removed torch ranking and n_threads

Author: blondered

examples/9_model_configs_and_saving.ipynb

@@ -1876,18 +1876,9 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 18,
+   "execution_count": 2,
    "metadata": {},
    "outputs": [
-    {
-     "name": "stderr",
-     "output_type": "stream",
-     "text": [
-      "GPU available: True (cuda), used: True\n",
-      "TPU available: False, using: 0 TPU cores\n",
-      "HPU available: False, using: 0 HPUs\n"
-     ]
-    },
     {
      "data": {
       "text/plain": [
@@ -1912,8 +1903,6 @@
        " 'deterministic': False,\n",
        " 'recommend_batch_size': 256,\n",
        " 'recommend_device': None,\n",
-       " 'recommend_n_threads': 0,\n",
-       " 'recommend_use_torch_ranking': True,\n",
        " 'train_min_user_interactions': 2,\n",
        " 'item_net_block_types': ['rectools.models.nn.item_net.IdEmbeddingsItemNet',\n",
        "  'rectools.models.nn.item_net.CatFeaturesItemNet'],\n",
@@ -1930,7 +1919,7 @@
        " 'lightning_module_kwargs': None}"
       ]
      },
-     "execution_count": 18,
+     "execution_count": 2,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -1962,18 +1951,9 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 19,
+   "execution_count": 3,
    "metadata": {},
    "outputs": [
-    {
-     "name": "stderr",
-     "output_type": "stream",
-     "text": [
-      "GPU available: True (cuda), used: True\n",
-      "TPU available: False, using: 0 TPU cores\n",
-      "HPU available: False, using: 0 HPUs\n"
-     ]
-    },
     {
      "data": {
       "text/plain": [
@@ -1998,8 +1978,6 @@
        " 'deterministic': False,\n",
        " 'recommend_batch_size': 256,\n",
        " 'recommend_device': None,\n",
-       " 'recommend_n_threads': 0,\n",
-       " 'recommend_use_torch_ranking': True,\n",
        " 'train_min_user_interactions': 2,\n",
        " 'item_net_block_types': ['rectools.models.nn.item_net.IdEmbeddingsItemNet',\n",
        "  'rectools.models.nn.item_net.CatFeaturesItemNet'],\n",
@@ -2016,7 +1994,7 @@
        " 'lightning_module_kwargs': None}"
       ]
      },
-     "execution_count": 19,
+     "execution_count": 3,
      "metadata": {},
      "output_type": "execute_result"
     }

examples/tutorials/transformers_advanced_training_guide.ipynb

@@ -1068,7 +1068,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 19,
+   "execution_count": 62,
    "metadata": {},
    "outputs": [
     {
@@ -1095,8 +1095,6 @@
        " 'deterministic': False,\n",
        " 'recommend_batch_size': 256,\n",
        " 'recommend_device': None,\n",
-       " 'recommend_n_threads': 0,\n",
-       " 'recommend_use_torch_ranking': True,\n",
        " 'train_min_user_interactions': 2,\n",
        " 'item_net_block_types': ['rectools.models.nn.item_net.IdEmbeddingsItemNet',\n",
        "  'rectools.models.nn.item_net.CatFeaturesItemNet'],\n",
@@ -1115,7 +1113,7 @@
        " 'mask_prob': 0.15}"
       ]
      },
-     "execution_count": 19,
+     "execution_count": 62,
      "metadata": {},
      "output_type": "execute_result"
     }

examples/tutorials/transformers_customization_guide.ipynb

@@ -177,8 +177,6 @@ class TransformerModelConfig(ModelConfig):
     deterministic: bool = False
     recommend_batch_size: int = 256
     recommend_torch_device: tp.Optional[str] = None
-    recommend_n_threads: int = 0
-    recommend_use_torch_ranking: bool = True
     train_min_user_interactions: int = 2
     item_net_block_types: ItemNetBlockTypes = (IdEmbeddingsItemNet, CatFeaturesItemNet)
     item_net_constructor_type: ItemNetConstructorType = SumOfEmbeddingsConstructor
@@ -234,8 +232,6 @@ def __init__(  # pylint: disable=too-many-arguments, too-many-locals
         deterministic: bool = False,
         recommend_batch_size: int = 256,
         recommend_torch_device: tp.Optional[str] = None,
-        recommend_n_threads: int = 0,
-        recommend_use_torch_ranking: bool = True,
         train_min_user_interactions: int = 2,
         item_net_block_types: tp.Sequence[tp.Type[ItemNetBase]] = (IdEmbeddingsItemNet, CatFeaturesItemNet),
         item_net_constructor_type: tp.Type[ItemNetConstructorBase] = SumOfEmbeddingsConstructor,
@@ -271,8 +267,6 @@ def __init__(  # pylint: disable=too-many-arguments, too-many-locals
         self.deterministic = deterministic
         self.recommend_batch_size = recommend_batch_size
         self.recommend_torch_device = recommend_torch_device
-        self.recommend_n_threads = recommend_n_threads
-        self.recommend_use_torch_ranking = recommend_use_torch_ranking
         self.train_min_user_interactions = train_min_user_interactions
         self.item_net_block_types = item_net_block_types
         self.item_net_constructor_type = item_net_constructor_type
@@ -452,8 +446,6 @@ def _recommend_u2i(
             k=k,
             filter_viewed=filter_viewed,
             dataset=dataset,
-            n_threads=self.recommend_n_threads,
-            use_torch_ranking=self.recommend_use_torch_ranking,
             torch_device=self.recommend_torch_device,
         )
 
