Source: src/olm/nn/attention/__init__.py:1
Classes
AttentionBase(embed_dim: int, num_heads: int, dropout: float = 0.0, bias: bool = True)
Bases: Module, ABC
Source: src/olm/nn/attention/base.py:8
Abstract base class for attention mechanisms.
Provides the common structure for attention layers, including QKV projections
and output projection. Subclasses must implement the specific attention logic
in compute_attention.
Attributes
embed_dim(int): Total dimension of the model.num_heads(int): Number of parallel attention heads.head_dim(int): Dimension of each attention head.scale(float): Scaling factor for dot products (1 / sqrt(head_dim)).dropout(nn.Dropout): Dropout layer applied to attention weights.q_proj(Linear): Linear projection for Query.k_proj(Linear): Linear projection for Key.v_proj(Linear): Linear projection for Value.out_proj(Linear): Linear projection for Output.
Methods
compute_attention(self, q: torch.Tensor, k: torch.Tensor, v: torch.Tensor, mask: torch.Tensor | None = None) -> torch.Tensor
Source: src/olm/nn/attention/base.py:57
Computes the attention scores and output.
Parameters
q(torch.Tensor): Query tensor [batch, heads, seq, head_dim].k(torch.Tensor): Key tensor [batch, heads, seq, head_dim].v(torch.Tensor): Value tensor [batch, heads, seq, head_dim].mask(torch.Tensor, optional): Attention mask. Defaults to None.
Returns
torch.Tensor: The attention output [batch, heads, seq, head_dim].
forward(self, x: torch.Tensor, mask: torch.Tensor | None = None) -> torch.Tensor
Source: src/olm/nn/attention/base.py:73
Standard forward pass for attention layers.
Projects input to Q, K, V, calls compute_attention, and projects output.
Parameters
x(torch.Tensor): Input tensor [batch, seq, embed_dim].mask(torch.Tensor, optional): Attention mask. Defaults to None.
Returns
torch.Tensor: Output tensor [batch, seq, embed_dim].
AttentionwithRoPEBase(embed_dim: int, num_heads: int, max_seq_len: int, dropout: float = 0.0, bias: bool = True, rope_theta: float = 10000.0)
Bases: Module, ABC
Source: src/olm/nn/attention/base.py:95
Abstract base class for attention mechanisms with Rotary Positional Embedding.
Provides the common structure for attention layers, including QKV projections
and output projection. Subclasses must implement the specific attention logic
in compute_attention.
Attributes
embed_dim(int): Total dimension of the model.num_heads(int): Number of parallel attention heads.head_dim(int): Dimension of each attention head.scale(float): Scaling factor for dot products (1 / sqrt(head_dim)).dropout(nn.Dropout): Dropout layer applied to attention weights.q_proj(Linear): Linear projection for Query.k_proj(Linear): Linear projection for Key.v_proj(Linear): Linear projection for Value.out_proj(Linear): Linear projection for Output.
Methods
compute_attention(self, q: torch.Tensor, k: torch.Tensor, v: torch.Tensor, mask: torch.Tensor | None = None) -> torch.Tensor
Source: src/olm/nn/attention/base.py:148
Computes the attention scores and output.
Parameters
q(torch.Tensor): Query tensor [batch, heads, seq, head_dim].k(torch.Tensor): Key tensor [batch, heads, seq, head_dim].v(torch.Tensor): Value tensor [batch, heads, seq, head_dim].mask(torch.Tensor, optional): Attention mask. Defaults to None.
Returns
torch.Tensor: The attention output [batch, heads, seq, head_dim].
forward(self, x: torch.Tensor, mask: torch.Tensor | None = None) -> torch.Tensor
Source: src/olm/nn/attention/base.py:164
Standard forward pass for attention layers.
Projects input to Q, K, V, calls compute_attention, and projects output.
Parameters
x(torch.Tensor): Input tensor [batch, seq, embed_dim].mask(torch.Tensor, optional): Attention mask. Defaults to None.
Returns
torch.Tensor: Output tensor [batch, seq, embed_dim].
