Make more components TorchScript-able, add tracing

openfold/model/evoformer.py

@@ -71,11 +71,25 @@ class MSATransition(nn.Module):
         m = self.linear_2(m) * mask
         return m
 
+    @torch.jit.ignore
+    def _chunk(self,
+        m: torch.Tensor,
+        mask: torch.Tensor,
+        chunk_size: int,
+    ) -> torch.Tensor:
+         return chunk_layer(
+             self._transition,
+             {"m": m, "mask": mask},
+             chunk_size=chunk_size,
+             no_batch_dims=len(m.shape[:-2]),
+         )
+
+
     def forward(
         self,
         m: torch.Tensor,
-        mask: torch.Tensor = None,
-        chunk_size: int = None,
+        mask: Optional[torch.Tensor] = None,
+        chunk_size: Optional[int] = None,
     ) -> torch.Tensor:
         """
         Args:
@@ -95,16 +109,10 @@ class MSATransition(nn.Module):
 
         m = self.layer_norm(m)
 
-        inp = {"m": m, "mask": mask}
         if chunk_size is not None:
-            m = chunk_layer(
-                self._transition,
-                inp,
-                chunk_size=chunk_size,
-                no_batch_dims=len(m.shape[:-2]),
-            )
+            m = self._chunk(m, mask, chunk_size)
         else:
-            m = self._transition(**inp)
+            m = self._transition(m, mask)
 
         return m
 
@@ -201,9 +209,11 @@ class EvoformerBlock(nn.Module):
         z: torch.Tensor,
         msa_mask: torch.Tensor,
         pair_mask: torch.Tensor,
-        chunk_size: int,
+        chunk_size: Optional[int] = None,
         _mask_trans: bool = True,
     ) -> Tuple[torch.Tensor, torch.Tensor]:
+        #print(torch.cuda.memory_summary())
+        
         # DeepMind doesn't mask these transitions in the source, so _mask_trans
         # should be disabled to better approximate the exact activations of
         # the original.

openfold/model/msa.py

@@ -16,7 +16,7 @@
 import math
 import torch
 import torch.nn as nn
-from typing import Optional
+from typing import Optional, List
 
 from openfold.model.primitives import Linear, Attention, GlobalAttention
 from openfold.utils.tensor_utils import (
@@ -63,6 +63,8 @@ class MSAAttention(nn.Module):
 
         self.layer_norm_m = nn.LayerNorm(self.c_in)
 
+        self.layer_norm_z = None
+        self.linear_z = None
         if self.pair_bias:
             self.layer_norm_z = nn.LayerNorm(self.c_z)
             self.linear_z = Linear(
@@ -73,7 +75,25 @@ class MSAAttention(nn.Module):
             self.c_in, self.c_in, self.c_in, self.c_hidden, self.no_heads
         )
 
-    def forward(self, m, chunk_size, z=None, mask=None):
+    @torch.jit.ignore
+    def _chunk(self, 
+        m: torch.Tensor,
+        biases: List[torch.Tensor],
+        chunk_size: int,
+    ) -> torch.Tensor:
+        return chunk_layer(
+            self.mha,
+            {"q_x": m, "k_x": m, "v_x": m, "biases": biases},
+            chunk_size=chunk_size,
+            no_batch_dims=len(m.shape[:-2]),
+        )
+
+    def forward(self, 
+        m: torch.Tensor, 
+        z: Optional[torch.Tensor] = None, 
+        mask: Optional[torch.Tensor] = None, 
+        chunk_size: Optional[int] = None,
+    ) -> torch.Tensor:
         """
         Args:
             m:
@@ -83,6 +103,11 @@ class MSAAttention(nn.Module):
                 pair_bias is True
             mask:
                 [*, N_seq, N_res] MSA mask
+            chunk_size:
+                Size of chunks into which the inputs are split along their
+                batch dimensions. A low value decreases memory overhead at the 
+                cost of slower execution. Chunking is not performed by default.
+                
         """
         # [*, N_seq, N_res, C_m]
         m = self.layer_norm_m(m)
@@ -106,7 +131,11 @@ class MSAAttention(nn.Module):
         
         biases = [bias]
 
-        if self.pair_bias:
+        if (self.pair_bias and 
+            z is not None and                       # For the 
+            self.layer_norm_z is not None and       # benefit of
+            self.linear_z is not None               # TorchScript
+        ):
             # [*, N_res, N_res, C_z]
             z = self.layer_norm_z(z)
 
@@ -118,16 +147,10 @@ class MSAAttention(nn.Module):
 
             biases.append(z)
 
