate is found, returns `limit`.
-
-    Args:
-        x (int): The integer to round up.
-        limit (int): The upper bound (must be a multiple of 128).
-
-    Returns:
-        int: The rounded value according to the rules above.
-
-    Raises:
-        AssertionError: If `limit` is less than 128 or not a multiple of 128.
-    """
-    assert limit >= 128 and limit % 128 == 0
-    if x <= 128:
-        return 128
-    if x < limit:
-        return (x + 127) // 128 * 128
-    for candidate in range(limit, 511, -128):
-        if x % candidate == 0:
-            return candidate
-    return limit
-
-
-def _get_tiling_size_for_gmm_kernel(m: int, k: int, n: int,
-                                    g: int) -> tuple[int, int, int]:
-    """
-    Calculate optimal tiling sizes for a GMM kernel in a Mixture of Experts
-    (MoE) setting.
-
-    Args:
-        m (int): The total number of tokens.
-        n (int): The output feature dimension.
-        k (int): The input feature dimension.
-        g (int): The number of experts.
-
-    Returns:
-        tuple[int, int, int]: A tuple (tm, tk, tn)
-    """
-
-    # TODO(Chengji): increase the upper limit tiling size of m when we can set
-    # the vmem size to be used for gmm kernel.
-    # NOTE: In average each expert has m // g tokens, but as it might be
-    # unbalanced, here we doubled the token size when choosing tiling size of m.
-    # 2m//g can be either greater or less than 512. If there are 32 tokens and
-    # topk=2, m=topk * num_tokens=64, in this case, 2*m//g will be less than
-    # 512.
-    tm = _round_up_to_multiple_of_128_within_limit(2 * m // g, 512)
-    tm = min(tm, m)  # there's a requirement that m % tm == 0
-    # k/n correspond to n_input_features/n_output_features in the matmul so they
-    # are normally greater than 2048, unless the num shards is large.
-    tk = _round_up_to_multiple_of_128_within_limit(k, 2048)
-    tn = _round_up_to_multiple_of_128_within_limit(n, 2048)
-    return tm, tk, tn
-
-
-def tensor_sharded_gmm_merged_column_parallel(
-    lhs: jax.Array,
-    rhs: jax.Array,
-    rhs_bias: jax.Array | None,
-    group_sizes: jax.Array,
-    mesh: Mesh,
-) -> tuple[jax.Array, jax.Array]:
-
-    def _gmm(lhs, rhs, group_sizes):
-        m, g, n, k = lhs.shape[0], *rhs.shape
-        tm, tk, tn = _get_tiling_size_for_gmm_kernel(m, k, n, g)
-        return gmm(
-            lhs,
-            rhs,
-            group_sizes,
-            preferred_element_type=lhs.dtype,
-            tiling=(tm, tk, tn),
-            transpose_rhs=True,
-            group_offset=jnp.array(0),
-        )
-
-    gmm_result = shard_map(
-        _gmm,
-        mesh=mesh,
-        in_specs=(P("data", None), P(None, "model", None), P("data")),
-        out_specs=(P("data", "model")),
-        check_vma=False,
-    )(lhs, rhs, group_sizes)
-
-    if rhs_bias is not None:
-
-        def _add_bias(gmm_result_local, rhs_bias_local, group_sizes_global):
-            rhs_bias = jnp.repeat(
-                rhs_bias_local,
-                group_sizes_global,
-                0,
-                total_repeat_length=gmm_result_local.shape[0])
-            return gmm_result_local + rhs_bias
-
-        gmm_result = shard_map(
-            _add_bias,
-            mesh=mesh,
-            in_specs=(P("data", "model"), P(None, "model"), P("data")),
-            out_specs=(P("data", "model")),
-        )(gmm_result, rhs_bias, group_sizes)
-    gmm_result = gmm_result.astype(lhs.dtype)
-
-    tp_size = mesh.shape["model"]
-    intermediate_size = gmm_result.shape[-1] // 2
-    output_sizes = [intermediate_size, intermediate_size]
