sglang 0.4.6.post2__py3-none-any.whl → 0.4.6.post3__py3-none-any.whl

sglang/srt/layers/moe/ep_moe/kernels.py

@@ -5,16 +5,23 @@ import torch
 import triton
 import triton.language as tl
 
-from sglang.srt.distributed import get_tensor_model_parallel_rank
 from sglang.srt.layers.quantization.fp8_kernel import per_token_group_quant_fp8
 from sglang.srt.utils import is_cuda
 
+logger = logging.getLogger(__name__)
+
 _is_cuda = is_cuda()
 if _is_cuda:
     from sglang.srt.layers.quantization.fp8_kernel import (
-        sglang_per_token_group_quant_fp8,
+        sglang_per_token_group_quant_fp8 as per_token_group_quant_fp8,
     )
-logger = logging.getLogger(__name__)
+
+    try:
+        from deep_gemm import ceil_div
+    except ImportError:
+        logger.error(f"Failed to import ceil_div from deep_gemm.")
+
+import triton.language as tl
 
 
 @triton.jit
@@ -654,10 +661,7 @@ def grouped_gemm_triton(
     if block_shape is not None:
         assert len(block_shape) == 2
         block_n, block_k = block_shape[0], block_shape[1]
-        if _is_cuda:
-            a, scale_a = sglang_per_token_group_quant_fp8(a, block_k)
-        else:
-            a, scale_a = per_token_group_quant_fp8(a, block_k)
+        a, scale_a = per_token_group_quant_fp8(a, block_k)
 
