[WhisperForCausalLM] Add WhisperForCausalLM for speculative decoding (#27195)

* finish

* add tests

* fix all tests

* [Assistant Decoding] Add test

* fix more

* better

* finish

* Apply suggestions from code review

Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>

* finish

---------

Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>

docs/source/en/model_doc/whisper.md

@@ -88,6 +88,11 @@ The original code can be found [here](https://github.com/openai/whisper).
     - forward
     - generate
 
+## WhisperForCausalLM
+
+[[autodoc]] WhisperForCausalLM
+    - forward
+
 ## WhisperForAudioClassification
 
 [[autodoc]] WhisperForAudioClassification

docs/source/en/tasks/language_modeling.md

@@ -37,7 +37,7 @@ You can finetune other architectures for causal language modeling following the
 Choose one of the following architectures:
 
 <!--This tip is automatically generated by `make fix-copies`, do not fill manually!-->
-[BART](../model_doc/bart), [BERT](../model_doc/bert), [Bert Generation](../model_doc/bert-generation), [BigBird](../model_doc/big_bird), [BigBird-Pegasus](../model_doc/bigbird_pegasus), [BioGpt](../model_doc/biogpt), [Blenderbot](../model_doc/blenderbot), [BlenderbotSmall](../model_doc/blenderbot-small), [BLOOM](../model_doc/bloom), [CamemBERT](../model_doc/camembert), [CodeLlama](../model_doc/code_llama), [CodeGen](../model_doc/codegen), [CPM-Ant](../model_doc/cpmant), [CTRL](../model_doc/ctrl), [Data2VecText](../model_doc/data2vec-text), [ELECTRA](../model_doc/electra), [ERNIE](../model_doc/ernie), [Falcon](../model_doc/falcon), [Fuyu](../model_doc/fuyu), [GIT](../model_doc/git), [GPT-Sw3](../model_doc/gpt-sw3), [OpenAI GPT-2](../model_doc/gpt2), [GPTBigCode](../model_doc/gpt_bigcode), [GPT Neo](../model_doc/gpt_neo), [GPT NeoX](../model_doc/gpt_neox), [GPT NeoX Japanese](../model_doc/gpt_neox_japanese), [GPT-J](../model_doc/gptj), [LLaMA](../model_doc/llama), [Marian](../model_doc/marian), [mBART](../model_doc/mbart), [MEGA](../model_doc/mega), [Megatron-BERT](../model_doc/megatron-bert), [Mistral](../model_doc/mistral), [MPT](../model_doc/mpt), [MusicGen](../model_doc/musicgen), [MVP](../model_doc/mvp), [OpenLlama](../model_doc/open-llama), [OpenAI GPT](../model_doc/openai-gpt), [OPT](../model_doc/opt), [Pegasus](../model_doc/pegasus), [Persimmon](../model_doc/persimmon), [PLBart](../model_doc/plbart), [ProphetNet](../model_doc/prophetnet), [QDQBert](../model_doc/qdqbert), [Reformer](../model_doc/reformer), [RemBERT](../model_doc/rembert), [RoBERTa](../model_doc/roberta), [RoBERTa-PreLayerNorm](../model_doc/roberta-prelayernorm), [RoCBert](../model_doc/roc_bert), [RoFormer](../model_doc/roformer), [RWKV](../model_doc/rwkv), [Speech2Text2](../model_doc/speech_to_text_2), [Transformer-XL](../model_doc/transfo-xl), [TrOCR](../model_doc/trocr), [XGLM](../model_doc/xglm), [XLM](../model_doc/xlm), [XLM-ProphetNet](../model_doc/xlm-prophetnet), [XLM-RoBERTa](../model_doc/xlm-roberta), [XLM-RoBERTa-XL](../model_doc/xlm-roberta-xl), [XLNet](../model_doc/xlnet), [X-MOD](../model_doc/xmod)
+[BART](../model_doc/bart), [BERT](../model_doc/bert), [Bert Generation](../model_doc/bert-generation), [BigBird](../model_doc/big_bird), [BigBird-Pegasus](../model_doc/bigbird_pegasus), [BioGpt](../model_doc/biogpt), [Blenderbot](../model_doc/blenderbot), [BlenderbotSmall](../model_doc/blenderbot-small), [BLOOM](../model_doc/bloom), [CamemBERT](../model_doc/camembert), [CodeLlama](../model_doc/code_llama), [CodeGen](../model_doc/codegen), [CPM-Ant](../model_doc/cpmant), [CTRL](../model_doc/ctrl), [Data2VecText](../model_doc/data2vec-text), [ELECTRA](../model_doc/electra), [ERNIE](../model_doc/ernie), [Falcon](../model_doc/falcon), [Fuyu](../model_doc/fuyu), [GIT](../model_doc/git), [GPT-Sw3](../model_doc/gpt-sw3), [OpenAI GPT-2](../model_doc/gpt2), [GPTBigCode](../model_doc/gpt_bigcode), [GPT Neo](../model_doc/gpt_neo), [GPT NeoX](../model_doc/gpt_neox), [GPT NeoX Japanese](../model_doc/gpt_neox_japanese), [GPT-J](../model_doc/gptj), [LLaMA](../model_doc/llama), [Marian](../model_doc/marian), [mBART](../model_doc/mbart), [MEGA](../model_doc/mega), [Megatron-BERT](../model_doc/megatron-bert), [Mistral](../model_doc/mistral), [MPT](../model_doc/mpt), [MusicGen](../model_doc/musicgen), [MVP](../model_doc/mvp), [OpenLlama](../model_doc/open-llama), [OpenAI GPT](../model_doc/openai-gpt), [OPT](../model_doc/opt), [Pegasus](../model_doc/pegasus), [Persimmon](../model_doc/persimmon), [PLBart](../model_doc/plbart), [ProphetNet](../model_doc/prophetnet), [QDQBert](../model_doc/qdqbert), [Reformer](../model_doc/reformer), [RemBERT](../model_doc/rembert), [RoBERTa](../model_doc/roberta), [RoBERTa-PreLayerNorm](../model_doc/roberta-prelayernorm), [RoCBert](../model_doc/roc_bert), [RoFormer](../model_doc/roformer), [RWKV](../model_doc/rwkv), [Speech2Text2](../model_doc/speech_to_text_2), [Transformer-XL](../model_doc/transfo-xl), [TrOCR](../model_doc/trocr), [Whisper](../model_doc/whisper), [XGLM](../model_doc/xglm), [XLM](../model_doc/xlm), [XLM-ProphetNet](../model_doc/xlm-prophetnet), [XLM-RoBERTa](../model_doc/xlm-roberta), [XLM-RoBERTa-XL](../model_doc/xlm-roberta-xl), [XLNet](../model_doc/xlnet), [X-MOD](../model_doc/xmod)
 