-        mha_inputs = {"q_x": m, "k_x": m, "v_x": m, "biases": biases}
         if chunk_size is not None:
-            m = chunk_layer(
-                self.mha,
-                mha_inputs,
-                chunk_size=chunk_size,
-                no_batch_dims=len(m.shape[:-2]),
-            )
+            m = self._chunk(m, biases, chunk_size)
         else:
-            m = self.mha(**mha_inputs)
+            m = self.mha(q_x=m, k_x=m, v_x=m, biases=biases)
 
         return m
 
@@ -161,9 +184,12 @@ class MSARowAttentionWithPairBias(MSAAttention):
         )
 
 
-class MSAColumnAttention(MSAAttention):
+class MSAColumnAttention(nn.Module):
     """
     Implements Algorithm 8.
+
+    By rights, this should also be a subclass of MSAAttention. Alas,
+    most inheritance isn't supported by TorchScript.
     """
 
     def __init__(self, c_m, c_hidden, no_heads, inf=1e9):
@@ -178,7 +204,14 @@ class MSAColumnAttention(MSAAttention):
             inf:
                 Large number used to construct attention masks
         """
-        super(MSAColumnAttention, self).__init__(
+        super(MSAColumnAttention, self).__init__()
+        
+        self.c_m = c_m
+        self.c_hidden = c_hidden
+        self.no_heads = no_heads
+        self.inf = inf
+
+        self._msa_att = MSAAttention(
             c_in=c_m,
             c_hidden=c_hidden,
             no_heads=no_heads,
@@ -187,31 +220,40 @@ class MSAColumnAttention(MSAAttention):
             inf=inf,
         )
 
-    def forward(self, m, chunk_size, mask=None):
+    def forward(self, 
+        m: torch.Tensor, 
+        mask: Optional[torch.Tensor] = None, 
+        chunk_size: Optional[int] = None
+    ) -> torch.Tensor:
         """
         Args:
             m:
                 [*, N_seq, N_res, C_m] MSA embedding
             mask:
                 [*, N_seq, N_res] MSA mask
+            chunk_size:
+                Size of chunks into which the inputs are split along their
+                batch dimensions. A low value decreases memory overhead at the 
+                cost of slower execution. Chunking is not performed by default.
         """ 
         # [*, N_res, N_seq, C_in]
         m = m.transpose(-2, -3)
         if mask is not None:
             mask = mask.transpose(-1, -2)
 
-        m = super().forward(m, chunk_size=chunk_size, mask=mask)
+        m = self._msa_att(m, mask=mask, chunk_size=chunk_size)
 
         # [*, N_seq, N_res, C_in]
         m = m.transpose(-2, -3)
         if mask is not None:
             mask = mask.transpose(-1, -2)
+
         return m
 
 
 class MSAColumnGlobalAttention(nn.Module):
     def __init__(
-        self, c_in, c_hidden, no_heads, inf=1e9, eps=1e-10
+        self, c_in, c_hidden, no_heads, inf=1e9, eps=1e-10,
     ):
         super(MSAColumnGlobalAttention, self).__init__()
 
@@ -231,8 +273,28 @@ class MSAColumnGlobalAttention(nn.Module):
             eps=eps,
         )
 
+    @torch.jit.ignore
+    def _chunk(self,
+        m: torch.Tensor,
+        mask: torch.Tensor,
+        chunk_size: int,
+    ) -> torch.Tensor:
+        mha_input = {
+            "m": m,
+            "mask": mask,
+        }
+        return chunk_layer(
+            self.global_attention,
+            mha_input,
+            chunk_size=chunk_size,
+            no_batch_dims=len(m.shape[:-2]),
+        )
+
     def forward(
-        self, m: torch.Tensor, chunk_size, mask: Optional[torch.Tensor] = None
+        self, 
+        m: torch.Tensor, 
+        mask: Optional[torch.Tensor] = None, 
+        chunk_size: Optional[int] = None,
     ) -> torch.Tensor:
         n_seq, n_res, c_in = m.shape[-3:]
 
@@ -251,19 +313,10 @@ class MSAColumnGlobalAttention(nn.Module):
         # [*, N_res, N_seq, C_in]
         m = self.layer_norm_m(m)
 
-        mha_input = {
-            "m": m,
-            "mask": mask,
-        }
         if chunk_size is not None:
-            m = chunk_layer(
-                self.global_attention,
-                mha_input,
-                chunk_size=chunk_size,
-                no_batch_dims=len(m.shape[:-2]),
-            )
+            m = self._chunk(m, mask, chunk_size) 
         else:
-            m = self.global_attention(m=mha_input["m"], mask=mha_input["mask"])
+            m = self.global_attention(m=m, mask=mask)
 