         assert triton.cdiv(a.shape[-1], block_k) == scale_a.shape[-1]
         assert triton.cdiv(b.shape[-2], block_n) == scale_b.shape[-2]
@@ -707,3 +711,334 @@ def grouped_gemm_triton(
         **config,
     )
     return c
+
+
+@triton.jit
+def _fwd_kernel_ep_scatter_1(
+    num_recv_tokens_per_expert,
+    expert_start_loc,
+    m_indices,
+    num_experts: tl.constexpr,
+    BLOCK_E: tl.constexpr,
+    BLOCK_EXPERT_NUM: tl.constexpr,
+):
+    cur_expert = tl.program_id(0)
+
+    offset_cumsum = tl.arange(0, BLOCK_EXPERT_NUM)
+    tokens_per_expert = tl.load(
+        num_recv_tokens_per_expert + offset_cumsum,
+        mask=offset_cumsum < num_experts,
+        other=0,
+    )
+    cumsum = tl.cumsum(tokens_per_expert) - tokens_per_expert
+    tl.store(expert_start_loc + offset_cumsum, cumsum, mask=offset_cumsum < num_experts)
+
+    cur_expert_start = tl.load(expert_start_loc + cur_expert)
+    cur_expert_token_num = tl.load(num_recv_tokens_per_expert + cur_expert)
+
+    m_indices_start_ptr = m_indices + cur_expert_start
+    off_expert = tl.arange(0, BLOCK_E)
+
+    for start_m in tl.range(0, cur_expert_token_num, BLOCK_E, num_stages=4):
+        tl.store(
+            m_indices_start_ptr + start_m + off_expert,
+            cur_expert,
+        )
+
+
+@triton.jit
+def _fwd_kernel_ep_scatter_2(
+    total_token_num,
+    expert_start_loc,
+    recv_x,
+    recv_x_stride0,
+    recv_x_stride1,
+    recv_x_scale,
+    recv_x_scale_stride0,
+    recv_x_scale_stride1,
+    recv_topk,
+    recv_topk_stride0,
+    recv_topk_stride1,
+    output_tensor,
+    output_tensor_stride0,
+    output_tensor_stride1,
+    output_tensor_scale,
+    output_tensor_scale_stride0,
+    output_tensor_scale_stride1,
+    output_index,
+    output_index_stride0,
+    output_index_stride1,
+    topk_num: tl.constexpr,
+    HIDDEN_SIZE: tl.constexpr,
+    HIDDEN_SIZE_PAD: tl.constexpr,
+    SCALE_HIDDEN_SIZE: tl.constexpr,
+    SCALE_HIDDEN_SIZE_PAD: tl.constexpr,
+):
+    start_token_id = tl.program_id(0)
+    grid_num = tl.num_programs(0)
+
+    offset_in = tl.arange(0, HIDDEN_SIZE_PAD)
+    mask = offset_in < HIDDEN_SIZE
+
+    offset_in_s = tl.arange(0, SCALE_HIDDEN_SIZE_PAD)
+    mask_s = offset_in_s < SCALE_HIDDEN_SIZE
+
+    for token_id in range(start_token_id, total_token_num, grid_num):
+        to_copy = tl.load(recv_x + token_id * recv_x_stride0 + offset_in, mask=mask)
+        to_copy_s = tl.load(
+            recv_x_scale + token_id * recv_x_scale_stride0 + offset_in_s, mask=mask_s
+        )
+
+        for topk_index in tl.range(0, topk_num, 1, num_stages=4):
+            expert_id = tl.load(recv_topk + token_id * recv_topk_stride0 + topk_index)
+            if expert_id >= 0:
+                dest_token_index = tl.atomic_add(expert_start_loc + expert_id, 1)
+                tl.store(
+                    output_index + token_id * output_index_stride0 + topk_index,
+                    dest_token_index,
+                )
+                output_tensor_ptr = (
+                    output_tensor + dest_token_index * output_tensor_stride0
+                )
+                output_tensor_scale_ptr = (
+                    output_tensor_scale + dest_token_index * output_tensor_scale_stride0
+                )
+                tl.store(output_tensor_ptr + offset_in, to_copy, mask=mask)
+                tl.store(output_tensor_scale_ptr + offset_in_s, to_copy_s, mask=mask_s)
+
+
+# copy from https://github.com/ModelTC/lightllm/blob/main/lightllm/common/fused_moe/deepep_scatter_gather.py
+@torch.no_grad()
+def ep_scatter(
+    recv_x: torch.Tensor,
+    recv_x_scale: torch.Tensor,
+    recv_topk: torch.Tensor,
+    num_recv_tokens_per_expert: torch.Tensor,
+    expert_start_loc: torch.Tensor,
+    output_tensor: torch.Tensor,
+    output_tensor_scale: torch.Tensor,
+    m_indices: torch.Tensor,
+    output_index: torch.Tensor,
+):
+    BLOCK_E = 128  # token num of per expert is aligned to 128
+    BLOCK_D = 128  # block size of quantization
+    num_warps = 8
+    num_experts = num_recv_tokens_per_expert.shape[0]
+    hidden_size = recv_x.shape[1]
+    # grid = (triton.cdiv(hidden_size, BLOCK_D), num_experts)
+    grid = num_experts
+
+    assert m_indices.shape[0] % BLOCK_E == 0
+
+    _fwd_kernel_ep_scatter_1[(grid,)](
+        num_recv_tokens_per_expert,
+        expert_start_loc,
+        m_indices,
+        num_experts=num_experts,
+        num_warps=num_warps,
+        BLOCK_E=BLOCK_E,
+        BLOCK_EXPERT_NUM=triton.next_power_of_2(num_experts),
+    )
+
+    grid = min(recv_topk.shape[0], 1024 * 8)
+
+    _fwd_kernel_ep_scatter_2[(grid,)](
+        recv_topk.shape[0],
+        expert_start_loc,
+        recv_x,
+        recv_x.stride(0),
+        recv_x.stride(1),
+        recv_x_scale,
+        recv_x_scale.stride(0),
+        recv_x_scale.stride(1),
+        recv_topk,
+        recv_topk.stride(0),
+        recv_topk.stride(1),
+        output_tensor,
+        output_tensor.stride(0),
+        output_tensor.stride(1),
+        output_tensor_scale,
+        output_tensor_scale.stride(0),
+        output_tensor_scale.stride(1),
+        output_index,
+        output_index.stride(0),
+        output_index.stride(1),
+        topk_num=recv_topk.shape[1],
+        num_warps=num_warps,
+        HIDDEN_SIZE=hidden_size,
+        HIDDEN_SIZE_PAD=triton.next_power_of_2(hidden_size),
+        SCALE_HIDDEN_SIZE=hidden_size // BLOCK_D,
+        SCALE_HIDDEN_SIZE_PAD=triton.next_power_of_2(hidden_size // BLOCK_D),
+    )
+    return
+
+
+@triton.jit
+def _fwd_kernel_ep_gather(
