liger-kernel 0.6.0__py3-none-any.whl → 0.6.2__py3-none-any.whl

liger_kernel/chunked_loss/dpo_loss.py

@@ -13,6 +13,7 @@ class LigerFusedLinearDPOFunction(LigerFusedLinearPreferenceBase):
         ref_chosen_logps=None,
         ref_rejected_logps=None,
         beta=0.1,
+        loss_type="sigmoid",
     ):
         """
         Paper: https://arxiv.org/pdf/2305.18290
@@ -48,8 +49,50 @@ class LigerFusedLinearDPOFunction(LigerFusedLinearPreferenceBase):
         chosen_rewards = beta * chosen_logratios
         rejected_rewards = beta * rejected_logratios
 
-        logits_diff = beta * (chosen_logratios - rejected_logratios)
-        loss = -F.logsigmoid(logits_diff).sum() / (full_target.shape[0] // 2)
+        if loss_type == "sigmoid":
+            logits_diff = beta * (chosen_logratios - rejected_logratios)
+            loss = -F.logsigmoid(logits_diff).sum() / (full_target.shape[0] // 2)
+
+        elif loss_type == "apo_zero":
+            # Eqn (7) of the APO paper (https://huggingface.co/papers/2408.06266)
+            # Use this loss when you believe the chosen outputs are better than your model's default output
+            losses_chosen = 1 - F.sigmoid(beta * chosen_logratios)  # Increase chosen likelihood
+            losses_rejected = F.sigmoid(beta * rejected_logratios)
+            losses = losses_chosen + losses_rejected
+            loss = losses.sum() / (full_target.shape[0] // 2)
+
+        elif loss_type == "apo_down":
+            # Eqn (8) of the APO paper (https://huggingface.co/papers/2408.06266)
+            # Use this loss when you believe the chosen outputs are worse than your model's default output.
+            # Decrease chosen likelihood and decrease rejected likelihood more
+            losses_chosen = F.sigmoid(beta * chosen_logratios)
+            losses_rejected = 1 - F.sigmoid(beta * (chosen_logratios - rejected_logratios))
+            losses = losses_chosen + losses_rejected
+            loss = losses.sum() / (full_target.shape[0] // 2)
+
+        elif loss_type == "sppo_hard":
+            # In the paper (https://huggingface.co/papers/2405.00675), SPPO employs a soft probability approach,
+            # estimated using the PairRM score. The probability calculation is conducted outside of the trainer class.
+            # The version described here is the hard probability version, where P in Equation (4.7) of Algorithm 1 is
+            # set to 1 for the winner and 0 for the loser.
+            a = chosen_logps - ref_chosen_logps
+            b = rejected_logps - ref_rejected_logps
+            losses = (a - 0.5 / beta) ** 2 + (b + 0.5 / beta) ** 2
+            loss = losses.sum() / (full_target.shape[0] // 2)
+
+        elif loss_type == "nca_pair":
+            losses = (
+                -F.logsigmoid(chosen_rewards)
+                - 0.5 * F.logsigmoid(-chosen_rewards)
+                - 0.5 * F.logsigmoid(-rejected_rewards)
+            )
+            loss = losses.sum() / (full_target.shape[0] // 2)
+
+        else:
+            raise ValueError(
+                f"Unsupported loss_type: {loss_type}. Supported types are: sigmoid, apo_zero, apo_down, sppo_hard, nca_pair"
+            )
+
         return loss, chosen_rewards, rejected_rewards
 
     @classmethod
@@ -70,6 +113,7 @@ class LigerFusedLinearDPOFunction(LigerFusedLinearPreferenceBase):
         use_ref_model=True,
         average_log_prob=False,
         chunk_size=1,
+        loss_type="sigmoid",
     ):
         """
         Fused linear layer with DPO loss.
@@ -108,12 +152,13 @@ class LigerFusedLinearDPOFunction(LigerFusedLinearPreferenceBase):
             ref_bias=ref_bias,
             average_log_prob=average_log_prob,
             chunk_size=chunk_size,
+            loss_type=loss_type,
         )
 
