liger-kernel-nightly 0.5.10.dev20250624183504__py3-none-any.whl → 0.6.4.dev20251121224847__py3-none-any.whl

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

@@ -25,10 +25,18 @@ def fused_linear_cross_entropy_forward(
     reduction="mean",
     softcap=None,
     return_z_loss=False,
+    accum_dtype=None,
+    use_token_scaling=False,
+    return_token_accuracy=False,
 ):
     assert isinstance(return_z_loss, bool), f"return_z_loss must be True or False. Got: {return_z_loss}"
+    assert isinstance(return_token_accuracy, bool), (
+        f"return_token_accuracy must be True or False. Got: {return_token_accuracy}"
+    )
     device = _input.device
 
+    input_requires_grad = _input.requires_grad
+
     # inputs have shape: BT x H
     # materialized activations will have shape: BT x V
     # the increase in memory = BT x V
@@ -44,12 +52,23 @@ 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 input_requires_grad:
+        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
+    else:
+        grad_weight = None
+        grad_bias = 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
+    token_accuracy_1d = torch.zeros(BT, dtype=torch.float32, device=device) if return_token_accuracy else None
 
     # TODO: evaluate how CUDA synchronization caused by .item() affects the speed
     target_mask = target != ignore_index
@@ -82,9 +101,40 @@ def fused_linear_cross_entropy_forward(
 
         n_rows = logits_chunk.shape[0]
 
+        # Compute predicted probabilities for token scaling if needed
+        if use_token_scaling:
+            # Compute softmax probabilities for scaling
+            # We need to compute this before the cross entropy kernel modifies logits_chunk
+            logits_for_softmax = logits_chunk.detach().clone()  # Detach to avoid gradient flow
+            if softcap is not None:
+                logits_for_softmax = softcap * torch.tanh(logits_for_softmax / softcap)
+
+            # Compute softmax to get predicted probabilities
+            probs = torch.softmax(logits_for_softmax, dim=-1)
+
+            # Get predicted probabilities for token scaling, handling ignored targets
+            valid_target_mask = target_chunk != ignore_index
+            valid_targets = target_chunk[valid_target_mask]
+
+            if len(valid_targets) > 0:
+                # Gather probabilities only for valid targets
+                valid_probs = probs[valid_target_mask]
+                pred_probs_valid = torch.gather(valid_probs, -1, valid_targets.unsqueeze(-1)).squeeze(-1)
+
+                # Create full tensor with zeros for ignored targets
+                pred_probs = torch.zeros_like(target_chunk, dtype=probs.dtype, device=probs.device)
+                pred_probs[valid_target_mask] = pred_probs_valid
+            else:
+                # All targets are ignored
+                pred_probs = torch.zeros_like(target_chunk, dtype=probs.dtype, device=probs.device)
+
+            # Store the scaling factors
+            scaling_factors = pred_probs.detach()  # Detach to ensure no gradient flow
+
         # unreduced loss
         loss_1d_slice = loss_1d[start_idx:end_idx]  # chunk_size,
         z_loss_1d_slice = z_loss_1d[start_idx:end_idx] if return_z_loss else None
+        token_accuracy_1d_slice = token_accuracy_1d[start_idx:end_idx] if return_token_accuracy else None
 
         # ensure _input and target are contiguous
         logits_chunk = logits_chunk.contiguous()
@@ -100,6 +150,10 @@ def fused_linear_cross_entropy_forward(
             loss_ptr=loss_1d_slice,
             z_loss_ptr=z_loss_1d_slice,
             loss_stride=loss_1d_slice.stride(-1),  # always 1
+            token_accuracy_ptr=token_accuracy_1d_slice,
+            token_accuracy_stride=token_accuracy_1d_slice.stride(-1)
+            if return_token_accuracy
+            else 0,  # always 1 if accuracy is enabled
             n_cols=V,
             n_non_ignore=total_n_non_ignore,
             sum_non_ignore_weight=total_sum_non_ignore_ce_weight,
@@ -110,35 +164,43 @@ def fused_linear_cross_entropy_forward(
             reduction=reduction,
             softcap=softcap,
             RETURN_Z_LOSS=return_z_loss,
+            RETURN_TOKEN_ACCURACY=return_token_accuracy,
             HAS_WEIGHT=True if ce_weight is not None else False,
             HAS_SOFTCAPPING=True if softcap is not None else False,
+            HAS_GRADIENTS=input_requires_grad,
             BLOCK_SIZE=BLOCK_SIZE,
             num_warps=32 if not is_hip() else 16,
         )
 
+        # Apply token scaling if requested
+        if use_token_scaling:
+            loss_1d_slice = loss_1d_slice * scaling_factors
+            if return_z_loss:
+                z_loss_1d_slice = z_loss_1d_slice * scaling_factors
+
         loss_1d[start_idx:end_idx] = loss_1d_slice
         if return_z_loss:
             z_loss_1d[start_idx:end_idx] = z_loss_1d_slice
+        if return_token_accuracy:
+            token_accuracy_1d[start_idx:end_idx] = token_accuracy_1d_slice
         grad_logits_chunk = logits_chunk  # chunk_size x V
 
-        grad_input[start_idx:end_idx] = grad_logits_chunk @ weight
+        # Apply token scaling to gradients if requested
+        if use_token_scaling:
+            # Expand scaling factors to match gradient dimensions
+            scaling_factors_expanded = scaling_factors.unsqueeze(-1)  # chunk_size x 1
+            grad_logits_chunk = grad_logits_chunk * scaling_factors_expanded
 