 
 

src/transformers/__init__.py

@@ -3090,6 +3090,7 @@ else:
         [
             "WHISPER_PRETRAINED_MODEL_ARCHIVE_LIST",
             "WhisperForAudioClassification",
+            "WhisperForCausalLM",
             "WhisperForConditionalGeneration",
             "WhisperModel",
             "WhisperPreTrainedModel",
@@ -6845,6 +6846,7 @@ if TYPE_CHECKING:
         from .models.whisper import (
             WHISPER_PRETRAINED_MODEL_ARCHIVE_LIST,
             WhisperForAudioClassification,
+            WhisperForCausalLM,
             WhisperForConditionalGeneration,
             WhisperModel,
             WhisperPreTrainedModel,

src/transformers/generation/utils.py

@@ -1626,6 +1626,10 @@ class GenerationMixin:
             if not model_kwargs["use_cache"]:
                 raise ValueError("assisted generate requires `use_cache=True`")
 
+            assistant_accepts_encoder_outputs = "encoder_outputs" in set(
+                inspect.signature(assistant_model.forward).parameters.keys()
+            )
+
             # 11. If the assistant model is an encoder-decoder, prepare its encoder outputs
             if assistant_model.config.is_encoder_decoder and "assistant_encoder_outputs" not in model_kwargs:
                 assistant_model_kwargs = copy.deepcopy(model_kwargs)
@@ -1637,6 +1641,17 @@ class GenerationMixin:
                 )
                 model_kwargs["assistant_encoder_outputs"] = assistant_model_kwargs["encoder_outputs"]
 
+            if (
+                not assistant_model.config.is_encoder_decoder
+                and assistant_accepts_encoder_outputs
+                and "encoder_outputs" in model_kwargs
+            ):
+                # some assistants might be assymetric (many more enc layers than dec layers)
+                # encoder-decoder models that share the exact same encoder as the teacher
+                # in this case the assistant only needs to load the light-weight decoder,
+                # but still requires `encoder_outputs` to be passed
+                model_kwargs["assistant_encoder_outputs"] = model_kwargs["encoder_outputs"]
+
             # 12. run assisted generate
             return self.assisted_decoding(
                 input_ids,
@@ -4368,6 +4383,11 @@ class GenerationMixin:
         else:
             num_assistant_tokens = assistant_model.generation_config.num_assistant_tokens
 