     @staticmethod
     def backward(ctx, *grad_output):
         grads = LigerFusedLinearPreferenceBase.backward(ctx, grad_output)[:4]
-        return *grads, None, None, None, None, None, None, None, None, None, None
+        return *grads, None, None, None, None, None, None, None, None, None, None, None
 
 
 class LigerFusedLinearDPOLoss(torch.nn.Module):
@@ -130,6 +175,7 @@ class LigerFusedLinearDPOLoss(torch.nn.Module):
         use_ref_model: bool = True,
         average_log_prob: bool = False,
         chunk_size: int = 1,
+        loss_type: str = "sigmoid",
     ):
         """
         Args:
@@ -149,6 +195,10 @@ class LigerFusedLinearDPOLoss(torch.nn.Module):
         self.use_ref_model = use_ref_model
         self.average_log_prob = average_log_prob
         self.chunk_size = chunk_size
+        self.loss_type = loss_type
+        supported_loss_types = {"sigmoid", "apo_zero", "apo_down", "sppo_hard", "nca_pair"}
+        if self.loss_type not in supported_loss_types:
+            raise ValueError(f"Unsupported loss_type: {self.loss_type}. Supported types are: {supported_loss_types}")
 
     def forward(
         self,
@@ -175,4 +225,5 @@ class LigerFusedLinearDPOLoss(torch.nn.Module):
             self.use_ref_model,
             self.average_log_prob,
             self.chunk_size,
+            self.loss_type,
         )

