File size: 6,328 Bytes
a371c7e | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 | from __future__ import annotations
import json
import pathlib
import tempfile
import unittest
import torch
from safetensors.torch import load_file, save_file
RELEASE_ROOT = pathlib.Path(__file__).resolve().parents[1]
import sys
sys.path.insert(0, str(RELEASE_ROOT))
from evotemplate_na.inference import load_artifact # noqa: E402
from evotemplate_na.model import EvoTemplateConfig, EvoTemplateNA # noqa: E402
from hub_loader import HubLoadError, load_release, load_release_spec # noqa: E402
from hub_predict import normalize_input, predict # noqa: E402
def tiny_model_config() -> dict:
return {
"config_version": 1,
"model_type": "evotemplate_na",
"student": {
"model_dim": 16,
"num_layers": 1,
"num_heads": 2,
"ffn_dim": 32,
"max_chains": 4,
"dropout": 0.0,
"rotary_base": 10000.0,
"vocab_size": 11,
"polymer_vocab_size": 3,
},
"pair_decoder": {
"token_dim": 16,
"pair_rank": 4,
"pair_dim": 8,
"num_pair_blocks": 1,
"relative_position_dim": 4,
"max_relative_position": 8,
"base_embedding_dim": 4,
"chain_pair_embedding_dim": 2,
"polymer_pair_embedding_dim": 2,
"slot_embedding_dim": 4,
"num_slots": 4,
"pb_bins": 40,
"auxiliary_distance_bins": 10,
"pairing_classes": 3,
"block_size": 2,
"dropout": 0.0,
"allow_cross_chain": True,
},
}
def tokenizer_config(max_chains: int) -> dict:
return {
"schema_version": 1,
"max_chains": max_chains,
"vocabulary": [
"A", "C", "G", "U", "T", "N", "MASK", "BOS", "EOS", "PAD",
"CHAIN_BOUNDARY",
],
"one_token_per_nucleotide": True,
"explicit_chain_boundary": True,
"reject_t_in_rna": True,
"reject_u_in_dna": True,
"silent_ut_conversion": False,
}
def build_tiny_release(root: pathlib.Path) -> None:
model_config = tiny_model_config()
tokenizer = tokenizer_config(4)
torch.manual_seed(17)
model = EvoTemplateNA(EvoTemplateConfig.from_dict(model_config))
state = {
name: value.detach().to(dtype=torch.bfloat16).contiguous()
for name, value in model.state_dict().items()
}
save_file(state, str(root / "model.safetensors"))
config = {
"schema_version": 1,
"model_type": "evotemplate_na",
"precision": "bfloat16",
"model_config": model_config,
"checkpoint_file": "model.safetensors",
"quantization": None,
}
(root / "config.json").write_text(json.dumps(config), encoding="utf-8")
(root / "tokenizer.json").write_text(
json.dumps(tokenizer), encoding="utf-8"
)
class HubRuntimeTest(unittest.TestCase):
def test_real_release_spec_loads_slim_metadata(self) -> None:
spec = load_release_spec(RELEASE_ROOT)
self.assertEqual(spec.precision, "bfloat16")
self.assertEqual(spec.model_config.student.max_chains, 128)
def test_tiny_bf16_safetensors_round_trip_and_k4_artifact(self) -> None:
with tempfile.TemporaryDirectory() as temporary:
root = pathlib.Path(temporary)
release = root / "release"
release.mkdir()
build_tiny_release(release)
loaded = load_release(release, device="cpu", dtype="float32")
self.assertEqual(loaded.dtype, torch.float32)
request = root / "input.json"
request.write_text(
json.dumps(
{
"name": "tiny-rna-dna",
"chains": [
{
"chain_id": "R",
"polymer_type": "RNA",
"sequence": "AG",
},
{
"chain_id": "D",
"polymer_type": "DNA",
"sequence": "CT",
},
],
}
),
encoding="utf-8",
)
output = root / "artifact"
result = predict(
model_dir=release,
input_path=request,
output_directory=output,
device="cpu",
dtype="float32",
block_size=2,
)
self.assertEqual(result.length, 4)
artifact = load_artifact(output)
self.assertEqual(artifact.arrays["dgram_39"].shape, (4, 4, 4, 39))
def test_strict_loader_rejects_incomplete_safetensors(self) -> None:
with tempfile.TemporaryDirectory() as temporary:
release = pathlib.Path(temporary)
build_tiny_release(release)
weights = release / "model.safetensors"
state = load_file(str(weights), device="cpu")
state.pop(next(iter(state)))
replacement = release / "replacement.safetensors"
save_file(state, str(replacement))
replacement.replace(weights)
with self.assertRaisesRegex(HubLoadError, "key set differs"):
load_release(release, device="cpu", dtype="float32")
def test_af3_input_expands_na_chain_ids_and_ignores_other_molecules(self) -> None:
payload = {
"dialect": "alphafold3",
"name": "af3-input",
"sequences": [
{"protein": {"id": "P", "sequence": "MKT"}},
{"rna": {"id": ["R", "R.1"], "sequence": "AGCU"}},
{"dna": {"id": "D", "sequence": "ACGT"}},
{"ligand": {"id": "L", "ccdCodes": ["MG"]}},
],
}
name, chains, pairs, max_length = normalize_input(
payload, fallback_artifact_id="fallback"
)
self.assertEqual(name, "af3-input")
self.assertEqual([chain["chain_id"] for chain in chains], ["R", "R.1", "D"])
self.assertEqual(len(pairs), 6)
self.assertEqual(max_length, 12)
if __name__ == "__main__":
unittest.main()
|