FlashAttention(embed_dim: int, num_heads: int, dropout: float = 0.0, causal: bool = False, use_flash_attn: bool | None = None)
Bases: olm.nn.attention.base.AttentionBase
Source: src/olm/nn/attention/flash.py:12
Flash Attention implementation for efficient attention computation.
Uses PyTorch's native scaled_dot_product_attention (which includes Flash Attention 2 optimizations) when available, or falls back to a memory-efficient implementation.
Flash Attention provides:
- O(N) memory complexity instead of O(N²) for sequence length N
- Faster computation through kernel fusion and tiling
- Exact attention (not an approximation)
- Support for causal masking without materializing the full attention matrix
Reference: "FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness" (Dao et al., 2022) and "FlashAttention-2: Faster Attention with Better Parallelism and Work Partitioning" (Dao, 2023)
Parameters
embed_dim: Total dimension of the modelnum_heads: Number of parallel attention headsdropout: Dropout probability on attention weights (default: 0.0)causal: If True, applies causal masking for autoregressive models (default: False)use_flash_attn: Force enable/disable flash attention. If None, auto-detect (default: None)
Example
attn = FlashAttention(embed_dim=512, num_heads=8, causal=True)
x = torch.randn(2, 128, 512) # (batch, seq_len, embed_dim)
output = attn(x)
output.shape
torch.Size([2, 128, 512])
Methods
compute_attention(self, q: torch.Tensor, k: torch.Tensor, v: torch.Tensor, mask: torch.Tensor | None = None) -> torch.Tensor
Source: src/olm/nn/attention/flash.py:73
Computes attention using Flash Attention when available.
Parameters
q: Query tensor [batch, heads, seq, head_dim]k: Key tensor [batch, heads, seq, head_dim]v: Value tensor [batch, heads, seq, head_dim]mask: Optional attention mask [batch, heads, seq, seq] or [batch, 1, seq, seq]
Returns
Attention output [batch, heads, seq, head_dim]
extra_repr(self) -> str
Source: src/olm/nn/attention/flash.py:206
String representation of the module.
forward(self, x: torch.Tensor, mask: torch.Tensor | None = None) -> torch.Tensor
Source: src/olm/nn/attention/flash.py:177
Forward pass with Flash Attention.
Parameters
x: Input tensor [batch, seq_len, embed_dim]mask: Optional attention mask
Returns
Output tensor [batch, seq_len, embed_dim]
FlashAttentionwithRoPE(embed_dim: int, num_heads: int, max_seq_len: int, dropout: float = 0.0, causal: bool = False, bias: bool = True, rope_theta: float = 10000.0, use_flash_attn: bool | None = None)
Bases: olm.nn.attention.base.AttentionwithRoPEBase
Source: src/olm/nn/attention/flash.py:215
Flash Attention implementation for efficient attention computation.
Uses PyTorch's native scaled_dot_product_attention (which includes Flash Attention 2 optimizations) when available, or falls back to a memory-efficient implementation.
Flash Attention provides:
- O(N) memory complexity instead of O(N²) for sequence length N
- Faster computation through kernel fusion and tiling
- Exact attention (not an approximation)
- Support for causal masking without materializing the full attention matrix
Reference: "FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness" (Dao et al., 2022) and "FlashAttention-2: Faster Attention with Better Parallelism and Work Partitioning" (Dao, 2023)
Parameters
embed_dim: Total dimension of the modelnum_heads: Number of parallel attention headsdropout: Dropout probability on attention weights (default: 0.0)causal: If True, applies causal masking for autoregressive models (default: False)use_flash_attn: Force enable/disable flash attention. If None, auto-detect (default: None)
Example
attn = FlashAttention(embed_dim=512, num_heads=8, causal=True)
x = torch.randn(2, 128, 512) # (batch, seq_len, embed_dim)
output = attn(x)
output.shape
torch.Size([2, 128, 512])
Methods
compute_attention(self, q: torch.Tensor, k: torch.Tensor, v: torch.Tensor, mask: torch.Tensor | None = None) -> torch.Tensor
Source: src/olm/nn/attention/flash.py:286
Computes attention using Flash Attention when available.