-    return slice_sharded_tensor_for_concatenation(gmm_result, output_sizes,
-                                                  tp_size)
-
-
-def tensor_sharded_gmm_row_parallel(
-    lhs: jax.Array,
-    rhs: jax.Array,
-    rhs_bias: jax.Array | None,
-    group_sizes: jax.Array,
-    mesh: Mesh,
-) -> jax.Array:
-
-    def _gmm_all_reduce(lhs, rhs, group_sizes):
-        m, g, n, k = lhs.shape[0], *rhs.shape
-        tm, tk, tn = _get_tiling_size_for_gmm_kernel(m, k, n, g)
-        out = gmm(
-            lhs,
-            rhs,
-            group_sizes,
-            preferred_element_type=lhs.dtype,
-            tiling=(tm, tk, tn),
-            transpose_rhs=True,
-            group_offset=jnp.array(0),
-        )
-        return jax.lax.psum(out, axis_name="model")
-
-    gmm_result = shard_map(
-        _gmm_all_reduce,
-        mesh=mesh,
-        in_specs=(P("data", "model"), P(None, None, "model"), P("data")),
-        out_specs=(P("data")),
-        check_vma=False,
-    )(lhs, rhs, group_sizes)
-
-    if rhs_bias is not None:
-
-        def _add_bias(gmm_result_local, rhs_bias_local, group_sizes_global):
-            rhs_bias = jnp.repeat(
-                rhs_bias_local,
-                group_sizes_global,
-                0,
-                total_repeat_length=gmm_result_local.shape[0])
-            return gmm_result_local + rhs_bias
-
-        gmm_result = shard_map(
-            _add_bias,
-            mesh=mesh,
-            in_specs=(P("data"), P(), P("data")),
-            out_specs=(P("data")),
-        )(gmm_result, rhs_bias, group_sizes)
-
-    return gmm_result.astype(lhs.dtype)
-
-
-def expert_sharded_gmm(
-    lhs: jax.Array,
-    rhs: jax.Array,
-    group_sizes: jax.Array,
-    mesh: Mesh,
-) -> jax.Array:
-    ep_size = mesh.shape["model"]
-
-    num_experts = rhs.shape[0]
-    num_experts_per_shard = num_experts // ep_size
-    group_offset = jnp.arange(0, num_experts, num_experts_per_shard)
-
-    def _gmm(lhs, rhs, group_sizes, group_offset):
-        m, g, n, k = lhs.shape[0], *rhs.shape
-        tm, tk, tn = _get_tiling_size_for_gmm_kernel(m, k, n, g)
-
-        gmm_res = gmm(
-            lhs=lhs,
-            rhs=rhs,
-            group_sizes=group_sizes,
-            preferred_element_type=lhs.dtype,
-            tiling=(tm, tk, tn),
-            transpose_rhs=True,
-            group_offset=group_offset[0],
-        )
-        return gmm_res
-
-    # The result from gmm on each shard has the same shape, but only the rows
-    # for this shard has non-zero values. Taking below as an working example:
-    #       A, A, A, A     0, 0, 0, 0     0, 0, 0, 0     0, 0, 0, 0
-    #       A, A, A, A     0, 0, 0, 0     0, 0, 0, 0     0, 0, 0, 0
-    #       A, A, A, A     0, 0, 0, 0     0, 0, 0, 0     0, 0, 0, 0
-    #       0, 0, 0, 0     B, B, B, B     0, 0, 0, 0     0, 0, 0, 0
-    #       0, 0, 0, 0     B, B, B, B     0, 0, 0, 0     0, 0, 0, 0
-    #       0, 0, 0, 0     0, 0, 0, 0     C, C, C, C     0, 0, 0, 0
-    #       0, 0, 0, 0     0, 0, 0, 0     C, C, C, C     0, 0, 0, 0
-    #       0, 0, 0, 0     0, 0, 0, 0     C, C, C, C     0, 0, 0, 0
-    #       0, 0, 0, 0     0, 0, 0, 0     C, C, C, C     0, 0, 0, 0
-    #       0, 0, 0, 0     0, 0, 0, 0     C, C, C, C     0, 0, 0, 0
-    #       0, 0, 0, 0     0, 0, 0, 0     0, 0, 0, 0     D, D, D, D
-    #       0, 0, 0, 0     0, 0, 0, 0     0, 0, 0, 0     D, D, D, D
-    #       0, 0, 0, 0     0, 0, 0, 0     0, 0, 0, 0     D, D, D, D
-    #       0, 0, 0, 0     0, 0, 0, 0     0, 0, 0, 0     D, D, D, D
-    #        shard-0        shard-1        shard-2        shard-3
-    # Each shards has 3 (row A), 2 (row B), 5 (row C) and 4 (row D).