         # [*, N_seq, N_res, C_in]
         m = m.transpose(-2, -3)

openfold/model/outer_product_mean.py

@@ -14,6 +14,8 @@
 # limitations under the License.
 
 from functools import partial
+from typing import Optional
+
 import torch
 import torch.nn as nn
 
@@ -38,6 +40,7 @@ class OuterProductMean(nn.Module):
         """
         super(OuterProductMean, self).__init__()
 
+        self.c_m = c_m
         self.c_z = c_z
         self.c_hidden = c_hidden
         self.eps = eps
@@ -52,14 +55,43 @@ class OuterProductMean(nn.Module):
         outer = torch.einsum("...bac,...dae->...bdce", a, b)
 
         # [*, N_res, N_res, C * C]
-        outer = outer.reshape(*outer.shape[:-2], -1)
+        outer = outer.reshape(outer.shape[:-2] + (-1,))
 
         # [*, N_res, N_res, C_z]
         outer = self.linear_out(outer)
 
         return outer
 
-    def forward(self, m, chunk_size, mask=None):
+    @torch.jit.ignore
+    def _chunk(self, 
+        a: torch.Tensor, 
+        b: torch.Tensor, 
+        chunk_size: int
+    ) -> torch.Tensor:
+        # Since the "batch dim" in this case is not a true batch dimension
+        # (in that the shape of the output depends on it), we need to
+        # iterate over it ourselves
+        a_reshape = a.reshape((-1,) + a.shape[-3:])
+        b_reshape = b.reshape((-1,) + b.shape[-3:])
+        out = []
+        for a_prime, b_prime in zip(a_reshape, b_reshape):
+            outer = chunk_layer(
+                partial(self._opm, b=b_prime),
+                {"a": a_prime},
+                chunk_size=chunk_size,
+                no_batch_dims=1,
+            )
+            out.append(outer)
+        outer = torch.stack(out, dim=0)
+        outer = outer.reshape(a.shape[:-3] + outer.shape[1:])
+
+        return outer
+
+    def forward(self, 
+        m: torch.Tensor, 
+        mask: Optional[torch.Tensor] = None,
+        chunk_size: Optional[int] = None
+    ) -> torch.Tensor:
         """
         Args:
             m:
@@ -84,22 +116,7 @@ class OuterProductMean(nn.Module):
         b = b.transpose(-2, -3)
 
         if chunk_size is not None:
-            # Since the "batch dim" in this case is not a true batch dimension
-            # (in that the shape of the output depends on it), we need to
-            # iterate over it ourselves
-            a_reshape = a.reshape(-1, *a.shape[-3:])
-            b_reshape = b.reshape(-1, *b.shape[-3:])
-            out = []
-            for a_prime, b_prime in zip(a_reshape, b_reshape):
-                outer = chunk_layer(
-                    partial(self._opm, b=b_prime),
-                    {"a": a_prime},
-                    chunk_size=chunk_size,
-                    no_batch_dims=1,
-                )
-                out.append(outer)
-            outer = torch.stack(out, dim=0)
-            outer = outer.reshape(*a.shape[:-3], *outer.shape[1:])
+            outer = self._chunk(a, b, chunk_size)
         else:
             outer = self._opm(a, b)
 

openfold/model/pair_transition.py

@@ -12,6 +12,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+from typing import Optional
 
 import torch
 import torch.nn as nn
@@ -54,7 +55,25 @@ class PairTransition(nn.Module):
 
         return z
 
-    def forward(self, z, chunk_size, mask=None):
+    @torch.jit.ignore
+    def _chunk(self,
+        z: torch.Tensor,
+        mask: torch.Tensor,
+        chunk_size: int,
+    ) -> torch.Tensor:
+        return chunk_layer(
+            self._transition,
+            {"z": z, "mask": mask},
+            chunk_size=chunk_size,
+            no_batch_dims=len(z.shape[:-2]),
+        )
+
+
+    def forward(self, 
+        z: torch.Tensor, 
+        mask: Optional[torch.Tensor] = None,
+        chunk_size: Optional[int] = None,
+    ) -> torch.Tensor:
         """
         Args:
             z:
@@ -72,15 +91,9 @@ class PairTransition(nn.Module):
         # [*, N_res, N_res, C_z]
         z = self.layer_norm(z)
 