+    total_token_num,
+    input_tensor,
+    input_tensor_stride0,
+    input_tensor_stride1,
+    recv_topk_ids,
+    recv_topk_ids_stride0,
+    recv_topk_ids_stride1,
+    recv_topk_weight,
+    recv_topk_weight_stride0,
+    recv_topk_weight_stride1,
+    input_index,
+    input_index_stride0,
+    input_index_stride1,
+    output_tensor,
+    output_tensor_stride0,
+    output_tensor_stride1,
+    topk_num: tl.constexpr,
+    BLOCK_D: tl.constexpr,
+):
+    cur_block = tl.program_id(0)
+    start_cur_token = tl.program_id(1)
+    grid_num = tl.num_programs(1)
+
+    for cur_token in range(start_cur_token, total_token_num, grid_num):
+        off_d = tl.arange(0, BLOCK_D)
+        accumulator = tl.zeros([BLOCK_D], dtype=tl.float32)
+        for topk_index in range(0, topk_num):
+            expert_id = tl.load(
+                recv_topk_ids + cur_token * recv_topk_ids_stride0 + topk_index
+            )
+            if expert_id >= 0:
+                source_token_index = tl.load(
+                    input_index + cur_token * input_index_stride0 + topk_index
+                )
+                acc_weight = tl.load(
+                    recv_topk_weight + cur_token * recv_topk_weight_stride0 + topk_index
+                )
+                tmp = tl.load(
+                    input_tensor
+                    + source_token_index * input_tensor_stride0
+                    + cur_block * BLOCK_D
+                    + off_d
+                )
+                accumulator += tmp.to(tl.float32) * acc_weight
+
+        tl.store(
+            output_tensor
+            + cur_token * output_tensor_stride0
+            + cur_block * BLOCK_D
+            + off_d,
+            accumulator.to(output_tensor.dtype.element_ty),
+        )
+
+
+@torch.no_grad()
+def ep_gather(
+    input_tensor: torch.Tensor,
+    recv_topk_ids: torch.Tensor,
+    recv_topk_weight: torch.Tensor,
+    input_index: torch.Tensor,
+    output_tensor: torch.Tensor,
+):
+    BLOCK_D = 1024  # block size of quantization
+    num_warps = 2
+    num_tokens = output_tensor.shape[0]
+    hidden_size = input_tensor.shape[1]
+    assert hidden_size % BLOCK_D == 0
+    grid = (triton.cdiv(hidden_size, BLOCK_D), min(num_tokens, 1024))
+    _fwd_kernel_ep_gather[grid](
+        num_tokens,
+        input_tensor,
+        input_tensor.stride(0),
+        input_tensor.stride(1),
+        recv_topk_ids,
+        recv_topk_ids.stride(0),
+        recv_topk_ids.stride(1),
+        recv_topk_weight,
+        recv_topk_weight.stride(0),
+        recv_topk_weight.stride(1),
+        input_index,
+        input_index.stride(0),
+        input_index.stride(1),
+        output_tensor,
+        output_tensor.stride(0),
+        output_tensor.stride(1),
+        topk_num=recv_topk_ids.shape[1],
+        num_warps=num_warps,
+        BLOCK_D=BLOCK_D,
+    )
+    return
+
+
+# copy from
+# https://github.com/deepseek-ai/DeepGEMM/blob/bd2a77552886b98c205af12f8d7d2d61247c4b27/deep_gemm/jit_kernels/utils.py#L58
+def get_tma_aligned_size(x: int, element_size: int) -> int:
+    """
+    Global memory address of TMA must be 16-byte aligned.
+    Since we use column-major layout for the LHS scaling tensor,
+        the M-axis of the LHS scaling tensor needs to be padded to a multiple of 16 bytes.
+
+    Arguments:
+        x: original M-axis shape of the LHS scaling tensor.
+        element_size: element size of the LHS scaling tensor.
+
+    Returns:
+        M-axis shape of the LHS scaling tensor after padding.
+    """
+    tma_alignment_bytes = 16
+    assert tma_alignment_bytes % element_size == 0
+    alignment = tma_alignment_bytes // element_size
+    return ceil_div(x, alignment) * alignment
+
+
+@triton.jit
+def _tma_align_input_scale_kernel(
+    input_scale_ptr,
+    output_ptr,
+    m,
+    k_div_block_size,
+    input_scale_stride_m,
+    input_scale_stride_k,
+    output_stride_m,
+    output_stride_k,
+    BLOCK_SIZE_K: tl.constexpr,
+):
+    pid_m = tl.program_id(axis=0)
+    grid_m = tl.num_programs(0)
+    k_offsets = tl.arange(0, BLOCK_SIZE_K)
+
+    for m_base in range(pid_m, m, grid_m):
+        input_offset = (
+            input_scale_ptr
+            + m_base * input_scale_stride_m
+            + k_offsets * input_scale_stride_k
+        )
+        input_data = tl.load(input_offset, mask=k_offsets < k_div_block_size)
+
+        output_offset = (
+            output_ptr + k_offsets * output_stride_k + m_base * output_stride_m
+        )
+        tl.store(output_offset, input_data, mask=k_offsets < k_div_block_size)
+
+
+# copy from https://github.com/ModelTC/lightllm/blob/main/lightllm/common/quantization/triton_quant/fp8/fp8act_quant_kernel.py
+def tma_align_input_scale(input_scale: torch.Tensor):
+    assert input_scale.dim() == 2
+    m, k_div_block_size = input_scale.shape
+    padd_m = get_tma_aligned_size(m, input_scale.element_size())
+    output = torch.empty(
+        (k_div_block_size, padd_m), dtype=input_scale.dtype, device=input_scale.device
+    )
+
+    grid_m = min(m, 8192)
+    BLOCK_SIZE_K = triton.next_power_of_2(k_div_block_size)
+
+    _tma_align_input_scale_kernel[(grid_m,)](
+        input_scale_ptr=input_scale,
+        output_ptr=output,
+        m=m,
+        k_div_block_size=k_div_block_size,
+        input_scale_stride_m=input_scale.stride(0),
+        input_scale_stride_k=input_scale.stride(1),
+        output_stride_m=output.stride(1),  # Note: these are swapped
+        output_stride_k=output.stride(0),  # for column-major
+        BLOCK_SIZE_K=BLOCK_SIZE_K,
+    )
+    return output.t()[:m]