-    gmm_res = shard_map(
-        _gmm,
-        mesh=mesh,
-        in_specs=(P(), P("model", None, None), P(), P("model")),
-        out_specs=(P("model", None)),
-        check_vma=False,
-    )(lhs, rhs, group_sizes, group_offset)
-
-    # For i-th shard, it is responsible groups (AKA experts) from
-    # i*num_experts_per_shard to (i+1)*num_experts_per_shard We sum them up to
-    # get total rows in that shard, and that is the size for shard to send to
-    # its peers. This is also the number of non-zero rows from the gmm results.
-    # In the working example, send_sizes would be [3, 2, 5, 4].
-
-    # group_sizes has shape of [num_tokens_per_shard * num_experts_per_shard].
-    # So reshaping to [num_tokens_per_shard, num_experts_per_shard] and applying
-    # sum(axis=1) will get desired send_sizes shaped [num_tokens_per_shard].
-    send_sizes = group_sizes.reshape(-1, num_experts_per_shard).sum(axis=1)
-    # In the working example, input_offsets would be [0, 3, 5, 10]
-    input_offsets = jnp.concatenate((jnp.array([0]), send_sizes.cumsum()[:-1]))
-    output_offsets = input_offsets
-    recv_sizes = send_sizes
-
-    def _ragged_all_to_all(operand, input_offsets, send_sizes, output_offsets,
-                           recv_sizes):
-        output = jnp.zeros_like(operand)
-
-        # input_offsets, send_sizes and output_offsets are sharded and there is
-        # only 1 elemnt in each shard, we are taking the 0-th element from them
-        # just so that jnp.repeat generates the arrays with correct shape.
-        input_offsets_of_shard = jnp.repeat(input_offsets[0], ep_size)
-        send_sizes_of_shard = jnp.repeat(send_sizes[0], ep_size)
-        output_offsets_of_shard = jnp.repeat(output_offsets[0], ep_size)
-
-        # recv_sizes is replicated across shards, because all the shards receive
-        # the same data and write to the output in the same way (same
-        # output_offsets and same recv_sizes) and thus generates replicated
-        # output.
-        recv_sizes_of_shard = recv_sizes
-
-        # In the working example, for each shard, the values of the offsets and
-        # sizes would be:
-        #                                shard-0         shard-1         shard-2         shard-3
-        # input_offsets_of_shard       [0, 0, 0, 0]    [3, 3, 3, 3]    [5, 5, 5, 5]    [10,10,10,10]
-        # send_sizes_of_shard          [3, 3, 3, 3]    [2, 2, 2, 2]    [5, 5, 5, 5]    [4, 4, 4, 4 ]
-        # output_offsets_of_shard      [0, 0, 0, 0]    [0, 0, 0, 0]    [0, 0, 0, 0]    [10,10,10,10]
-        # recv_sizes_of_shard          [3, 2, 5, 4]    [3, 2, 5, 4]    [3, 2, 5, 4]    [3, 2, 5, 4]
-        return jax.lax.ragged_all_to_all(operand,
-                                         output,
-                                         input_offsets_of_shard,
-                                         send_sizes_of_shard,
-                                         output_offsets_of_shard,
-                                         recv_sizes_of_shard,
-                                         axis_name="model")
-
-    # Use ragged_all_to_all to send the result from gmm for each expert to all
-    # the shards.  In the working example, the result would be:
-    #       A, A, A, A     A, A, A, A     A, A, A, A     A, A, A, A
-    #       A, A, A, A     A, A, A, A     A, A, A, A     A, A, A, A
-    #       A, A, A, A     A, A, A, A     A, A, A, A     A, A, A, A
-    #       B, B, B, B     B, B, B, B     B, B, B, B     B, B, B, B
-    #       B, B, B, B     B, B, B, B     B, B, B, B     B, B, B, B
-    #       C, C, C, C     C, C, C, C     C, C, C, C     C, C, C, C
-    #       C, C, C, C     C, C, C, C     C, C, C, C     C, C, C, C
-    #       C, C, C, C     C, C, C, C     C, C, C, C     C, C, C, C
-    #       C, C, C, C     C, C, C, C     C, C, C, C     C, C, C, C
-    #       C, C, C, C     C, C, C, C     C, C, C, C     C, C, C, C