+        # check if assistant model accepts encoder_outputs
+        assistant_accepts_encoder_outputs = "encoder_outputs" in set(
+            inspect.signature(assistant_model.forward).parameters.keys()
+        )
+
         # init values
         logits_processor = logits_processor if logits_processor is not None else LogitsProcessorList()
         logits_warper = logits_warper if logits_warper is not None else LogitsProcessorList()
@@ -4454,9 +4474,13 @@ class GenerationMixin:
                             encoder_outputs=model_kwargs["assistant_encoder_outputs"],
                         )
                     else:
+                        encoder_kwargs = {}
+
+                        if assistant_accepts_encoder_outputs and "assistant_encoder_outputs" in model_kwargs:
+                            encoder_kwargs["encoder_outputs"] = model_kwargs["assistant_encoder_outputs"]
+
                         assistant_model_outputs = assistant_model(
-                            assist_inputs,
+                            assist_inputs, past_key_values=model_kwargs["assistant_past_key_values"], **encoder_kwargs
-                            past_key_values=model_kwargs["assistant_past_key_values"],
                         )
                 else:
                     if assistant_model.config.is_encoder_decoder:
@@ -4465,7 +4489,12 @@ class GenerationMixin:
                             encoder_outputs=model_kwargs["assistant_encoder_outputs"],
                         )
                     else:
-                        assistant_model_outputs = assistant_model(candidate_input_ids)
+                        encoder_kwargs = {}
+
+                        if assistant_accepts_encoder_outputs and "assistant_encoder_outputs" in model_kwargs:
+                            encoder_kwargs["encoder_outputs"] = model_kwargs["assistant_encoder_outputs"]
+
+                        assistant_model_outputs = assistant_model(candidate_input_ids, **encoder_kwargs)
 
                 # 1.2. greedily select the next candidate token
                 model_kwargs["assistant_past_key_values"] = assistant_model_outputs.past_key_values

src/transformers/models/auto/modeling_auto.py

@@ -438,6 +438,7 @@ MODEL_FOR_CAUSAL_LM_MAPPING_NAMES = OrderedDict(
         ("speech_to_text_2", "Speech2Text2ForCausalLM"),
         ("transfo-xl", "TransfoXLLMHeadModel"),
         ("trocr", "TrOCRForCausalLM"),
+        ("whisper", "WhisperForCausalLM"),
         ("xglm", "XGLMForCausalLM"),
         ("xlm", "XLMWithLMHeadModel"),
         ("xlm-prophetnet", "XLMProphetNetForCausalLM"),

src/transformers/models/whisper/__init__.py

@@ -46,6 +46,7 @@ except OptionalDependencyNotAvailable:
 else:
     _import_structure["modeling_whisper"] = [
         "WHISPER_PRETRAINED_MODEL_ARCHIVE_LIST",
+        "WhisperForCausalLM",
         "WhisperForConditionalGeneration",
         "WhisperModel",
         "WhisperPreTrainedModel",
@@ -102,6 +103,7 @@ if TYPE_CHECKING:
         from .modeling_whisper import (
             WHISPER_PRETRAINED_MODEL_ARCHIVE_LIST,
             WhisperForAudioClassification,
+            WhisperForCausalLM,
             WhisperForConditionalGeneration,
             WhisperModel,
             WhisperPreTrainedModel,

src/transformers/models/whisper/modeling_whisper.py

@@ -29,6 +29,7 @@ from ...modeling_attn_mask_utils import _prepare_4d_causal_attention_mask
 from ...modeling_outputs import (
     BaseModelOutput,
     BaseModelOutputWithPastAndCrossAttentions,
+    CausalLMOutputWithCrossAttentions,
     Seq2SeqLMOutput,
     Seq2SeqModelOutput,
     SequenceClassifierOutput,
@@ -945,6 +946,8 @@ class WhisperDecoder(WhisperPreTrainedModel):
         config: WhisperConfig
     """
 
+    main_input_name = "input_ids"
+
     def __init__(self, config: WhisperConfig):
         super().__init__(config)
         self.dropout = config.dropout
@@ -1028,7 +1031,8 @@ class WhisperDecoder(WhisperPreTrainedModel):
 