-        inp = {"z": z, "mask": mask}
         if chunk_size is not None:
-            z = chunk_layer(
-                self._transition,
-                inp,
-                chunk_size=chunk_size,
-                no_batch_dims=len(z.shape[:-2]),
-            )
+            z = self._chunk(z, mask, chunk_size)
         else:
-            z = self._transition(**inp)
+            z = self._transition(z=z, mask=mask)
 
         return z

openfold/model/structure_module.py

@@ -155,17 +155,16 @@ class InvariantPointAttention(nn.Module):
     """
     Implements Algorithm 22.
     """
-
     def __init__(
         self,
-        c_s,
-        c_z,
-        c_hidden,
-        no_heads,
-        no_qk_points,
-        no_v_points,
-        inf=1e5,
-        eps=1e-8,
+        c_s: int,
+        c_z: int,
+        c_hidden: int,
+        no_heads: int,
+        no_qk_points: int,
+        no_v_points: int,
+        inf: float = 1e5,
+        eps: float = 1e-8,
     ):
         """
         Args:

openfold/model/template.py

@@ -12,9 +12,10 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
-
 from functools import partial
 import math
+from typing import Optional, List
+
 import torch
 import torch.nn as nn
 
@@ -71,7 +72,32 @@ class TemplatePointwiseAttention(nn.Module):
             gating=False,
         )
 
-    def forward(self, t, z, chunk_size, template_mask=None):
+    def _chunk(self,
+        z: torch.Tensor,
+        t: torch.Tensor,
+        biases: List[torch.Tensor],
+        chunk_size: int,
+    ) -> torch.Tensor:
+        mha_inputs = {
+            "q_x": z,
+            "k_x": t,
+            "v_x": t,
+            "biases": biases,
+        }
+        return chunk_layer(
+            self.mha,
+            mha_inputs,
+            chunk_size=chunk_size,
+            no_batch_dims=len(z.shape[:-2]),
+        )
+
+
+    def forward(self, 
+        t: torch.Tensor, 
+        z: torch.Tensor, 
+        template_mask: Optional[torch.Tensor] = None,
+        chunk_size: Optional[int] = None
+    ) -> torch.Tensor:
         """
         Args:
             t:
@@ -95,21 +121,11 @@ class TemplatePointwiseAttention(nn.Module):
         t = permute_final_dims(t, (1, 2, 0, 3))
 
         # [*, N_res, N_res, 1, C_z]
-        mha_inputs = {
-            "q_x": z,
-            "k_x": t,
-            "v_x": t,
-            "biases": [bias],
-        }
+        biases = [bias]
         if chunk_size is not None:
-            z = chunk_layer(
-                self.mha,
-                mha_inputs,
-                chunk_size=chunk_size,
-                no_batch_dims=len(z.shape[:-2]),
-            )
+            z = self._chunk(z, t, biases, chunk_size)
         else:
-            z = self.mha(**mha_inputs)
+            z = self.mha(q_x=z, k_x=t, v_x=t, biases=biases)
 
         # [*, N_res, N_res, C_z]
         z = z.squeeze(-2)
@@ -120,13 +136,13 @@ class TemplatePointwiseAttention(nn.Module):
 class TemplatePairStackBlock(nn.Module):
     def __init__(
         self,
-        c_t,
-        c_hidden_tri_att,
-        c_hidden_tri_mul,
-        no_heads,
-        pair_transition_n,
-        dropout_rate,
-        inf,
+        c_t: int,
+        c_hidden_tri_att: int,
+        c_hidden_tri_mul: int,
+        no_heads: int,
+        pair_transition_n: int,
+        dropout_rate: float,
+        inf: float,
         **kwargs,
     ):
         super(TemplatePairStackBlock, self).__init__()
@@ -169,7 +185,12 @@ class TemplatePairStackBlock(nn.Module):
             self.pair_transition_n,
         )
 
-    def forward(self, z, mask, chunk_size, _mask_trans=True):
+    def forward(self, 
+        z: torch.Tensor, 
+        mask: torch.Tensor, 
+        chunk_size: Optional[int] = None, 
+        _mask_trans: bool = True
+    ):
         single_templates = [
             t.unsqueeze(-4) for t in torch.unbind(z, dim=-4)
         ]
@@ -208,8 +229,8 @@ class TemplatePairStackBlock(nn.Module):
             )
             single = single + self.pair_transition(
                 single,
+                mask=single_mask if _mask_trans else None,
                 chunk_size=chunk_size,
-                mask=single_mask if _mask_trans else None
             )
 
             single_templates[i] = single

openfold/model/torchscript.py

-from typing import Optional, Sequence
+# Copyright 2021 AlQuraishi Laboratory
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from typing import Optional, Sequence, Tuple
 