-    #       D, D, D, D     D, D, D, D     D, D, D, D     D, D, D, D
-    #       D, D, D, D     D, D, D, D     D, D, D, D     D, D, D, D
-    #       D, D, D, D     D, D, D, D     D, D, D, D     D, D, D, D
-    #       D, D, D, D     D, D, D, D     D, D, D, D     D, D, D, D
-    #        shard-0        shard-1        shard-2        shard-3
-    return shard_map(
-        _ragged_all_to_all,
-        mesh=mesh,
-        in_specs=(P("model", None), P("model"), P("model"), P("model"), P()),
-        out_specs=(P()),
-        check_vma=False,
-    )(gmm_res, input_offsets, send_sizes, output_offsets, recv_sizes)
-
-
-@functools.partial(
-    jax.jit,
-    static_argnames=(
-        "topk",
-        "renormalize",
-        "mesh",
-        "use_ep",
-        "activation",
-    ),
-)
-def fused_moe_func(
-    hidden_states: jax.Array,
-    w1: jax.Array,
-    w2: jax.Array,
-    w1_bias: jax.Array | None,
-    w2_bias: jax.Array | None,
-    gating_output: jax.Array,
-    topk: int,
-    renormalize: bool,
+def fused_moe_apply(
+    layer: torch.nn.Module,
+    x: torch.Tensor,
+    router_logits: torch.Tensor,
+    moe_backend: FusedMoEBackend,
     mesh: Mesh,
-    use_ep: bool,
-    activation: str,
-) -> jax.Array:
-    """
-    Route tokens in hidden_states into each experts based on routing
-    information in gating_out and performs moe with w1 and w2 weights.
-
-    Args:
-        hidden_states: [num_tokens, hidden_size]
-        w1: first moe weights [num_experts, intermediate_size * 2, hidden_size]
-        w2: second moe weights [num_experts, hidden_size, intermediate_size]
-        w1_bias: optional bias of w1 [num_experts, intermediate_size * 2]
-        w2_bias: optional bias of w2 [num_experts, hidden_size]
-        gating_output: routing information of tokens [num_tokens, num_experts]
-        topk: number of experts to choose per token.
-        renormalize: normalize gating_output.
-        mesh: mesh to perform moe.
-        use_ep: use expert parallelism.
-        activation: activation function to perform on the output of w1.
-
-    Returns:
-        Output of moe operation [num_tokens, hidden_size]
-    """
-    if use_ep and (w1_bias is not None or w2_bias is not None):
-        raise NotImplementedError(
-            "Bias is not supported when using expert parallelism.")
-
-    num_tokens = hidden_states.shape[0]
-    global_num_experts, hidden_size, intermediate_size = w2.shape
-    dtype = hidden_states.dtype
-
-    assert (num_tokens * topk) % 16 == 0, (
-        "The kernel requires num_tokens * topk to be a multiple of "
-        f"16 but got {num_tokens}*{topk}={num_tokens*topk}")
-    assert hidden_states.shape == (num_tokens, hidden_size)
-    assert gating_output.shape == (num_tokens, global_num_experts)
-    assert w1.shape == (global_num_experts, intermediate_size * 2, hidden_size)
-
-    topk_weights = jax.nn.softmax(gating_output.astype(jnp.float32), axis=-1)
-    topk_weights, topk_indices = jax.lax.top_k(topk_weights, k=topk)
-    if renormalize:
-        topk_weights = topk_weights / topk_weights.sum(axis=-1, keepdims=True)
-    topk_weights = topk_weights.astype(dtype)
-
-    def _process_tokens_locally(hidden_states_local, topk_indices_local):
-        num_tokens_local = hidden_states_local.shape[0]
-        topk_indices_flat = topk_indices_local.flatten()
-        topk_argsort_indices = jnp.argsort(topk_indices_flat)
-        topk_argsort_revert_indices = jnp.argsort(topk_argsort_indices)
-        token_indices = jnp.arange(num_tokens_local,
-                                   dtype=jnp.int32).repeat(topk)
-        token_indices_sorted = token_indices[topk_argsort_indices]
-        group_sizes_local = jnp.bincount(topk_indices_flat,
-                                         length=global_num_experts)
-
-        x = hidden_states_local[token_indices_sorted]
-        return x, group_sizes_local, topk_argsort_revert_indices
-