Parameters
q: Query tensor [batch, heads, seq, head_dim]k: Key tensor [batch, heads, seq, head_dim]v: Value tensor [batch, heads, seq, head_dim]mask: Optional attention mask [batch, heads, seq, seq] or [batch, 1, seq, seq]
Returns
Attention output [batch, heads, seq, head_dim]
extra_repr(self) -> str
Source: src/olm/nn/attention/flash.py:428
String representation of the module.
forward(self, x: torch.Tensor, mask: torch.Tensor | None = None) -> torch.Tensor
Source: src/olm/nn/attention/flash.py:390
Forward pass with Flash Attention and RoPE.
Parameters
x: Input tensor [batch, seq_len, embed_dim]mask: Optional attention mask
Returns
Output tensor [batch, seq_len, embed_dim]
GroupedQueryAttention(embed_dim: int, num_heads: int, num_kv_heads: int, max_seq_len: int, head_dim: int | None = None, dropout: float = 0.0, rope_theta: float = 10000.0, use_bias: bool = False, qkv_bias: bool = False, use_qk_norm: bool = False, rms_norm_eps: float = 1e-06, attention_scale: float | None = None, attn_logit_softcap: float | None = None)
Bases: Module
Source: src/olm/nn/attention/gqa.py:11
Grouped Query Attention (GQA) with Rotary Positional Embeddings.
GQA is a distinct attention mechanism where the number of Key/Value heads is smaller than the number of Query heads. This reduces memory bandwidth usage during inference (smaller KV cache) while maintaining performance close to Multi-Head Attention (MHA).
If num_kv_heads == num_heads, this is equivalent to MHA. If num_kv_heads == 1, this is equivalent to Multi-Query Attention (MQA).
Parameters
embed_dim(int): Total dimension of the model.num_heads(int): Number of Query heads.num_kv_heads(int): Number of Key/Value heads. Must divide num_heads.max_seq_len(int): Maximum sequence length for RoPE.dropout(float, optional): Dropout probability. Defaults to 0.0.rope_theta(float, optional): Base frequency for RoPE. Defaults to 10000.0.use_bias(bool, optional): Whether to use bias in linear projections. Defaults to False.
Methods
forward(self, x: torch.Tensor, mask: torch.Tensor | None = None) -> torch.Tensor
Source: src/olm/nn/attention/gqa.py:104
Forward pass of Grouped Query Attention.
Parameters
x(torch.Tensor): Input tensor of shape [batch, seq_len, embed_dim].mask(torch.Tensor, optional): Attention mask of shape [batch, 1, seq_len, seq_len] or [batch, seq_len, seq_len]. Defaults to None.
Returns
torch.Tensor: Output tensor of shape [batch, seq_len, embed_dim].
MultiHeadAttention(embed_dims: int, num_heads: int, dropout: float = 0.0, causal: bool = False)
Bases: olm.nn.attention.base.AttentionBase
Source: src/olm/nn/attention/mha.py:8
Implements Multi-Head Attention (MHA) as described in "Attention Is All You Need".
Splits the input into multiple heads, computes scaled dot-product attention for each, and concatenates the results. Supports causal masking for autoregressive models.
Parameters
embed_dims(int): Total dimension of the model.num_heads(int): Number of parallel attention heads.dropout(float, optional): Dropout probability on attention weights. Defaults to 0.0.causal(bool, optional): If True, applies a causal mask. Defaults to False.
Attributes
scale(float): Scaling factor (1 / sqrt(head_dim)).causal(bool): Whether to apply a causal mask.
Methods
forward(self, x: torch.Tensor, mask: torch.Tensor | None = None) -> torch.Tensor (inherited from AttentionBase)
Source: src/olm/nn/attention/base.py:73
Standard forward pass for attention layers.
Projects input to Q, K, V, calls compute_attention, and projects output.
Parameters
x(torch.Tensor): Input tensor [batch, seq, embed_dim].mask(torch.Tensor, optional): Attention mask. Defaults to None.
Returns
torch.Tensor: Output tensor [batch, seq, embed_dim].
compute_attention(self, q: torch.Tensor, k: torch.Tensor, v: torch.Tensor, mask: torch.Tensor | None = None) -> torch.Tensor
Source: src/olm/nn/attention/mha.py:29
Computes the scaled dot-product attention.
