Option to spare computations

inseq-team · Feb 27, 2024 · de4b304 · de4b304
1 parent def38cf
commit de4b304
Show file tree

Hide file tree

Showing 2 changed files with 32 additions and 20 deletions.
diff --git a/inseq/attr/feat/ops/value_zeroing.py b/inseq/attr/feat/ops/value_zeroing.py
@@ -312,6 +312,8 @@ def attribute(
         cross_zeroed_units_indices: Optional[OneOrMoreIndicesDict] = None,
         encoder_hidden_states: Optional[MultiLayerEmbeddingsTensor] = None,
         decoder_hidden_states: Optional[MultiLayerEmbeddingsTensor] = None,
+        output_decoder_self_scores: bool = True,
+        output_encoder_self_scores: bool = True,
     ) -> TensorOrTupleOfTensorsGeneric:
         """Perform attribution using the Value Zeroing method.
 
@@ -334,6 +336,11 @@ def attribute(
                 encoder-decoders models. Default: None.
             decoder_hidden_states (:obj:`torch.Tensor`, optional): A tensor of shape ``[batch_size, num_layers + 1,
                 target_seq_len, hidden_size]`` containing hidden states of the decoder.
+            output_decoder_self_scores (:obj:`bool`, optional): Whether to produce scores derived from zeroing the
+                decoder self-attention value vectors in encoder-decoder models. Cannot be false for decoder-only, or
+                if target-side attribution is requested using `attribute_target=True`. Default: True.
+            output_encoder_self_scores (:obj:`bool`, optional): Whether to produce scores derived from zeroing the
+                encoder self-attention value vectors in encoder-decoder models. Default: True.
 
         Returns:
             `TensorOrTupleOfTensorsGeneric`: Attribution outputs for source-only or source + target feature attribution
@@ -343,29 +350,32 @@ def attribute(
                 f"Similarity metric {similarity_metric} not available."
                 f"Available metrics: {','.join(self.SIMILARITY_METRICS.keys())}"
             )
-
-        decoder_scores = self.compute_modules_post_zeroing_similarity(
-            inputs=inputs,
-            additional_forward_args=additional_forward_args,
-            hidden_states=decoder_hidden_states,
-            attention_module_name=self.forward_func.config.self_attention_module,
-            similarity_metric=similarity_metric,
-            mode=ValueZeroingModule.DECODER.value,
-            zeroed_units_indices=decoder_zeroed_units_indices,
-            use_causal_mask=True,
-        )
-        # Encoder-decoder models also perform zeroing on the encoder self-attention and cross-attention values
-        # Adapted from https://github.com/hmohebbi/ContextMixingASR/blob/master/scoring/valueZeroing.py
-        if self.forward_func.is_encoder_decoder:
-            encoder_scores = self.compute_modules_post_zeroing_similarity(
+        decoder_scores = None
+        if not self.forward_func.is_encoder_decoder or output_decoder_self_scores or len(inputs) > 1:
+            decoder_scores = self.compute_modules_post_zeroing_similarity(
                 inputs=inputs,
                 additional_forward_args=additional_forward_args,
-                hidden_states=encoder_hidden_states,
+                hidden_states=decoder_hidden_states,
                 attention_module_name=self.forward_func.config.self_attention_module,
                 similarity_metric=similarity_metric,
-                mode=ValueZeroingModule.ENCODER.value,
-                zeroed_units_indices=encoder_zeroed_units_indices,
+                mode=ValueZeroingModule.DECODER.value,
+                zeroed_units_indices=decoder_zeroed_units_indices,
+                use_causal_mask=True,
             )
+        # Encoder-decoder models also perform zeroing on the encoder self-attention and cross-attention values
+        # Adapted from https://github.com/hmohebbi/ContextMixingASR/blob/master/scoring/valueZeroing.py
+        if self.forward_func.is_encoder_decoder:
+            encoder_scores = None
+            if output_encoder_self_scores:
+                encoder_scores = self.compute_modules_post_zeroing_similarity(
+                    inputs=inputs,
+                    additional_forward_args=additional_forward_args,
+                    hidden_states=encoder_hidden_states,
+                    attention_module_name=self.forward_func.config.self_attention_module,
+                    similarity_metric=similarity_metric,
+                    mode=ValueZeroingModule.ENCODER.value,
+                    zeroed_units_indices=encoder_zeroed_units_indices,
+                )
             cross_scores = self.compute_modules_post_zeroing_similarity(
                 inputs=inputs,
                 additional_forward_args=additional_forward_args,

diff --git a/inseq/attr/feat/perturbation_attribution.py b/inseq/attr/feat/perturbation_attribution.py
@@ -153,8 +153,10 @@ def attribute_step(
                 target_attributions = decoder_self_scores.to("cpu")
             else:
                 target_attributions = None
-                sequence_scores["decoder_self_scores"] = decoder_self_scores.to("cpu")
-            sequence_scores["encoder_self_scores"] = encoder_self_scores.to("cpu")
+                if decoder_self_scores is not None:
+                    sequence_scores["decoder_self_scores"] = decoder_self_scores.to("cpu")
+            if encoder_self_scores is not None:
+                sequence_scores["encoder_self_scores"] = encoder_self_scores.to("cpu")
             return MultiDimensionalFeatureAttributionStepOutput(
                 source_attributions=decoder_cross_scores.to("cpu"),
                 target_attributions=target_attributions,