-    x, group_sizes, topk_argsort_revert_indices = shard_map(
-        _process_tokens_locally,
-        mesh=mesh,
-        in_specs=(P("data", None), P("data", None)),
-        out_specs=(P("data", None), P("data"), P("data")),
-    )(hidden_states, topk_indices)
-
-    if use_ep:
-        x = expert_sharded_gmm(
-            x,
-            w1,
-            group_sizes,
-            mesh=mesh,
-        )
-        x1, x2 = jnp.split(x, 2, -1)
-
-        x = activation_fn(activation, x1, x2)
-
-        x = expert_sharded_gmm(
-            x,
-            w2,
-            group_sizes,
-            mesh=mesh,
-        )
-    else:
-        x1, x2 = tensor_sharded_gmm_merged_column_parallel(
-            x,
-            w1,
-            w1_bias,
-            group_sizes,
-            mesh=mesh,
-        )
-
-        x = activation_fn(activation, x1, x2)
-
-        x = tensor_sharded_gmm_row_parallel(
-            x,
-            w2,
-            w2_bias,
-            group_sizes,
-            mesh=mesh,
-        )
-
-    def _finalize_output(x_local, topk_argsort_revert_indices_local,
-                         topk_weights_local):
-        x_local = x_local[topk_argsort_revert_indices_local].reshape(
-            -1, topk, hidden_size)
-        x_local = x_local * jnp.expand_dims(topk_weights_local, axis=-1)
-        x_local = x_local.sum(axis=-2)
-        return x_local
-
-    x = shard_map(
-        _finalize_output,
-        mesh=mesh,
-        in_specs=(P("data", None), P("data"), P("data", None)),
-        out_specs=(P("data", None)),
-    )(x, topk_argsort_revert_indices, topk_weights)
-
-    return x[:num_tokens, :hidden_size]
+    extra_backend_kwargs: dict,
+) -> torch.Tensor:
+    assert isinstance(layer, FusedMoE)
+    if layer.scoring_func != "softmax":
+        raise NotImplementedError("Only softmax is supported for scoring_func")
+
+    x = jax_view(x)
+    gating_output = jax_view(router_logits)
+
+    w13_weight = jax_view(layer.w13_weight)
+    w13_weight_scale = jax_view(getattr(layer, "w13_weight_scale", None))
+    w13_bias = jax_view(getattr(layer, "w13_bias", None))
+    w2_weight = jax_view(layer.w2_weight)
+    w2_weight_scale = jax_view(getattr(layer, "w2_weight_scale", None))
+    w2_bias = jax_view(getattr(layer, "w2_bias", None))
+
+    with jax.named_scope(layer._get_name()):
+        match moe_backend:
+            case FusedMoEBackend.FUSED_MOE:
+                actual_hidden_size = x.shape[-1]
+                padding_size = w13_weight.shape[-2] - actual_hidden_size
+                x = jnp.pad(x, ((0, 0), (0, padding_size)))
+                output = fused_ep_moe(
+                    mesh=mesh,
+                    tokens=x,
+                    w1=w13_weight,
+                    w2=w2_weight,
+                    w1_scale=w13_weight_scale,
+                    w2_scale=w2_weight_scale,
+                    b1=w13_bias,
+                    b2=w2_bias,
+                    gating_output=gating_output,
+                    top_k=layer.top_k,
+                    renormalize_topk_logits=layer.renormalize,
+                    act_fn=layer.activation,
+                    **extra_backend_kwargs,
+                )[:, :actual_hidden_size]
+            case FusedMoEBackend.GMM_EP | FusedMoEBackend.GMM_TP:
+                output = fused_moe_func(
+                    hidden_states=x,
+                    w1=w13_weight,
+                    w2=w2_weight,
+                    w1_scale=w13_weight_scale,
+                    w2_scale=w2_weight_scale,
+                    w1_bias=w13_bias,
+                    w2_bias=w2_bias,
+                    gating_output=gating_output,
+                    topk=layer.top_k,
+                    renormalize=layer.renormalize,
+                    mesh=mesh,
+                    use_ep=layer.use_ep,
+                    activation=layer.activation,
+                )
+
+        return torch_view(output)