@@ -471,8 +463,6 @@ def _recommend_i2i(
             target_ids=target_ids,
             sorted_item_ids_to_recommend=sorted_item_ids_to_recommend,
             k=k,
-            n_threads=self.recommend_n_threads,
-            use_torch_ranking=self.recommend_use_torch_ranking,
             torch_device=self.recommend_torch_device,
         )
 

rectools/models/nn/transformers/base.py

@@ -270,25 +270,10 @@ class BERT4RecModel(TransformerModelBase[BERT4RecModelConfig]):
         If you want to change this parameter after model is initialized,
         you can manually assign new value to model `recommend_batch_size` attribute.
     recommend_torch_device : {"cpu", "cuda", "cuda:0", ...}, default ``None``
-        String representation for `torch.device` used for torch model inference.
-        When `recommend_use_torch_ranking` is set to ``True`` (default) this device is also used
-        for items ranking while preparing recommendations using `TorchRanker`.
+        String representation for `torch.device` used for model inference.
         When set to ``None``, "cuda" will be used if it is available, "cpu" otherwise.
         If you want to change this parameter after model is initialized,
         you can manually assign new value to model `recommend_torch_device` attribute.
-    recommend_use_torch_ranking : bool, default ``True``
-        Use `TorchRanker` for items ranking while preparing recommendations.
-        When set to ``True`` (default), device specified in `recommend_torch_device` is used
-        for items ranking.
-        When set to ``False``, multi-threaded cpu ranking will be used with `ImplicitRanker`. You
-        can specify number of threads using `recommend_n_threads` argument.
-        If you want to change this parameter after model is initialized,
-        you can manually assign new value to model `recommend_use_torch_ranking` attribute.
-    recommend_n_threads : int, default 0
-        Number of threads to use for cpu items ranking with `ImplicitRanker`. Omitted if
-        `recommend_use_torch_ranking` is set to ``True`` (default).
-        If you want to change this parameter after model is initialized,
-        you can manually assign new value to model `recommend_n_threads` attribute.
     data_preparator_kwargs: optional(dict), default ``None``
         Additional keyword arguments to pass during `data_preparator_type` initialization.
         Make sure all dict values have JSON serializable types.

rectools/models/nn/transformers/bert4rec.py

@@ -23,7 +23,7 @@
 from rectools import ExternalIds
 from rectools.dataset.dataset import Dataset, DatasetSchemaDict
 from rectools.models.base import InternalRecoTriplet
-from rectools.models.rank import Distance, ImplicitRanker, Ranker, TorchRanker
+from rectools.models.rank import Distance, TorchRanker
 from rectools.types import InternalIdsArray
 
 from .data_preparator import TransformerDataPreparatorBase
@@ -113,8 +113,6 @@ def _recommend_u2i(
         k: int,
         dataset: Dataset,  # [n_rec_users x n_items + n_item_extra_tokens]
         filter_viewed: bool,
-        use_torch_ranking: bool,
-        n_threads: int,
         torch_device: tp.Optional[str],
         *args: tp.Any,
         **kwargs: tp.Any,
@@ -127,8 +125,6 @@ def _recommend_i2i(
         target_ids: InternalIdsArray,
         sorted_item_ids_to_recommend: InternalIdsArray,
         k: int,
-        use_torch_ranking: bool,
-        n_threads: int,
         torch_device: tp.Optional[str],
         *args: tp.Any,
         **kwargs: tp.Any,
@@ -331,8 +327,6 @@ def _recommend_u2i(
         k: int,
         dataset: Dataset,  # [n_rec_users x n_items + n_item_extra_tokens]
         filter_viewed: bool,
-        use_torch_ranking: bool,
-        n_threads: int,
         torch_device: tp.Optional[str],
     ) -> InternalRecoTriplet:
         """Recommend to users."""
@@ -342,26 +336,12 @@ def _recommend_u2i(
 
         user_embs, item_embs = self._get_user_item_embeddings(recommend_dataloader, torch_device)
 