 import torch
 import torch.nn as nn
 
+from openfold.model.dropout import (
+    DropoutRowwise,
+    DropoutColumnwise,
+)
+from openfold.model.evoformer import (
+    EvoformerBlock,
+    EvoformerStack,
+)
+from openfold.model.outer_product_mean import OuterProductMean
+from openfold.model.msa import (
+    MSARowAttentionWithPairBias, 
+    MSAColumnAttention,
+    MSAColumnGlobalAttention,
+)
+from openfold.model.pair_transition import PairTransition
 from openfold.model.primitives import Attention, GlobalAttention
+from openfold.model.structure_module import (
+    InvariantPointAttention,
+    BackboneUpdate,
+)
+from openfold.model.template import TemplatePairStackBlock
+from openfold.model.triangular_attention import (
+    TriangleAttentionStartingNode,
+    TriangleAttentionEndingNode,
+)
+from openfold.model.triangular_multiplicative_update import (
+    TriangleMultiplicationOutgoing,
+    TriangleMultiplicationIncoming,
+)
+
+
+def script_preset_(model: torch.nn.Module):
+    """
+    TorchScript a handful of low-level but frequently used submodule types 
+    that are known to be scriptable.
+
+    Args:
+        model: 
+            A torch.nn.Module. It should contain at least some modules from 
+            this repository, or this function won't do anything.
+    """
+    script_submodules_(
+        model, 
+        [
+            nn.Dropout,
+            Attention,
+            GlobalAttention,
+            EvoformerBlock,
+            #TemplatePairStackBlock,
+        ], 
+        attempt_trace=False,
+        batch_dims=None,
+    ) 
+
+    
+def _get_module_device(module: torch.nn.Module) -> torch.device:
+    """
+    Fetches the device of a module, assuming that all of the module's
+    parameters reside on a single device
+
+    Args:
+        module: A torch.nn.Module
+    Returns:
+        The module's device
+    """
+    return next(module.parameters()).device
+
+
+def _trace_module(module, batch_dims=None):
+    if(batch_dims is None):
+        batch_dims = ()
+
+    # Stand-in values
+    n_seq = 10
+    n_res = 10
+
+    device = _get_module_device(module)
+
+    def msa(channel_dim):
+        return torch.rand(
+            (*batch_dims, n_seq, n_res, channel_dim),
+            device=device,
+        )
+
+    def pair(channel_dim):
+        return torch.rand(
+            (*batch_dims, n_res, n_res, channel_dim),
+            device=device,
+        )
 
+    if(isinstance(module, MSARowAttentionWithPairBias)):
+        inputs = {
+            "forward": (
+                msa(module.c_in), # m
+                pair(module.c_z), # z
+                torch.randint(
+                    0, 2, 
+                    (*batch_dims, n_seq, n_res)
+                ), # mask
+            ),
+        }
+    elif(isinstance(module, MSAColumnAttention)):
+        inputs = {
+            "forward": (
+                msa(module.c_in), # m
+                torch.randint(
+                    0, 2, 
+                    (*batch_dims, n_seq, n_res)
+                ), # mask
+            ),
+        }
+    elif(isinstance(module, OuterProductMean)):
+        inputs = {
+            "forward": (
+                msa(module.c_m),
+                torch.randint(
+                    0, 2,
+                    (*batch_dims, n_seq, n_res)
+                )
+            )
+        }
+        module = OPM(module)
+    else:
+        raise TypeError(
+            f"tracing is not supported for modules of type {type(module)}"
+        )
+
+    return torch.jit.trace_module(module, inputs)
+
+
+def _script_submodules_helper_(
+    model,
+    types,
+    attempt_trace,
+    to_trace,
+):
+    for name, child in model.named_children():
+        if(types is None or any(isinstance(child, t) for t in types)):
+            try:
+                scripted = torch.jit.script(child)
+                setattr(model, name, scripted)
+                continue
+            except (RuntimeError, torch.jit.frontend.NotSupportedError) as e:
+                if(attempt_trace):
+                    to_trace.add(type(child))
+                else:
+                    raise e
+        
+        _script_submodules_helper_(child, types, attempt_trace, to_trace)
+
+
+def _trace_submodules_(
+    model,
+    types,
+    batch_dims=None,
+):
+    for name, child in model.named_children():
+        if(any(isinstance(child, t) for t in types)):
+            traced = _trace_module(child, batch_dims=batch_dims)
+            setattr(model, name, traced)
+        else:
+            _trace_submodules_(child, types, batch_dims=batch_dims)
 
-def script_primitives_(model):
-    script_submodules_(model, [Attention, GlobalAttention])
 
 def script_submodules_(
     model: nn.Module,
     types: Optional[Sequence[type]] = None,
+    attempt_trace: Optional[bool] = True,
+    batch_dims: Optional[Tuple[int]] = None,
 ):
     """
     Convert all submodules whose types match one of those in the input 
@@ -21,11 +195,22 @@ def script_submodules_(
     When types is None, all submodules are scripted.
 