liger_kernel/ops/fused_add_rms_norm.py

@@ -0,0 +1,412 @@
+import math
+import operator
+
+import torch
+import triton
+import triton.language as tl
+
+from liger_kernel.ops.utils import calculate_settings
+from liger_kernel.ops.utils import compare_version
+from liger_kernel.ops.utils import ensure_contiguous
+from liger_kernel.ops.utils import torch_to_triton_dtype
+
+if compare_version("triton", operator.ge, "3.0.0"):
+    try:
+        # typical import path with dispatch available
+        from triton.language.extra.libdevice import rsqrt
+    except ModuleNotFoundError:
+        # for working with NGC containers
+        from triton.language.extra.cuda.libdevice import rsqrt
+else:
+    from triton.language.math import rsqrt
+
+
+_CASTING_MODE_NONE: tl.constexpr = tl.constexpr(-1)
+_CASTING_MODE_LLAMA: tl.constexpr = tl.constexpr(0)
+_CASTING_MODE_GEMMA: tl.constexpr = tl.constexpr(1)
+
+
+@triton.jit
+def _fused_add_rms_norm_forward_kernel(
+    Y_ptr,
+    Y_row_stride,
+    S_ptr,  # output residual
+    S_row_stride,
+    X_ptr,
+    X_row_stride,
+    R_ptr,  # input residual
+    R_row_stride,
+    W_ptr,
+    W_row_stride,
+    RSTD_ptr,
+    RSTD_row_stride,
+    n_cols,
+    eps,
+    offset,
+    casting_mode: tl.constexpr,  # constexpr so the `if` blocks can be optimized out
+    BLOCK_SIZE: tl.constexpr,
+):
+    """
+    This kernel computes the following:
+    1. hidden_states = residual + hidden_states
+    2. residual = hidden_states
+    3. hidden_states = rmsnorm(hidden_states)
+
+    This is a commonly used pattern in the decoder layers of LLMs.
+    Some examples:
+    1. https://github.com/huggingface/transformers/blob/0dc2df5ddafe3cb5824ad24e85beba13e0aa6726/src/transformers/models/qwen3/modeling_qwen3.py#L271
+    2. https://github.com/huggingface/transformers/blob/0dc2df5ddafe3cb5824ad24e85beba13e0aa6726/src/transformers/models/llama4/modeling_llama4.py#L393
+
+    This kernel is inspired by the rms_norm forward kernel, and is adapted to support the residual addition in the forward pass.
+    The backward pass is also adapted to support the residual addition in the backward pass.
+    """
+
+    row_idx = tl.program_id(0).to(tl.int64)
+    col_offsets = tl.arange(0, BLOCK_SIZE)
+    mask = col_offsets < n_cols
+
+    Y_ptr += row_idx * Y_row_stride
+    S_ptr += row_idx * S_row_stride
+    X_ptr += row_idx * X_row_stride
+    R_ptr += row_idx * R_row_stride
+    RSTD_ptr += row_idx * RSTD_row_stride
+
+    X_row = tl.load(X_ptr + col_offsets, mask=mask, other=0)
+    R_row = tl.load(R_ptr + col_offsets, mask=mask, other=0)
+    S_row = X_row + R_row
+    tl.store(S_ptr + col_offsets, S_row, mask=mask)
+    S_row_dtype = S_row.dtype
+    W_row = tl.load(W_ptr + col_offsets, mask=mask, other=0)
+
+    # On Llama, only rstd is computed on fp32
+    if casting_mode == _CASTING_MODE_LLAMA:
+        S_row = S_row.to(tl.float32)
+
+    # Gemma computes everything on fp32, and then casts back the output to the original dtype
+    if casting_mode == _CASTING_MODE_GEMMA:
+        W_row = W_row.to(tl.float32)
+        S_row = S_row.to(tl.float32)
+
+    if casting_mode == _CASTING_MODE_NONE:
+        eps = eps.to(S_row_dtype)
+        offset = offset.to(S_row_dtype)
+
+    mean_square = tl.sum(S_row * S_row, axis=0) / n_cols
+    rstd = rsqrt(mean_square + eps)
+
+    # We can save time by caching rms with minimal memory overhead
+    # because rms is much smaller compared to X_row, as rms is for each row.
+    # However, on the computation side, it can save 4 operations (*, sum, /, sqrt).
+    tl.store(RSTD_ptr, rstd)
+
+    S_row = S_row * rstd