-        ranker: Ranker
-        if use_torch_ranking:
-            ranker = TorchRanker(
-                distance=Distance.DOT,
-                device=item_embs.device,
-                subjects_factors=user_embs[user_ids],
-                objects_factors=item_embs,
-            )
-        else:
-            user_embs_np = user_embs.detach().cpu().numpy()
-            item_embs_np = item_embs.detach().cpu().numpy()
-            del user_embs, item_embs
-            torch.cuda.empty_cache()
-            ranker = ImplicitRanker(
-                Distance.DOT,
-                user_embs_np[user_ids],  # [n_rec_users, n_factors]
-                item_embs_np,  # [n_items + n_item_extra_tokens, n_factors]
-                num_threads=n_threads,
-                use_gpu=False,
-            )
+        ranker = TorchRanker(
+            distance=Distance.DOT,
+            device=item_embs.device,
+            subjects_factors=user_embs[user_ids],
+            objects_factors=item_embs,
+        )
 
         user_ids_indices, all_reco_ids, all_scores = ranker.rank(
             subject_ids=np.arange(len(user_ids)),  # n_rec_users
@@ -377,31 +357,16 @@ def _recommend_i2i(
         target_ids: InternalIdsArray,
         sorted_item_ids_to_recommend: InternalIdsArray,
         k: int,
-        use_torch_ranking: bool,
-        n_threads: int,
         torch_device: tp.Optional[str],
     ) -> InternalRecoTriplet:
         """Recommend to items."""
         self._prepare_for_inference(torch_device)
         with torch.no_grad():
             item_embs = self.torch_model.item_model.get_all_embeddings()
 
-        ranker: Ranker
-        if use_torch_ranking:
-            ranker = TorchRanker(
-                distance=self.i2i_dist, device=item_embs.device, subjects_factors=item_embs, objects_factors=item_embs
-            )
-        else:
-            item_embs_np = item_embs.detach().cpu().numpy()
-            del item_embs
-            ranker = ImplicitRanker(
-                self.i2i_dist,
-                item_embs_np,  # [n_items + n_item_extra_tokens, n_factors]
-                item_embs_np,  # [n_items + n_item_extra_tokens, n_factors]
-                num_threads=n_threads,
-                use_gpu=False,
-            )
-
+        ranker = TorchRanker(
+            distance=self.i2i_dist, device=item_embs.device, subjects_factors=item_embs, objects_factors=item_embs
+        )
         torch.cuda.empty_cache()
         return ranker.rank(
             subject_ids=target_ids,  # model internal

rectools/models/nn/transformers/lightning.py

@@ -350,25 +350,10 @@ class SASRecModel(TransformerModelBase[SASRecModelConfig]):
         If you want to change this parameter after model is initialized,
         you can manually assign new value to model `recommend_batch_size` attribute.
     recommend_torch_device : {"cpu", "cuda", "cuda:0", ...}, default ``None``
-        String representation for `torch.device` used for torch model inference.
-        When `recommend_use_torch_ranking` is set to ``True`` (default) this device is also used
-        for items ranking while preparing recommendations using `TorchRanker`.
+        String representation for `torch.device` used for model inference.
         When set to ``None``, "cuda" will be used if it is available, "cpu" otherwise.
         If you want to change this parameter after model is initialized,
         you can manually assign new value to model `recommend_torch_device` attribute.
-    recommend_use_torch_ranking : bool, default ``True``
-        Use `TorchRanker` for items ranking while preparing recommendations.
-        When set to ``True`` (default), device specified in `recommend_torch_device` is used
-        for items ranking.
-        When set to ``False``, multi-threaded cpu ranking will be used with `ImplicitRanker`. You
-        can specify number of threads using `recommend_n_threads` argument.
-        If you want to change this parameter after model is initialized,
-        you can manually assign new value to model `recommend_use_torch_ranking` attribute.
-    recommend_n_threads : int, default 0
-        Number of threads to use for cpu items ranking with `ImplicitRanker`. Omitted if
-        `recommend_use_torch_ranking` is set to ``True`` (default).
-        If you want to change this parameter after model is initialized,
-        you can manually assign new value to model `recommend_n_threads` attribute.
     data_preparator_kwargs: optional(dict), default ``None``
         Additional keyword arguments to pass during `data_preparator_type` initialization.
         Make sure all dict values have JSON serializable types.