sglang/srt/layers/moe/ep_moe/layer.py

@@ -4,11 +4,19 @@ from typing import Callable, List, Optional, Tuple
 import torch
 from torch.nn import Module
 
+from sglang.srt.layers.quantization.deep_gemm import _ENABLE_JIT_DEEPGEMM
+
 try:
     from deep_gemm import (
         get_col_major_tma_aligned_tensor,
+        m_grouped_gemm_fp8_fp8_bf16_nt_contiguous,
         m_grouped_gemm_fp8_fp8_bf16_nt_masked,
     )
+    from sgl_kernel import silu_and_mul
+
+    from sglang.srt.layers.quantization.fp8_kernel import (
+        sglang_per_token_group_quant_fp8,
+    )
 
     use_deep_gemm = True
 except ImportError:
@@ -20,6 +28,8 @@ from sglang.srt.distributed import (
     get_tensor_model_parallel_world_size,
 )
 from sglang.srt.layers.moe.ep_moe.kernels import (
+    ep_gather,
+    ep_scatter,
     gelu_and_mul_triton_kernel,
     grouped_gemm_triton,
     post_reorder_triton_kernel,
@@ -27,6 +37,7 @@ from sglang.srt.layers.moe.ep_moe.kernels import (
     run_moe_ep_preproess,
     silu_and_mul_masked_post_quant_fwd,
     silu_and_mul_triton_kernel,
+    tma_align_input_scale,
 )
 from sglang.srt.layers.moe.fused_moe_triton import FusedMoeWeightScaleSupported
 from sglang.srt.layers.moe.fused_moe_triton.layer import FusedMoEMethodBase
@@ -842,15 +853,23 @@ class DeepEPMoE(EPMoE):
     def forward(
         self,
         hidden_states: torch.Tensor,
+        topk_idx: torch.Tensor,
+        topk_weights: torch.Tensor,
         reorder_topk_ids: torch.Tensor,
         seg_indptr: torch.Tensor,
         masked_m: torch.Tensor,
         expected_m: int,
+        num_recv_tokens_per_expert: List[int],
         forward_mode: ForwardMode,
     ):
         resolved_deepep_mode = self.deepep_mode.resolve(forward_mode)
         if resolved_deepep_mode == DeepEPMode.normal:
-            return self.forward_normal(hidden_states, reorder_topk_ids, seg_indptr)
+            if _ENABLE_JIT_DEEPGEMM:
+                return self.forward_deepgemm_contiguous(
+                    hidden_states, topk_idx, topk_weights, num_recv_tokens_per_expert
+                )
+            else:
+                return self.forward_normal(hidden_states, reorder_topk_ids, seg_indptr)
         elif resolved_deepep_mode == DeepEPMode.low_latency:
             return self.forward_deepgemm_masked(hidden_states, masked_m, expected_m)
         else:
@@ -969,6 +988,106 @@ class DeepEPMoE(EPMoE):
             )
         return down_output
 