+
+    # On Llama, the multiplication with the weight is done on the original dtype
+    if casting_mode == _CASTING_MODE_LLAMA:
+        S_row = S_row.to(S_row_dtype)
+
+    Y_row = S_row * (offset + W_row)
+
+    if casting_mode == _CASTING_MODE_GEMMA:
+        Y_row = Y_row.to(S_row_dtype)
+
+    tl.store(Y_ptr + col_offsets, Y_row, mask=mask)
+
+
+@triton.jit
+def _fused_add_rms_norm_backward_kernel(
+    dY_ptr,
+    dY_row_stride,
+    dS_out_ptr,
+    dS_out_row_stride,
+    dX_ptr,
+    dX_row_stride,
+    X_ptr,
+    X_row_stride,
+    X_dtype: tl.constexpr,
+    W_ptr,
+    W_row_stride,
+    RSTD_ptr,
+    RSTD_row_stride,
+    dW_ptr,
+    dW_row_stride,
+    n_rows,
+    n_cols,
+    offset,
+    rows_per_program: tl.constexpr,
+    casting_mode: tl.constexpr,
+    BLOCK_SIZE: tl.constexpr,
+    has_dS_out: tl.constexpr,
+):
+    """
+    This kernel is adapted from the rms_norm backward kernel, and is adapted to support the residual
+    addition in the backward pass. For the following code pattern:
+    1. hidden_states = residual + hidden_states
+    2. residual = hidden_states
+    3. hidden_states = rmsnorm(hidden_states)
+
+    The gradient of hidden_states and residual comes out be exactly same. The value of this gradient is
+    the sum of the gradient of the hidden_states in step 3 and the gradient of the residual in step 2.
+
+    The backward pass computation logic is same as the rms_norm backward kernel, except that the gradient
+    of the hidden_states in step 3 and the gradient of the residual in step 2 are summed up.
+    """
+
+    row_block_id = tl.program_id(0).to(tl.int64)
+    row_start = row_block_id * rows_per_program
+    row_end = min((row_block_id + 1) * rows_per_program, n_rows)
+    col_offsets = tl.arange(0, BLOCK_SIZE)
+    mask = col_offsets < n_cols
+
+    dW_row = tl.zeros((BLOCK_SIZE,), dtype=tl.float32)
+
+    dY_ptr += row_start * dY_row_stride
+    dX_ptr += row_start * dX_row_stride
+    if has_dS_out:
+        dS_out_ptr += row_start * dS_out_row_stride
+
+    X_ptr += row_start * X_row_stride
+    RSTD_ptr += row_start
+
+    W_row = tl.load(W_ptr + col_offsets, mask=mask, other=0.0)
+    W_row = W_row + offset
+
+    for _ in range(row_start, row_end):
+        dY_row = tl.load(dY_ptr + col_offsets, mask=mask, other=0.0)
+        X_row = tl.load(X_ptr + col_offsets, mask=mask, other=0.0)
+
+        # Get cached rms
+        rstd_row = tl.load(RSTD_ptr)
+
+        X_row = X_row.to(tl.float32)
+
+        # Different bacward graphs for different casting modes
+        if casting_mode == _CASTING_MODE_LLAMA:
+            m = (dY_row * W_row).to(tl.float32)
+
+        elif casting_mode == _CASTING_MODE_GEMMA:
+            dY_row = dY_row.to(tl.float32)
+            m = dY_row * W_row
+        else:
+            m = dY_row * W_row
+
+        dX_row = rstd_row * m
+
+        if has_dS_out:
+            dS_out_row = tl.load(dS_out_ptr + col_offsets, mask=mask, other=0.0)
+            dX_row += (rstd_row) * (
+                -(1 / n_cols) * rstd_row * rstd_row * tl.sum(m * X_row, axis=0) * X_row
+            ) + dS_out_row
+            dS_out_ptr += dS_out_row_stride
+        else:
+            dX_row += (rstd_row) * (-(1 / n_cols) * rstd_row * rstd_row * tl.sum(m * X_row, axis=0) * X_row)
+
+        # calculate the gradient of W
+        if casting_mode == _CASTING_MODE_LLAMA:
+            dW_row += dY_row * (X_row * rstd_row).to(X_dtype)
+        else:
+            # here X_row is already in fp32 (see previous if block)
+            dW_row += dY_row * (X_row * rstd_row)
+