     Args:
-            model: A torch.nn.Module
-            types: A list of types of submodules to script
+        model: 
+            A torch.nn.Module
+        types: 
+            A list of types of submodules to script
+        attempt_trace: 
+            Whether to attempt to trace specified modules if scripting 
+            fails. Recall that tracing eliminates all conditional 
+            logic---with great tracing comes the mild responsibility of 
+            having to remember to ensure that the modules in question 
+            perform the same computations no matter what.
     """
-    for name, child in model.named_children():
-        if(types is None or any(isinstance(child, t) for t in types)):
-            setattr(model, name, torch.jit.script(child))
-        else:
-            script_submodules_(child, types)
+    to_trace = set()
+
+    # Aggressively script as much as possible first...
+    _script_submodules_helper_(model, types, attempt_trace, to_trace)
+  
+    # ... and then trace stragglers.
+    if(attempt_trace and len(to_trace) > 0):
+        _trace_submodules_(model, to_trace, batch_dims=batch_dims)

openfold/model/triangular_attention.py

@@ -15,6 +15,8 @@
 
 from functools import partialmethod
 import math
+from typing import Optional, List
+
 import torch
 import torch.nn as nn
 
@@ -55,7 +57,30 @@ class TriangleAttention(nn.Module):
             self.c_in, self.c_in, self.c_in, self.c_hidden, self.no_heads
         )
 
-    def forward(self, x, chunk_size, mask=None):
+    @torch.jit.ignore
+    def _chunk(self,
+        x: torch.Tensor,
+        biases: List[torch.Tensor],
+        chunk_size: int,
+    ) -> torch.Tensor:
+        mha_inputs = {
+            "q_x": x,
+            "k_x": x,
+            "v_x": x,
+            "biases": biases,
+        }
+        return chunk_layer(
+            self.mha,
+            mha_inputs,
+            chunk_size=chunk_size,
+            no_batch_dims=len(x.shape[:-2]),
+        )
+
+    def forward(self, 
+        x: torch.Tensor, 
+        mask: Optional[torch.Tensor] = None,
+        chunk_size: Optional[int] = None
+    ) -> torch.Tensor:
         """
         Args:
             x:
@@ -86,21 +111,12 @@ class TriangleAttention(nn.Module):
         # [*, 1, H, I, J]
         triangle_bias = triangle_bias.unsqueeze(-4)
 
-        mha_inputs = {
-            "q_x": x,
-            "k_x": x,
-            "v_x": x,
-            "biases": [mask_bias, triangle_bias],
-        }
+        biases = [mask_bias, triangle_bias]
+
         if chunk_size is not None:
-            x = chunk_layer(
-                self.mha,
-                mha_inputs,
-                chunk_size=chunk_size,
-                no_batch_dims=len(x.shape[:-2]),
-            )
+            x = self._chunk(x, biases, chunk_size)
         else:
-            x = self.mha(**mha_inputs)
+            x = self.mha(q_x=x, k_x=x, v_x=x, biases=biases)
 
         if not self.starting:
             x = x.transpose(-2, -3)

openfold/model/triangular_multiplicative_update.py

@@ -14,6 +14,8 @@
 # limitations under the License.
 
 from functools import partialmethod
+from typing import Optional
+
 import torch
 import torch.nn as nn
 
@@ -25,7 +27,6 @@ class TriangleMultiplicativeUpdate(nn.Module):
     """
     Implements Algorithms 11 and 12.
     """
-
     def __init__(self, c_z, c_hidden, _outgoing=True):
         """
         Args:
@@ -51,39 +52,16 @@ class TriangleMultiplicativeUpdate(nn.Module):
 
         self.sigmoid = nn.Sigmoid()
 
-        cp = self._outgoing_matmul if self._outgoing else self._incoming_matmul
-        self.combine_projections = cp
-
-    def _outgoing_matmul(
-        self,
-        a: torch.Tensor,  # [*, N_i, N_k, C]
-        b: torch.Tensor,  # [*, N_j, N_k, C]
-    ):
-        # [*, C, N_i, N_j]
-        p = torch.matmul(
-            permute_final_dims(a, (2, 0, 1)),
-            permute_final_dims(b, (2, 1, 0)),
-        )
-
-        # [*, N_i, N_j, C]
-        return permute_final_dims(p, (1, 2, 0))
-
-    def _incoming_matmul(
-        self,
-        a: torch.Tensor,  # [*, N_k, N_i, C]
-        b: torch.Tensor,  # [*, N_k, N_j, C]
-    ):
-
-        # [*, C, N_i, N_j]
-        p = torch.matmul(
-            permute_final_dims(a, (2, 1, 0)),
-            permute_final_dims(b, (2, 0, 1)),
-        )
-
-        # [*, N_i, N_j, C]
-        return permute_final_dims(p, (1, 2, 0))
+    def _combine_projections(
+        a: torch.Tensor,
+        b: torch.Tensor,
+    ) -> torch.Tensor:
+        raise NotImplementedError("This method needs to be overridden")
 