+    def forward_deepgemm_contiguous(
+        self,
+        hidden_states_fp8: Tuple[torch.Tensor, torch.Tensor],
+        topk_idx,
+        topk_weights,
+        num_recv_tokens_per_expert: List[int],
+    ):
+        hidden_states_fp8, hidden_states_scale = hidden_states_fp8
+        assert self.quant_method is not None
+        assert self.activation == "silu"
+        if num_recv_tokens_per_expert is None:
+            return hidden_states_fp8.bfloat16()
+        all_tokens = sum(num_recv_tokens_per_expert)
+        if all_tokens <= 0:
+            return hidden_states_fp8.bfloat16()
+        M, K = hidden_states_fp8.size()
+        N = self.w13_weight.size(1)
+        scale_block_size = 128
+
+        gather_out = torch.empty_like(
+            hidden_states_fp8,
+            device=hidden_states_fp8.device,
+            dtype=torch.bfloat16,
+        )
+
+        input_tensor = [
+            torch.empty(
+                (all_tokens, K),
+                device=hidden_states_fp8.device,
+                dtype=hidden_states_fp8.dtype,
+            ),
+            torch.empty(
+                (all_tokens, K // 128),
+                device=hidden_states_fp8.device,
+                dtype=torch.float32,
+            ),
+        ]
+        m_indices = torch.empty(
+            all_tokens, device=hidden_states_fp8.device, dtype=torch.int32
+        )
+        output_index = torch.empty_like(topk_idx)
+
+        num_recv_tokens_per_expert_gpu = torch.tensor(
+            num_recv_tokens_per_expert,
+            dtype=torch.int32,
+            pin_memory=True,
+            device="cpu",
+        ).cuda(non_blocking=True)
+        expert_start_loc = torch.empty_like(num_recv_tokens_per_expert_gpu)
+
+        ep_scatter(
+            hidden_states_fp8,
+            hidden_states_scale,
+            topk_idx,
+            num_recv_tokens_per_expert_gpu,
+            expert_start_loc,
+            input_tensor[0],
+            input_tensor[1],
+            m_indices,
+            output_index,
+        )
+
+        gateup_output = torch.empty(
+            (all_tokens, N),
+            device=hidden_states_fp8.device,
+            dtype=torch.bfloat16,
+        )
+        input_tensor[1] = tma_align_input_scale(input_tensor[1])
+        m_grouped_gemm_fp8_fp8_bf16_nt_contiguous(
+            input_tensor, self.w13_weight_fp8, gateup_output, m_indices
+        )
+        down_input = torch.empty(
+            (
+                all_tokens,
+                N // 2,
+            ),
+            device=gateup_output.device,
+            dtype=torch.bfloat16,
+        )
+        silu_and_mul(gateup_output.view(-1, N), down_input)
+        down_output = torch.empty(
+            (all_tokens, K),
+            device=hidden_states_fp8.device,
+            dtype=torch.bfloat16,
+        )
+        down_input_fp8, down_input_scale = sglang_per_token_group_quant_fp8(
+            down_input, scale_block_size
+        )
+        down_input_scale = tma_align_input_scale(down_input_scale)
+        m_grouped_gemm_fp8_fp8_bf16_nt_contiguous(
+            (down_input_fp8, down_input_scale),
+            self.w2_weight_fp8,
+            down_output,
+            m_indices,
+        )
+
+        ep_gather(down_output, topk_idx, topk_weights, output_index, gather_out)
+
+        return gather_out
+
     def forward_deepgemm_masked(
         self,
         hidden_states_fp8: Tuple[torch.Tensor, torch.Tensor],