+        tl.store(dX_ptr + col_offsets, dX_row.to(X_dtype), mask=mask)
+
+        dY_ptr += dY_row_stride
+        dX_ptr += dX_row_stride
+        X_ptr += X_row_stride
+        RSTD_ptr += RSTD_row_stride
+
+    tl.store(dW_ptr + row_block_id * dW_row_stride + col_offsets, dW_row, mask=mask)
+
+
+_str_to_casting_mode = {
+    "llama": _CASTING_MODE_LLAMA.value,
+    "gemma": _CASTING_MODE_GEMMA.value,
+    "none": _CASTING_MODE_NONE.value,
+}
+
+
+def fused_add_rms_norm_forward(X, R, W, eps, offset, casting_mode):
+    if not isinstance(casting_mode, int):
+        assert casting_mode in _str_to_casting_mode, f"Invalid casting mode: {casting_mode}"
+        casting_mode = _str_to_casting_mode[casting_mode]
+    else:
+        assert casting_mode in _str_to_casting_mode.values(), f"Invalid casting mode: {casting_mode}"
+
+    shape = X.shape
+    dim = shape[-1]
+    X = X.view(-1, dim)
+    R = R.view(-1, dim)
+    n_rows, n_cols = X.shape
+    BLOCK_SIZE, num_warps = calculate_settings(n_cols)
+
+    Y = torch.empty((n_rows, n_cols), dtype=X.dtype, device=X.device)
+    S = torch.empty((n_rows, n_cols), dtype=X.dtype, device=X.device)
+    # RSTD is to cache rstd for each row
+    # RSTD is always computed/stored in fp32 if we are using Llama or Gemma casting mode
+    rstd_dtype = torch.float32 if casting_mode in (_CASTING_MODE_LLAMA.value, _CASTING_MODE_GEMMA.value) else X.dtype
+    RSTD = torch.empty(n_rows, dtype=rstd_dtype, device=X.device)
+
+    # Check constraints.
+    assert X.shape[1] == W.shape[0], "Incompatible hidden size dimension between tensor1.shape[1] and tensor2.shape[0]"
+
+    # XPU-specific optimization
+    kernel_args = {}
+    if X.device.type == "xpu":
+        kernel_args["grf_mode"] = "large"
+
+    # TODO: add _block_fused_add_rms_norm_forward_kernel
+    _fused_add_rms_norm_forward_kernel[(n_rows,)](
+        Y,
+        Y.stride(0),
+        S,
+        S.stride(0),
+        X,
+        X.stride(0),
+        R,
+        R.stride(0),
+        W,
+        W.stride(0),
+        RSTD,
+        RSTD.stride(0),
+        n_cols,
+        eps,
+        offset,
+        casting_mode,
+        BLOCK_SIZE=BLOCK_SIZE,
+        num_warps=num_warps,
+        **kernel_args,  # XPU-specific optimization
+    )
+
+    return Y.view(*shape), S.view(*shape), RSTD, BLOCK_SIZE, num_warps, casting_mode
+
+
+def fused_add_rms_norm_backward(dY, dS_out, S, W, RSTD, offset, casting_mode, BLOCK_SIZE, num_warps, in_place):
+    shape = dY.shape
+    dim = shape[-1]
+    dY = dY.view(-1, dim)
+    dS_out = dS_out.view(-1, dim)
+    S = S.view(-1, dim)
+    n_rows, n_cols = dY.shape
+
+    sm_count = 1
+    if S.device.type == "cuda":
+        sm_count = torch.cuda.get_device_properties(S.device).multi_processor_count
+    elif S.device.type == "xpu":
+        sm_count = torch.xpu.get_device_properties(S.device).gpu_eu_count
+
+    # fp32 for numerical stability especially.
+    _dW = torch.empty((sm_count, n_cols), dtype=torch.float32, device=W.device)
+
+    if n_cols > BLOCK_SIZE:
+        raise RuntimeError("This layer norm doesn't support feature dim >= 64KB.")
+    rows_per_program = math.ceil(n_rows / sm_count)
+    grid = (sm_count,)
+
+    if in_place is True:
+        dX = dY
+    else:
+        dX = torch.empty_like(dY)
+
+    # XPU-specific optimization
+    kernel_args = {}
+    if S.device.type == "xpu":
+        kernel_args["grf_mode"] = "large"
+
+    # TODO: add _block_fused_add_rms_norm_backward_kernel
+    _fused_add_rms_norm_backward_kernel[grid](
+        dY,
+        dY.stride(0),
+        dS_out,
+        dS_out.stride(0),
+        dX,
+        dX.stride(0),