-    def forward(self, z, mask=None):
+    def forward(self, 
+        z: torch.Tensor, 
+        mask: Optional[torch.Tensor] = None
+    ) -> torch.Tensor:
         """
         Args:
             x:
@@ -103,7 +81,7 @@ class TriangleMultiplicativeUpdate(nn.Module):
         a = a * mask
         b = self.linear_b_p(z) * self.sigmoid(self.linear_b_g(z))
         b = b * mask
-        x = self.combine_projections(a, b)
+        x = self._combine_projections(a, b)
         x = self.layer_norm_out(x)
         x = self.linear_z(x)
         g = self.sigmoid(self.linear_g(z))
@@ -116,19 +94,36 @@ class TriangleMultiplicationOutgoing(TriangleMultiplicativeUpdate):
     """
     Implements Algorithm 11.
     """
-
-    __init__ = partialmethod(
-        TriangleMultiplicativeUpdate.__init__,
-        _outgoing=True,
+    def _combine_projections(
+        self,
+        a: torch.Tensor,  # [*, N_i, N_k, C]
+        b: torch.Tensor,  # [*, N_j, N_k, C]
+    ):
+        # [*, C, N_i, N_j]
+        p = torch.matmul(
+            permute_final_dims(a, (2, 0, 1)),
+            permute_final_dims(b, (2, 1, 0)),
         )
 
+        # [*, N_i, N_j, C]
+        return permute_final_dims(p, (1, 2, 0))
+
 
 class TriangleMultiplicationIncoming(TriangleMultiplicativeUpdate):
     """
     Implements Algorithm 12.
     """
-
-    __init__ = partialmethod(
-        TriangleMultiplicativeUpdate.__init__,
-        _outgoing=False,
+    def _combine_projections(
+        self,
+        a: torch.Tensor,  # [*, N_k, N_i, C]
+        b: torch.Tensor,  # [*, N_k, N_j, C]
+    ):
+        # [*, C, N_i, N_j]
+        p = torch.matmul(
+            permute_final_dims(a, (2, 1, 0)),
+            permute_final_dims(b, (2, 0, 1)),
         )
+
+        # [*, N_i, N_j, C]
+        return permute_final_dims(p, (1, 2, 0))
+

openfold/utils/affine_utils.py

@@ -631,7 +631,7 @@ _qtr_ind_dict = {key: ind for ind, key in enumerate(_qtr_keys)}
 
 
 def _to_mat(pairs):
-    mat = torch.zeros((4, 4))
+    mat = np.zeros((4, 4))
     for pair in pairs:
         key, value = pair
         ind = _qtr_ind_dict[key]

openfold/utils/checkpointing.py

@@ -17,10 +17,11 @@ import torch
 import torch.utils.checkpoint
 from typing import Any, Tuple, List, Callable
 
+
 BLOCK_ARG = Any
 BLOCK_ARGS = List[BLOCK_ARG]
 
-
+@torch.jit.ignore
 def checkpoint_blocks(
     blocks: List[Callable],
     args: BLOCK_ARGS,

openfold/utils/import_weights.py

@@ -217,6 +217,11 @@ def import_jax_weights_(model, npz_path, version="model_1"):
         "attention": AttentionGatedParams(matt.mha),
     }
 
+    MSAColAttParams = lambda matt: {
+        "query_norm": LayerNormParams(matt._msa_att.layer_norm_m),
+        "attention": AttentionGatedParams(matt._msa_att.mha),
+    }
+
     MSAGlobalAttParams = lambda matt: {
         "query_norm": LayerNormParams(matt.layer_norm_m),
         "attention": GlobalAttentionParams(matt.global_attention),
@@ -270,7 +275,7 @@ def import_jax_weights_(model, npz_path, version="model_1"):
             msa_col_att_params = MSAGlobalAttParams(b.msa_att_col)
         else:
             col_att_name = "msa_column_attention"
-            msa_col_att_params = MSAAttParams(b.msa_att_col)
+            msa_col_att_params = MSAColAttParams(b.msa_att_col)
 
         d = {
             "msa_row_attention_with_pair_bias": MSAAttPairBiasParams(

openfold/utils/tensor_utils.py

@@ -107,6 +107,22 @@ def tree_map(fn, tree, leaf_type):
 
 tensor_tree_map = partial(tree_map, leaf_type=torch.Tensor)
 
