File size: 3,592 Bytes
388fd6e | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 | """Utility functions for TileLang selective scan operations."""
import torch
import torch.nn.functional as F
from typing import Tuple, Optional
def validate_tensor_shapes(
u: torch.Tensor,
A: torch.Tensor,
B: torch.Tensor,
C: torch.Tensor,
) -> Tuple[int, int, int, int]:
"""
Validate tensor shapes and return dimensions.
Args:
u: Input (batch, seq_len, state_dim)
A: State matrix (hidden_dim, hidden_dim)
B: Input matrix (hidden_dim, state_dim)
C: Output matrix (state_dim, hidden_dim)
Returns:
(batch_size, seq_len, state_dim, hidden_dim)
Raises:
RuntimeError: If tensor shapes are incompatible
"""
if len(u.shape) != 3:
raise RuntimeError(f"Input u must be 3D, got {len(u.shape)}D")
batch_size, seq_len, state_dim = u.shape
hidden_dim = A.shape[0]
if A.shape != (hidden_dim, hidden_dim):
raise RuntimeError(f"A shape mismatch: expected ({hidden_dim}, {hidden_dim}), got {A.shape}")
if B.shape != (hidden_dim, state_dim):
raise RuntimeError(f"B shape mismatch: expected ({hidden_dim}, {state_dim}), got {B.shape}")
if C.shape != (state_dim, hidden_dim):
raise RuntimeError(f"C shape mismatch: expected ({state_dim}, {hidden_dim}), got {C.shape}")
return batch_size, seq_len, state_dim, hidden_dim
def convert_to_supported_dtype(tensor: torch.Tensor) -> Tuple[torch.Tensor, bool]:
"""
Convert tensor to supported dtype if needed.
TileLang may not support all dtypes, so convert bfloat16/float16 to float32
if needed, and track whether conversion was done.
Args:
tensor: Input tensor
Returns:
(converted_tensor, was_converted)
"""
if tensor.dtype in (torch.float32, torch.float64):
return tensor, False
elif tensor.dtype in (torch.float16, torch.bfloat16):
# Return both since we'll need to convert back
return tensor, False
else:
raise RuntimeError(f"Unsupported dtype: {tensor.dtype}")
def ensure_contiguous(*tensors: torch.Tensor) -> Tuple[torch.Tensor, ...]:
"""Ensure tensors are contiguous in memory."""
return tuple(t.contiguous() if not t.is_contiguous() else t for t in tensors)
def check_device_consistency(*tensors: torch.Tensor) -> torch.device:
"""
Verify all tensors are on the same device.
Returns:
The device of the tensors
Raises:
RuntimeError: If tensors are on different devices
"""
if not tensors:
raise RuntimeError("No tensors provided")
device = tensors[0].device
for t in tensors[1:]:
if t.device != device:
raise RuntimeError(f"Device mismatch: {device} vs {t.device}")
return device
def check_dtype_consistency(*tensors: torch.Tensor) -> torch.dtype:
"""
Verify all tensors have compatible dtypes.
Returns:
The dtype of the tensors
Raises:
RuntimeError: If tensors have incompatible dtypes
"""
if not tensors:
raise RuntimeError("No tensors provided")
dtype = tensors[0].dtype
for t in tensors[1:]:
if t.dtype != dtype:
# Allow compatible types, but warn
if t.dtype not in (torch.float32, torch.float16, torch.bfloat16, torch.float64):
raise RuntimeError(f"Incompatible dtype: {dtype} vs {t.dtype}")
return dtype
|