+        S,
+        S.stride(0),
+        torch_to_triton_dtype[S.dtype],
+        W,
+        W.stride(0),
+        RSTD,
+        RSTD.stride(0),
+        _dW,
+        _dW.stride(0),
+        n_rows,
+        n_cols,
+        offset,
+        rows_per_program,
+        casting_mode,
+        BLOCK_SIZE=BLOCK_SIZE,
+        num_warps=num_warps,
+        has_dS_out=dS_out is not None,
+        **kernel_args,  # XPU-specific optimization
+    )
+
+    dX = dX.view(*shape)
+    dW = _dW.sum(dim=0).to(W.dtype)
+
+    return dX, dX, dW  # dR is equal to dX
+
+
+class LigerFusedAddRMSNormFunction(torch.autograd.Function):
+    """
+    Performs a fused operation that first adds a residual tensor to the hidden_states tensor (`X`), then applies RMSNorm (Root Mean Square Normalization) to the result using the weight tensor `W`, with optional offset and casting mode.
+
+    This class implements the following sequence, commonly used in transformer decoder layers:
+        1. hidden_states = residual + hidden_states
+        2. residual = hidden_states (after addition)
+        3. hidden_states = rmsnorm(hidden_states)
+
+    Both the normalized hidden_states and the updated residual are returned as outputs.
+
+    Some models use an 'offset' to shift the weight tensor `W` by a constant value. For example, Gemma
+    uses an offset of 1.0, so the computation becomes `(X / RMS(X)) * (W + 1.0)` instead of the usual
+    `(X / RMS(X)) * W`. You can pass the offset value as an argument to the forward function.
+
+    In addition, different models cast their inputs at different places during RMSNorm computation. For
+    example, Gemma casts everything to fp32 before starting the computation, while Llama casts only the
+    inverse RMS to fp32. You can specify the casting mode using the `casting_mode` argument. We currently
+    support the following casting modes (they match HuggingFace Transformers' implementations):
+    - 'llama': matches the Llama implementation, where only the inverse RMS is computed on fp32.
+    - 'gemma': matches the Gemma implementation, where everything is cast to fp32, then computed, then cast back to the original dtype.
+    - 'none': no casting is done. The computation is done in the original dtype. This saves memory and is slightly faster, but has more error w.r.t. the original implementation.
+
+    The `in_place` option determines whether to modify dY in-place to store dX. This defaults to `True` to save memory.
+    """
+
+    @staticmethod
+    @ensure_contiguous
+    def forward(ctx, X, R, W, eps, offset=0.0, casting_mode="llama", in_place=False):
+        """
+        X: (B, T, H) or (BxT, H)
+        W: (H,)
+        """
+        # TODO: add row_mode
+        Y, S, RSTD, BLOCK_SIZE, num_warps, casting_mode = fused_add_rms_norm_forward(X, R, W, eps, offset, casting_mode)
+        ctx.offset = offset
+        ctx.casting_mode = casting_mode
+        ctx.in_place = in_place
+        ctx.BLOCK_SIZE = BLOCK_SIZE
+        ctx.num_warps = num_warps
+        ctx.save_for_backward(S, W, RSTD)
+        return Y, S
+
+    @staticmethod
+    @ensure_contiguous
+    def backward(ctx, dY, dS_out):
+        """
+        Y: (B, T, H) or (BxT, H)
+        """
+        S, W, RSTD = ctx.saved_tensors
+        dX, dR, dW = fused_add_rms_norm_backward(
+            dY,
+            dS_out,
+            S,
+            W,
+            RSTD,
+            ctx.offset,
+            ctx.casting_mode,
+            ctx.BLOCK_SIZE,
+            ctx.num_warps,
+            ctx.in_place,
+        )
+
+        return dX, dR, dW, None, None, None, None, None