                 If `past_key_values` are used, the user can optionally input only the last `decoder_input_ids` (those
                 that don't have their past key value states given to this model) of shape `(batch_size, 1)` instead of
-                all `decoder_input_ids` of shape `(batch_size, sequence_length)`. inputs_embeds (`torch.FloatTensor` of
+                all `decoder_input_ids` of shape `(batch_size, sequence_length)`.
+            inputs_embeds (`torch.FloatTensor` of
                 shape `(batch_size, sequence_length, hidden_size)`, *optional*): Optionally, instead of passing
                 `input_ids` you can choose to directly pass an embedded representation. This is useful if you want more
                 control over how to convert `input_ids` indices into associated vectors than the model's internal
@@ -1811,6 +1815,247 @@ class WhisperForConditionalGeneration(WhisperPreTrainedModel):
         return timestamps
 
 
+class WhisperDecoderWrapper(WhisperPreTrainedModel):
+    """
+    This wrapper class is a helper class to correctly load pretrained checkpoints when the causal language model is
+    used in combination with the [`EncoderDecoderModel`] framework.
+    """
+
+    def __init__(self, config):
+        super().__init__(config)
+        config.is_encoder_decoder = False
+        self.decoder = WhisperDecoder(config)
+
+    def get_input_embeddings(self):
+        return self.decoder.embed_tokens
+
+    def set_input_embeddings(self, value):
+        self.decoder.embed_tokens = value
+
+    def forward(self, *args, **kwargs):
+        return self.decoder(*args, **kwargs)
+
+
+@add_start_docstrings(
+    """
+    Whisper decoder with with a language modeling head on top (linear layer with weights tied to the input embeddings).
+    """,
+    WHISPER_START_DOCSTRING,
+)
+class WhisperForCausalLM(WhisperPreTrainedModel):
+    _tied_weights_keys = ["proj_out.weight"]
+    main_input_name = "input_ids"
+
+    def __init__(self, config):
+        super().__init__(config)
+        config.is_encoder_decoder = False
+        self.model = WhisperDecoderWrapper(config)
+
+        self.proj_out = nn.Linear(config.hidden_size, config.vocab_size, bias=False)
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def get_output_embeddings(self):
+        return self.proj_out
+
+    def set_output_embeddings(self, new_embeddings):
+        self.proj_out = new_embeddings
+
+    def get_input_embeddings(self) -> nn.Module:
+        return self.model.get_input_embeddings()
+
+    def set_input_embeddings(self, value):
+        self.model.set_input_embeddings(value)
+
+    def set_decoder(self, decoder):
+        self.model.decoder = decoder
+
+    def get_decoder(self):
+        return self.model.decoder
+
+    @replace_return_docstrings(output_type=CausalLMOutputWithCrossAttentions, config_class=_CONFIG_FOR_DOC)
+    def forward(
+        self,
+        input_ids: torch.LongTensor = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        encoder_outputs: Optional[Tuple[torch.FloatTensor]] = None,
+        head_mask: Optional[torch.Tensor] = None,
+        cross_attn_head_mask: Optional[torch.Tensor] = None,
+        past_key_values: Optional[Tuple[Tuple[torch.FloatTensor]]] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        labels: Optional[torch.LongTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[Tuple, CausalLMOutputWithCrossAttentions]:
+        r"""
+        Args:
+            input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`):
+                Indices of input sequence tokens in the vocabulary. Padding will be ignored by default should you
+                provide it. Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and
+                [`PreTrainedTokenizer.__call__`] for details. [What are input IDs?](../glossary#input-ids)
+            attention_mask (`torch.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
+                Mask to avoid performing attention on padding token indices. Mask values selected in `[0, 1]`:
+                - 1 for tokens that are **not masked**,
+                - 0 for tokens that are **masked**.
+                [What are attention masks?](../glossary#attention-mask)
+            encoder_outputs  (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`, *optional*):
+                Sequence of hidden-states at the output of the last layer of the encoder. Used in the cross-attention
+                if the model is configured as a decoder.
+            head_mask (`torch.Tensor` of shape `(decoder_layers, decoder_attention_heads)`, *optional*):
+                Mask to nullify selected heads of the attention modules. Mask values selected in `[0, 1]`:
+                - 1 indicates the head is **not masked**,
+                - 0 indicates the head is **masked**.
+            cross_attn_head_mask (`torch.Tensor` of shape `(decoder_layers, decoder_attention_heads)`, *optional*):
+                Mask to nullify selected heads of the cross-attention modules. Mask values selected in `[0, 1]`:
+                - 1 indicates the head is **not masked**,
+                - 0 indicates the head is **masked**.
