File size: 4,208 Bytes
d766458 | 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 | # Copyright (c) Facebook, Inc. and its affiliates.
#
# This source code is licensed under the MIT license found in the
# LICENSE file in the root directory of this source tree.
import joblib
import jax.numpy as jnp
import numpy as np
import haiku as hk
import jax
from .modules import (
AxialTransformerLayer,
EmbedPosition,
MSAPositionEmbedding,
ContactPredictionHead,
LmHead,
)
from colabdesign.shared.prng import SafeKey
class MSATransformer(hk.Module):
def __init__(self, alphabet, config):
super().__init__()
self.alphabet_size = len(alphabet)
self.padding_idx = alphabet.padding_idx
self.mask_idx = alphabet.mask_idx
self.cls_idx = alphabet.cls_idx
self.eos_idx = alphabet.eos_idx
self.prepend_bos = alphabet.prepend_bos
self.append_eos = alphabet.append_eos
self.config = config
self.dropout = config.dropout
self.embed_tokens = hk.Embed(
vocab_size=self.alphabet_size,
embed_dim=self.config.embed_dim,
)
self.msa_position_embedding = MSAPositionEmbedding(self.config.embed_dim)
self.safe_key = SafeKey(hk.next_rng_key())
self.layers = [
AxialTransformerLayer(self.config)
for _ in range(self.config.layer_num)
]
self.contact_head = ContactPredictionHead(
self.config.layer_num * self.config.RowAtt.head_num,
self.prepend_bos,
self.append_eos,
eos_idx=self.eos_idx,
)
self.embed_positions = EmbedPosition(
self.config,
self.padding_idx,
)
self.emb_layer_norm_before = hk.LayerNorm(-1, create_scale=True, create_offset=True)
self.emb_layer_norm_after = hk.LayerNorm(-1, create_scale=True, create_offset=True)
self.lm_head = LmHead(
config=self.config,
output_dim=self.alphabet_size,
weight=self.embed_tokens.embeddings.transpose(),
)
def __call__(self, tokens):
num_alignments, seqlen = tokens.shape
padding_mask = jnp.equal(tokens, self.padding_idx) # R, C
x = self.embed_tokens(tokens)
x += self.embed_positions(tokens)
x += self.msa_position_embedding(tokens)
x = self.emb_layer_norm_before(x)
self.safe_key, use_key = self.safe_key.split()
x = hk.dropout(use_key.get(), self.dropout, x)
x = x * (1 - jnp.expand_dims(padding_mask, axis=-1))
row_attn_weights = []
col_attn_weights = []
for layer_idx, layer in enumerate(self.layers):
x = layer(
x,
self_attn_padding_mask=padding_mask,
)
x, col_attn, row_attn = x
col_attn_weights.append(col_attn)
row_attn_weights.append(row_attn)
x = self.emb_layer_norm_after(x)
x = self.lm_head(x)
result = {"logits": x}
# col_attentions: L x H x C x R x R
col_attentions = jnp.stack(col_attn_weights, 0)
# row_attentions: L x H x C x C
row_attentions = jnp.stack(row_attn_weights, 0)
result["col_attentions"] = col_attentions
result["row_attentions"] = row_attentions
contacts = self.contact_head(tokens, row_attentions)
result["contacts"] = contacts
return result
class RunModel:
'''container for msa transformer'''
def __init__(self, alphabet, config):
self.padding_idx = alphabet.padding_idx
def _forward(tokens):
model = MSATransformer(alphabet, config)
return model(tokens)
_forward_t = hk.transform(_forward)
self.init = jax.jit(_forward_t.init)
self.apply = jax.jit(_forward_t.apply)
self.key = jax.random.PRNGKey(42)
def load_params(self, path):
self.params = joblib.load(path)
def __call__(self, tokens):
assert tokens.ndim == 2
num_alignments, seqlen = tokens.shape
if num_alignments > 1024:
raise RuntimeError(
"Using model with MSA position embedding trained on maximum MSA "
f"depth of 1024, but received {num_alignments} alignments."
)
self.key, use_key = jax.random.split(self.key)
result = self.apply(self.params, use_key, tokens)
result_new = {}
for ikey in result.keys():
result_new[ikey] = np.array(result[ikey])
return result_new
|