liger_kernel/ops/fused_linear_cross_entropy.py

@@ -25,6 +25,7 @@ def fused_linear_cross_entropy_forward(
     reduction="mean",
     softcap=None,
     return_z_loss=False,
+    accum_dtype=None,
 ):
     assert isinstance(return_z_loss, bool), f"return_z_loss must be True or False. Got: {return_z_loss}"
     device = _input.device
@@ -44,10 +45,16 @@ def fused_linear_cross_entropy_forward(
     chunk_size = triton.next_power_of_2(triton.cdiv(BT, inc_factor))  # (BT + inc_factor - 1) // inc_factor
     num_chunks = triton.cdiv(BT, chunk_size)  # (BT + chunk_size - 1) // chunk_size
 
-    grad_weight = torch.zeros_like(weight, device=device) if weight.requires_grad else None
     grad_input = torch.zeros_like(_input, device=device)
-    grad_bias = torch.zeros_like(bias, device=device) if bias is not None else None
-    # we use fp32 for loss accumulator
+
+    # we use fp32 for loss and gradients accumulator
+    if accum_dtype is None:
+        grad_weight = torch.zeros_like(weight, device=device) if weight.requires_grad else None
+        grad_bias = torch.zeros_like(bias, device=device) if bias is not None else None
+    else:
+        grad_weight = torch.zeros_like(weight, dtype=accum_dtype, device=device) if weight.requires_grad else None
+        grad_bias = torch.zeros_like(bias, dtype=accum_dtype, device=device) if bias is not None else None
+
     loss_1d = torch.zeros(BT, dtype=torch.float32, device=device)
     z_loss_1d = torch.zeros(BT, dtype=_input.dtype, device=_input.device) if return_z_loss else None
 
@@ -124,16 +131,7 @@ def fused_linear_cross_entropy_forward(
         grad_input[start_idx:end_idx] = grad_logits_chunk @ weight
 
         if grad_weight is not None:
-            torch.addmm(
-                input=grad_weight,
-                mat1=logits_chunk.t().to(
-                    _input_chunk.dtype
-                ),  # In an autocast scenario without bias, differing logits_chunk data types will cause an addmm operation error.
-                mat2=_input_chunk,
-                out=grad_weight,
-                alpha=1.0,
-                beta=1.0,
-            )
+            grad_weight += torch.mm(grad_logits_chunk.t(), _input_chunk).float()
 
         if bias is not None:
             torch.add(
@@ -151,6 +149,11 @@ def fused_linear_cross_entropy_forward(
     else:
         loss = torch.sum(loss_1d)
         z_loss = torch.sum(z_loss_1d) if return_z_loss else None
+
+    # Cast back to original dtype
+    grad_weight = grad_weight.to(weight.dtype) if grad_weight is not None else None
+    grad_bias = grad_bias.to(bias.dtype) if grad_bias is not None else None
+
     return loss, z_loss, grad_input, grad_weight, grad_bias
 
 
@@ -217,6 +220,7 @@ class LigerFusedLinearCrossEntropyFunction(torch.autograd.Function):
         reduction="mean",
         softcap=None,
         return_z_loss: bool = False,
+        accum_dtype=None,
     ):
         """
         Fusing the last linear layer with cross-entropy loss
@@ -235,6 +239,8 @@ class LigerFusedLinearCrossEntropyFunction(torch.autograd.Function):
         ignore_index: the index to ignore in the target
         label_smoothing (float): The amount of smoothing when computing the loss, where 0.0 means no smoothing.
         reduction: reduction to apply
+        accum_dtype (torch.dtype): the dtype of intermediate result buffers for weight and bias gradient accumulations.
+            Recommended to set `accum_dtype` to higher precision, e.g. `torch.float32`, if the training is unstable with original dtype. Default: `None`, performing accumulations in original dtype
         """
 
         loss, z_loss, grad_input, grad_weight, grad_bias = fused_linear_cross_entropy_forward(
@@ -249,6 +255,7 @@ class LigerFusedLinearCrossEntropyFunction(torch.autograd.Function):
             reduction=reduction,
             softcap=softcap,
             return_z_loss=return_z_loss,
+            accum_dtype=accum_dtype,
         )
         # downcast to dtype and store for backward
         ctx.save_for_backward(
@@ -280,4 +287,5 @@ class LigerFusedLinearCrossEntropyFunction(torch.autograd.Function):
             None,
             None,
             None,
+            None,
         )