-        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,
-            )
+        if input_requires_grad:
+            grad_input[start_idx:end_idx] = grad_logits_chunk @ weight
 
-        if bias is not None:
+        if grad_weight is not None and input_requires_grad:
+            grad_weight += torch.mm(grad_logits_chunk.t(), _input_chunk).float()
+
+        if bias is not None and input_requires_grad:
             torch.add(
                 input=grad_bias,
-                other=logits_chunk.sum(dim=0),
+                other=grad_logits_chunk.sum(dim=0),
                 out=grad_bias,
                 alpha=1.0,
             )
@@ -148,10 +210,22 @@ def fused_linear_cross_entropy_forward(
     #     loss = loss_1d
     #     z_loss = z_loss_1d if return_z_loss else None
 
+    if reduction == "none":
+        # Return per-token losses
+        loss = loss_1d
+        z_loss = z_loss_1d if return_z_loss else None
+        token_accuracy = token_accuracy_1d if return_token_accuracy else None
     else:
         loss = torch.sum(loss_1d)
         z_loss = torch.sum(z_loss_1d) if return_z_loss else None
-    return loss, z_loss, grad_input, grad_weight, grad_bias
+        # For accuracy, we compute the mean across all non-ignored tokens
+        token_accuracy = torch.sum(token_accuracy_1d) / total_n_non_ignore if return_token_accuracy 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, token_accuracy, grad_input, grad_weight, grad_bias
 
 
 def fused_linear_cross_entropy_backward(grad_output, grad_input, grad_weight, grad_bias):
@@ -217,6 +291,9 @@ class LigerFusedLinearCrossEntropyFunction(torch.autograd.Function):
         reduction="mean",
         softcap=None,
         return_z_loss: bool = False,
+        accum_dtype=None,
+        use_token_scaling: bool = False,
+        return_token_accuracy: bool = False,
     ):
         """
         Fusing the last linear layer with cross-entropy loss
@@ -235,9 +312,15 @@ 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
+        use_token_scaling (bool): whether to scale each token's loss by its predicted probability (detached).
+            When True, each token's loss is multiplied by the model's predicted probability for that token's true class.
+            Default: False.
+        return_token_accuracy (bool): When `return_token_accuracy` is `True`, computes and returns per-token accuracy without materializing logits. Default: `False`
         """
 
-        loss, z_loss, grad_input, grad_weight, grad_bias = fused_linear_cross_entropy_forward(
+        loss, z_loss, token_accuracy, grad_input, grad_weight, grad_bias = fused_linear_cross_entropy_forward(
             _input=_input,
             weight=weight,
             target=target,
@@ -249,6 +332,9 @@ class LigerFusedLinearCrossEntropyFunction(torch.autograd.Function):
             reduction=reduction,
             softcap=softcap,
             return_z_loss=return_z_loss,
+            accum_dtype=accum_dtype,
+            use_token_scaling=use_token_scaling,
+            return_token_accuracy=return_token_accuracy,
         )
         # downcast to dtype and store for backward
         ctx.save_for_backward(
@@ -257,13 +343,16 @@ class LigerFusedLinearCrossEntropyFunction(torch.autograd.Function):
             grad_bias.detach() if bias is not None else None,
         )
         ctx.return_z_loss = return_z_loss
-        return loss, z_loss
+        ctx.return_token_accuracy = return_token_accuracy
+        return loss, z_loss, token_accuracy
 
     @staticmethod
     @amp_custom_bwd
-    def backward(ctx, grad_output, grad_output2):
+    def backward(ctx, grad_output, grad_output2, grad_output3):
         if ctx.return_z_loss:
             del grad_output2  # z_loss is only for logging
+        if ctx.return_token_accuracy:
+            del grad_output3  # token_accuracy is only for metrics
         (grad_input, grad_weight, grad_bias) = ctx.saved_tensors
         grad_input, grad_weight, grad_bias = fused_linear_cross_entropy_backward(
             grad_output, grad_input, grad_weight, grad_bias
@@ -280,4 +369,7 @@ class LigerFusedLinearCrossEntropyFunction(torch.autograd.Function):
             None,
             None,
             None,
+            None,
+            None,  # use_token_scaling
+            None,  # return_token_accuracy
         )

liger_kernel/ops/geglu.py

@@ -40,7 +40,7 @@ def _geglu_tanh_forward_kernel(a, b, c, stride, n_cols: tl.constexpr, BLOCK_SIZE
     tanh_arg = sqrt_2_over_pi * (a_row + 0.044715 * a_cubed)
     tanh_result = tanh(tanh_arg)
     geglu_a = 0.5 * a_row * (1 + tanh_result)
-    c_row = geglu_a * b_row
+    c_row = geglu_a.cast(b_row.dtype) * b_row
     tl.store(c + col_offsets, c_row, mask=mask)
 
 

liger_kernel/ops/grpo_loss.py

@@ -128,7 +128,9 @@ def _grpo_loss_fwd_kernel(
     per_token_loss1 = coef_1 * advantage
     per_token_loss2 = coef_2 * advantage
     per_token_loss = -tl.minimum(per_token_loss1, per_token_loss2)
-    is_clipped = per_token_loss1 < per_token_loss2
+    is_low_clipped = (coef_1 < 1 - EPS_LOW) & (advantage < 0)
+    is_high_clipped = (coef_1 > 1 + EPS_HIGH) & (advantage > 0)
+    is_clipped = is_low_clipped | is_high_clipped
 
     if BETA != 0.0:
         REF_LOGP += off_b * L + off_l