Spaces:
Running on Zero
Running on Zero
File size: 4,570 Bytes
454ecdd | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 | """Information Horizon Encoder - Causal transformer with linear attention."""
from __future__ import annotations
import torch
import torch.nn as nn
import torch.nn.functional as F
from typing import Optional, Dict, Any, List
from manifold.models.layers.attention import MultiHeadLinearAttention, RotaryPositionEncoding
class IHEBlock(nn.Module):
"""
Single IHE transformer block with linear attention + FFN.
Uses pre-norm architecture for training stability.
"""
def __init__(
self,
embed_dim: int = 256,
num_heads: int = 8,
ff_dim: int = 1024,
dropout: float = 0.1,
):
super().__init__()
self.norm1 = nn.LayerNorm(embed_dim)
self.norm2 = nn.LayerNorm(embed_dim)
self.attention = MultiHeadLinearAttention(
embed_dim=embed_dim,
num_heads=num_heads,
dropout=dropout,
causal=True,
use_rotary=True,
)
self.ffn = nn.Sequential(
nn.Linear(embed_dim, ff_dim),
nn.GELU(),
nn.Dropout(dropout),
nn.Linear(ff_dim, embed_dim),
nn.Dropout(dropout),
)
def forward(
self,
x: torch.Tensor,
mask: Optional[torch.Tensor] = None,
) -> Dict[str, torch.Tensor]:
"""
Forward pass through transformer block.
Args:
x: Input tensor [batch, seq, embed_dim]
mask: Optional attention mask [batch, seq]
Returns:
Dict with 'output' and 'attention_weights'
"""
normed = self.norm1(x)
attn_out = self.attention(normed, mask=mask)
x = x + attn_out["output"]
normed = self.norm2(x)
x = x + self.ffn(normed)
return {
"output": x,
"attention_weights": None,
}
class InformationHorizonEncoder(nn.Module):
"""
Multi-layer causal transformer for encoding player action sequences.
Uses linear attention O(T) and rotary position encoding.
Causal masking ensures actions can't see future information.
"""
def __init__(
self,
embed_dim: int = 256,
num_layers: int = 4,
num_heads: int = 8,
ff_dim: int = 1024,
dropout: float = 0.1,
max_seq_len: int = 128,
):
super().__init__()
self.embed_dim = embed_dim
self.num_layers = num_layers
self.num_heads = num_heads
self.max_seq_len = max_seq_len
head_dim = embed_dim // num_heads
self.pos_encoding = RotaryPositionEncoding(
dim=head_dim,
max_seq_len=max_seq_len,
)
self.layers = nn.ModuleList([
IHEBlock(
embed_dim=embed_dim,
num_heads=num_heads,
ff_dim=ff_dim,
dropout=dropout,
)
for _ in range(num_layers)
])
self.final_norm = nn.LayerNorm(embed_dim)
def forward(
self,
x: torch.Tensor,
mask: Optional[torch.Tensor] = None,
) -> Dict[str, torch.Tensor]:
"""
Encode action sequence through causal transformer layers.
Args:
x: Input tensor [batch, seq, embed_dim]
mask: Optional attention mask [batch, seq]
Returns:
Dict with 'encoding' and 'all_layer_outputs'
"""
all_layer_outputs: List[torch.Tensor] = []
for layer in self.layers:
layer_out = layer(x, mask=mask)
x = layer_out["output"]
all_layer_outputs.append(x)
encoding = self.final_norm(x)
return {
"encoding": encoding,
"all_layer_outputs": all_layer_outputs,
}
@classmethod
def from_config(cls, config: Any) -> "InformationHorizonEncoder":
"""
Create InformationHorizonEncoder from ModelConfig.
Args:
config: ModelConfig instance with IHE parameters
Returns:
Configured InformationHorizonEncoder instance
"""
return cls(
embed_dim=config.embed_dim,
num_layers=config.ihe_layers,
num_heads=config.ihe_heads,
ff_dim=config.ihe_ff_dim,
dropout=config.ihe_dropout,
max_seq_len=config.sequence_length,
)
|