Parameters
q(torch.Tensor): Query tensor of shape [batch, heads, seq, head_dim].k(torch.Tensor): Key tensor of shape [batch, heads, seq, head_dim].v(torch.Tensor): Value tensor of shape [batch, heads, seq, head_dim].mask(torch.Tensor, optional): Attention mask. Defaults to None.
Returns
torch.Tensor: The result of the attention mechanism applied to v.
MultiHeadAttentionwithALiBi(embed_dim: int, num_heads: int, dropout: float = 0.0, bias: bool = False, causal: bool = True, max_seq_len: int = 2048)
Bases: olm.nn.attention.base.AttentionBase
Source: src/olm/nn/attention/alibi.py:9
Multi-Head Attention with ALiBi (Attention with Linear Biases).
ALiBi adds a static, non-learned bias to attention scores based on the distance between query and key positions. This allows the model to extrapolate to longer sequence lengths than seen during training.
Parameters
embed_dim(int): Total dimension of the model.num_heads(int): Number of parallel attention heads.dropout(float, optional): Dropout probability. Defaults to 0.0.bias(bool, optional): Whether to use bias in linear projections. Defaults to False.causal(bool, optional): Whether to apply causal masking logic. Defaults to True.max_seq_len(int, optional): Max sequence length for precomputing ALiBi bias. Defaults to 2048.
Methods
forward(self, x: torch.Tensor, mask: torch.Tensor | None = None) -> torch.Tensor (inherited from AttentionBase)
Source: src/olm/nn/attention/base.py:73
Standard forward pass for attention layers.
Projects input to Q, K, V, calls compute_attention, and projects output.
Parameters
x(torch.Tensor): Input tensor [batch, seq, embed_dim].mask(torch.Tensor, optional): Attention mask. Defaults to None.
Returns
torch.Tensor: Output tensor [batch, seq, embed_dim].
compute_attention(self, q: torch.Tensor, k: torch.Tensor, v: torch.Tensor, mask: torch.Tensor | None = None) -> torch.Tensor
Source: src/olm/nn/attention/alibi.py:41
Computes attention scores with ALiBi bias.
MultiHeadAttentionwithRoPE(embed_dims: int, num_heads: int, max_seq_len: int, dropout: float = 0.0, causal: bool = False, bias: bool = True, rope_theta: float = 10000.0)
Bases: olm.nn.attention.base.AttentionwithRoPEBase
Source: src/olm/nn/attention/mha.py:58
Implements Multi-Head Attention (MHA) with Rotary Positional Embedding (RoPE).
Splits the input into multiple heads, computes scaled dot-product attention for each, and concatenates the results. Uses RoPE for positional information.
Parameters
embed_dims(int): Total dimension of the model.num_heads(int): Number of parallel attention heads.max_seq_len(int): Maximum sequence length.dropout(float, optional): Dropout probability on attention weights. Defaults to 0.0.causal(bool, optional): If True, applies a causal mask. Defaults to False.
Attributes
scale(float): Scaling factor (1 / sqrt(head_dim)).causal(bool): Whether to apply a causal mask.
Methods
forward(self, x: torch.Tensor, mask: torch.Tensor | None = None) -> torch.Tensor (inherited from AttentionwithRoPEBase)
Source: src/olm/nn/attention/base.py:164
Standard forward pass for attention layers.
Projects input to Q, K, V, calls compute_attention, and projects output.
Parameters
x(torch.Tensor): Input tensor [batch, seq, embed_dim].mask(torch.Tensor, optional): Attention mask. Defaults to None.
Returns
torch.Tensor: Output tensor [batch, seq, embed_dim].
compute_attention(self, q: torch.Tensor, k: torch.Tensor, v: torch.Tensor, mask: torch.Tensor | None = None) -> torch.Tensor
Source: src/olm/nn/attention/mha.py:80
Computes the scaled dot-product attention suited for RoPE.
Parameters
q(torch.Tensor): Query tensor of shape [batch, heads, seq, head_dim].k(torch.Tensor): Key tensor of shape [batch, heads, seq, head_dim].v(torch.Tensor): Value tensor of shape [batch, heads, seq, head_dim].mask(torch.Tensor, optional): Attention mask. Defaults to None.
Returns
torch.Tensor: The result of the attention mechanism applied to v.