| |
| |
| @@ -50,7 +50,8 @@ |
| |
| def __init__(self, |
| config: ml_collections.ConfigDict, |
| - params: Optional[Mapping[str, Mapping[str, np.ndarray]]] = None): |
| + params: Optional[Mapping[str, Mapping[str, np.ndarray]]] = None, |
| + is_training = False): |
| self.config = config |
| self.params = params |
| |
| @@ -58,7 +59,7 @@ |
| model = modules.AlphaFold(self.config.model) |
| return model( |
| batch, |
| - is_training=False, |
| + is_training=is_training, |
| compute_loss=False, |
| ensemble_representations=True) |
| |
| @@ -117,7 +118,7 @@ |
| logging.info('Output shape was %s', shape) |
| return shape |
| |
| - def predict(self, feat: features.FeatureDict) -> Mapping[str, Any]: |
| + def predict(self, feat: features.FeatureDict, random_seed=0) -> Mapping[str, Any]: |
| """Makes a prediction by inferencing the model on the provided features. |
| |
| Args: |
| @@ -128,14 +129,28 @@ |
| A dictionary of model outputs. |
| """ |
| self.init_params(feat) |
| - logging.info('Running predict with shape(feat) = %s', |
| - tree.map_structure(lambda x: x.shape, feat)) |
| - result = self.apply(self.params, jax.random.PRNGKey(0), feat) |
| - # This block is to ensure benchmark timings are accurate. Some blocking is |
| - # already happening when computing get_confidence_metrics, and this ensures |
| - # all outputs are blocked on. |
| - jax.tree_map(lambda x: x.block_until_ready(), result) |
| - result.update(get_confidence_metrics(result)) |
| - logging.info('Output shape was %s', |
| - tree.map_structure(lambda x: x.shape, result)) |
| - return result |
| + logging.info('Running predict with shape(feat) = %s', tree.map_structure(lambda x: x.shape, feat)) |
| + |
| + aatype = feat["aatype"] |
| + num_iters = self.config.model.num_recycle + 1 |
| + num_ensemble = self.config.data.eval.num_ensemble |
| + L = aatype.shape[1] |
| + result = {"prev":{'prev_msa_first_row': np.zeros([L,256]), |
| + 'prev_pair': np.zeros([L,L,128]), |
| + 'prev_pos': np.zeros([L,37,3])}} |
| + |
| + r = 0 |
| + key = jax.random.PRNGKey(random_seed) |
| + while r < num_iters: |
| + s = r * num_ensemble |
| + e = (r+1) * num_ensemble |
| + sub_feat = jax.tree_map(lambda x:x[s:e], feat) |
| + sub_feat["prev"] = result["prev"] |
| + result, _ = self.apply(self.params, key, sub_feat) |
| + del sub_feat |
| + confidences = get_confidence_metrics(result) |
| + result.update(confidences) |
| + r += 1 |
| + |
| + logging.info('Output shape was %s', tree.map_structure(lambda x: x.shape, result)) |
| + return result, (r-1,0) |
|
|