+def _fetch_dims(tree):
+    shapes = []
+    tree_type = type(tree)
+    if tree_type is dict:
+        for v in tree.values():
+            shapes.extend(_fetch_dims(v))
+    elif tree_type is list or tree_type is tuple:
+        for t in tree:
+            shapes.extend(_fetch_dims(t))
+    elif tree_type is torch.Tensor:
+        shapes.append(tree.shape)
+    else:
+        raise ValueError("Not supported")
+
+    return shapes
+
 
 def chunk_layer(
     layer: Callable,
@@ -141,33 +157,17 @@ def chunk_layer(
     if not (len(inputs) > 0):
         raise ValueError("Must provide at least one input")
 
-    def fetch_dims(tree):
-        shapes = []
-        tree_type = type(tree)
-        if tree_type is dict:
-            for v in tree.values():
-                shapes.extend(fetch_dims(v))
-        elif tree_type is list or tree_type is tuple:
-            for t in tree:
-                shapes.extend(fetch_dims(t))
-        elif tree_type is torch.Tensor:
-            shapes.append(tree.shape)
-        else:
-            raise ValueError("Not supported")
-
-        return shapes
-
-    initial_dims = [shape[:no_batch_dims] for shape in fetch_dims(inputs)]
+    initial_dims = [shape[:no_batch_dims] for shape in _fetch_dims(inputs)]
     orig_batch_dims = tuple([max(s) for s in zip(*initial_dims)])
 
-    def prep_inputs(t):
+    def _prep_inputs(t):
         # TODO: make this more memory efficient. This sucks
         if not sum(t.shape[:no_batch_dims]) == no_batch_dims:
             t = t.expand(*orig_batch_dims, *t.shape[no_batch_dims:])
         t = t.reshape(-1, *t.shape[no_batch_dims:])
         return t
 
-    flattened_inputs = tensor_tree_map(prep_inputs, inputs)
+    flattened_inputs = tensor_tree_map(_prep_inputs, inputs)
 
     flat_batch_dim = 1
     for d in orig_batch_dims:

run_pretrained_openfold.py

@@ -31,7 +31,7 @@ import torch
 from openfold.config import model_config
 from openfold.data import templates, feature_pipeline, data_pipeline
 from openfold.model.model import AlphaFold
-from openfold.model.torchscript import script_primitives_
+from openfold.model.torchscript import script_preset_
 from openfold.np import residue_constants, protein
 import openfold.np.relax.relax as relax
 from openfold.utils.import_weights import (
@@ -49,7 +49,7 @@ def main(args):
     model = AlphaFold(config)
     model = model.eval()
     import_jax_weights_(model, args.param_path)
-    script_primitives_(model)
+    script_preset_(model)
     model = model.to(args.model_device)
  
     template_featurizer = templates.TemplateHitFeaturizer(

train_openfold.py

@@ -2,7 +2,7 @@ import argparse
 import logging
 import os
 
-#os.environ["CUDA_VISIBLE_DEVICES"] = "6"
+os.environ["CUDA_VISIBLE_DEVICES"] = "6"
 #os.environ["MASTER_ADDR"]="10.119.81.14"
 #os.environ["MASTER_PORT"]="42069"
 #os.environ["NODE_RANK"]="0"
@@ -23,6 +23,7 @@ from openfold.data.data_modules import (
     DummyDataLoader,
 )
 from openfold.model.model import AlphaFold
+from openfold.model.torchscript import script_preset_
 from openfold.utils.callbacks import (
     EarlyStoppingVerbose,
 )
@@ -64,10 +65,6 @@ class OpenFoldWrapper(pl.LightningModule):
         # Compute loss
         loss = self.loss(outputs, batch)
 
-        #if(torch.isnan(loss) or torch.isinf(loss)):
-        #    logging.warning("loss is NaN. Skipping example...")
-        #    loss = loss.new_tensor(0., requires_grad=True)
-
         return {"loss": loss}
 
     def validation_step(self, batch, batch_idx):
@@ -121,6 +118,10 @@ def main(args):
         sd = {k[len("module."):]:v for k,v in sd.items()}
         model_module.load_state_dict(sd)
         logging.info("Successfully loaded model weights...")
+
+    # TorchScript components of the model
+    script_preset_(model_module)
+
     #data_module = DummyDataLoader("batch.pickle")
     data_module = OpenFoldDataModule(
         config=config.data,