hyper-connections 0.0.12__tar.gz → 0.0.14__tar.gz

{hyper_connections-0.0.12 → hyper_connections-0.0.14}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: hyper-connections
-Version: 0.0.12
+Version: 0.0.14
 Summary: Hyper-Connections
 Project-URL: Homepage, https://pypi.org/project/hyper-connections/
 Project-URL: Repository, https://github.com/lucidrains/hyper-connections

hyper_connections-0.0.14/hyper_connections/hyper_connections_with_multi_branch_inputs.py

@@ -0,0 +1,178 @@
+from __future__ import annotations
+from typing import Callable
+
+from functools import partial
+from random import randrange
+
+import torch
+from torch import nn
+from torch.nn import Module
+import torch.nn.functional as F
+from torch.utils._pytree import tree_flatten, tree_unflatten
+
+from einops import rearrange, repeat, reduce, einsum
+
+from hyper_connections.hyper_connections import Residual, StreamEmbed
+
+# helper functions
+
+def exists(v):
+    return v is not None
+
+def default(v, d):
+    return v if exists(v) else d
+
+def identity(t):
+    return t
+
+# main classes
+
+# hyper connection residual streams
+
+class HyperConnections(Module):
+    def __init__(
+        self,
+        num_residual_streams,
+        *,
+        dim,
+        branch: Module | None = None,
+        layer_index = None,
+        tanh = True,
+        channel_first = False,
+        num_branch_inputs = 1  # residuals will be linearly combined to multiple inputs, fed through the branch, then linearly combined back out to residuals
+    ):
+        """
+        Appendix J, Algorithm2 in - https://arxiv.org/abs/2409.19606
+        """
+        super().__init__()
+
+        self.branch = branch
+
+        # activation, seemingly results were wishy washy depending on using tanh or not
+
+        self.act = nn.Tanh() if tanh else nn.Identity()
+
+        self.norm = nn.RMSNorm(dim) # they used layernorm in paper, but rmsnorm is fine given what we know now
+
+        self.num_residual_streams = num_residual_streams
+        self.num_branch_inputs = num_branch_inputs
+
+        init_residual_index = default(layer_index, randrange(num_residual_streams)) % num_residual_streams # just choose one random residual stream if layer index not given
+
+        self.static_beta = nn.Parameter(torch.ones(num_residual_streams, num_branch_inputs))
+
+        init_alpha0 = torch.zeros((num_residual_streams, num_branch_inputs))
+        init_alpha0[init_residual_index, :] = 1.
+
+        self.static_alpha = nn.Parameter(torch.cat([init_alpha0, torch.eye(num_residual_streams)], dim = 1))
+
+        self.dynamic_alpha_fn = nn.Parameter(torch.zeros(dim, num_residual_streams + num_branch_inputs))
+        self.dynamic_alpha_scale = nn.Parameter(torch.ones(()) * 1e-2)
+        self.dynamic_beta_fn = nn.Parameter(torch.zeros(dim, num_branch_inputs))
+        self.dynamic_beta_scale = nn.Parameter(torch.ones(()) * 1e-2)
+
+        # channel first option
+
+        self.channel_first = channel_first
+
+    @classmethod
+    def get_expand_reduce_stream_functions(cls, num_streams):
+        expand_fn = partial(repeat, pattern = 'b ... -> (b s) ...', s = num_streams)
+        reduce_fn = partial(reduce, pattern = '(b s) ... -> b ...', reduction = 'sum', s = num_streams)
+
+        return expand_fn, reduce_fn
+
+    @classmethod
+    def get_init_and_expand_reduce_stream_functions(cls, num_streams, disable = False):
+
+        hyper_conn_klass = cls if not disable else Residual
+
+        init_hyper_conn_fn = partial(hyper_conn_klass, num_streams)
+        expand_reduce_fns = cls.get_expand_reduce_stream_functions(num_streams) if not disable else (identity, identity)
+
+        return (init_hyper_conn_fn, *expand_reduce_fns)
+
+    def width_connection(self, residuals):
+        num_streams, num_branch_inputs = self.num_residual_streams, self.num_branch_inputs
+
+        # width connection
+
+        if self.channel_first:
+            residuals = rearrange(residuals, 'b d ... -> b ... d')
+
+        residuals = rearrange(residuals, '(b s) ... d -> b ... s d', s = num_streams)
+
+        normed = self.norm(residuals)
+
+        # alpha for weighted sum of residuals going into branch
+
+        wc_weight = self.act(normed @ self.dynamic_alpha_fn)
+        dynamic_alpha = wc_weight * self.dynamic_alpha_scale
+        alpha = dynamic_alpha + self.static_alpha
+
+        # beta for weights from branch output back to residual streams
+
+        dc_weight = self.act(normed @ self.dynamic_beta_fn)
+        dynamic_beta = dc_weight * self.dynamic_beta_scale
+
+        beta = dynamic_beta + self.static_beta
+
+        mix_h = einsum(alpha, residuals, '... s t, ... s d -> ... t d')
+
+        branch_input, residuals = mix_h[..., :-num_streams, :], mix_h[..., -num_streams:, :]
+
+        branch_input = rearrange(branch_input, 'b ... i d -> (i b) ... d')
+
+        if self.channel_first:
+            branch_input = rearrange(branch_input, 'b ... d -> b d ...')
+
+        return branch_input, residuals, dict(beta = beta)
+
+    def depth_connection(self, branch_output, residuals, *, beta):
+        # 'depth' connection
+
+        if self.channel_first:
+            branch_output = rearrange(branch_output, 'b d ... -> b ... d')
+
+        branch_output = rearrange(branch_output, '(i b) ... -> i b ...', i = self.num_branch_inputs)
+
+        residuals = einsum(branch_output, beta, 'i b ... d, b ... s i -> b ... s d') + residuals
+
+        output = rearrange(residuals, 'b ... s d -> (b s) ... d')
+
+        if self.channel_first:
+            output = rearrange(output, 'b ... d -> b d ...')
+
+        return output
+
+    def decorate_branch(self, branch: Callable):
+        assert not exists(self.branch), 'branch was already wrapped on init'
+
+        def forward_and_add_residual(residual, *args, **kwargs):
+            branch_input, add_residual = self.forward(residual)
+
+            branch_output = branch(branch_input, *args, **kwargs)
+
+            residual = add_residual(branch_output)
+
+            return residual
+
+        return forward_and_add_residual
+
+    def forward(self, residuals, *branch_args, **branch_kwargs):
+
+        branch_input, residuals, residual_kwargs = self.width_connection(residuals)
+
+        def add_residual_fn(branch_out):
+            (branch_out, *rest), tree_spec = tree_flatten(branch_out)
+
+            branch_out = self.depth_connection(branch_out, residuals, **residual_kwargs)
+
+            return tree_unflatten((branch_out, *rest), tree_spec)
+
+        if not exists(self.branch):
+            return branch_input, add_residual_fn
+
+        branch_output = self.branch(branch_input, *branch_args, **branch_kwargs)
+
+        return add_residual_fn(branch_output)

{hyper_connections-0.0.12 → hyper_connections-0.0.14}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "hyper-connections"
-version = "0.0.12"
+version = "0.0.14"
 description = "Hyper-Connections"
 authors = [
     { name = "Phil Wang", email = "lucidrains@gmail.com" }