+            past_key_values (`tuple(tuple(torch.FloatTensor))`, *optional*, returned when `use_cache=True` is passed or when `config.use_cache=True`):
+                Tuple of `tuple(torch.FloatTensor)` of length `config.n_layers`, with each tuple having 2 tensors of
+                shape `(batch_size, num_heads, sequence_length, embed_size_per_head)`) and 2 additional tensors of
+                shape `(batch_size, num_heads, encoder_sequence_length, embed_size_per_head)`. The two additional
+                tensors are only required when the model is used as a decoder in a Sequence to Sequence model. Contains
+                pre-computed hidden-states (key and values in the self-attention blocks and in the cross-attention
+                blocks) that can be used (see `past_key_values` input) to speed up sequential decoding. If
+                `past_key_values` are used, the user can optionally input only the last `decoder_input_ids` (those that
+                don't have their past key value states given to this model) of shape `(batch_size, 1)` instead of all
+                `decoder_input_ids` of shape `(batch_size, sequence_length)`.
+            inputs_embeds (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`, *optional*):
+                Optionally, instead of passing `input_ids` you can choose to directly pass an embedded representation.
+                This is useful if you want more control over how to convert `input_ids` indices into associated vectors
+                than the model's internal embedding lookup matrix.
+            labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
+                Labels for computing the masked language modeling loss. Indices should either be in `[0, ...,
+                config.vocab_size]` or -100 (see `input_ids` docstring). Tokens with indices set to `-100` are ignored
+                (masked), the loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`.
+            use_cache (`bool`, *optional*):
+                If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding
+                (see `past_key_values`).
+                - 1 for tokens that are **not masked**,
+                - 0 for tokens that are **masked**.
+            output_attentions (`bool`, *optional*):
+                Whether or not to return the attentions tensors of all attention layers. See `attentions` under
+                returned tensors for more detail.
+            output_hidden_states (`bool`, *optional*):
+                Whether or not to return the hidden states of all layers. See `hidden_states` under returned tensors
+                for more detail.
+            return_dict (`bool`, *optional*):
+                Whether or not to return a [`~utils.ModelOutput`] instead of a plain tuple.
+
+        Returns:
+
+        Example:
+
+        ```python
+        >>> from transformers import WhisperForCausalLM, WhisperForConditionalGeneration, WhisperProcessor
+        >>> import torch
+        >>> from datasets import load_dataset
+
+        >>> processor = WhisperProcessor.from_pretrained("openai/whisper-large-v2")
+        >>> model = WhisperForConditionalGeneration.from_pretrained("openai/whisper-large-v2")
+
+        >>> assistant_model = WhisperForCausalLM.from_pretrained("distil-whisper/distil-large-v2")
+
+        >>> ds = load_dataset("hf-internal-testing/librispeech_asr_dummy", "clean", split="validation")
+        >>> sample = ds[0]["audio"]
+        >>> input_features = processor(
+        ...     sample["array"], sampling_rate=sample["sampling_rate"], return_tensors="pt"
+        ... ).input_features
+
+        >>> predicted_ids = model.generate(input_features, assistant_model=assistant_model)
+
+        >>> # decode token ids to text
+        >>> transcription = processor.batch_decode(predicted_ids, skip_special_tokens=True)
+        >>> transcription
+        ' Mr. Quilter is the apostle of the middle classes and we are glad to welcome his gospel.'
+        ```"""
+        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
+        output_hidden_states = (
+            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
+        )
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        # If the user passed a tuple or `BaseModelOutput` for encoder_outputs, we extract only the hidden states
+        if isinstance(encoder_outputs, (BaseModelOutput, tuple, list)):
+            encoder_outputs = encoder_outputs[0]
+
+        # decoder outputs consists of (dec_features, layer_state, dec_hidden, dec_attn)
+        outputs = self.model.decoder(
+            input_ids=input_ids,
+            attention_mask=attention_mask,
+            encoder_hidden_states=encoder_outputs,
+            head_mask=head_mask,
+            cross_attn_head_mask=cross_attn_head_mask,
+            past_key_values=past_key_values,
+            inputs_embeds=inputs_embeds,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+
+        logits = self.proj_out(outputs[0])
+
+        loss = None
+        if labels is not None:
+            labels = labels.to(logits.device)
+            loss_fct = CrossEntropyLoss()
+            loss = loss_fct(logits.view(-1, self.config.vocab_size), labels.view(-1))
+
+        if not return_dict:
+            output = (logits,) + outputs[1:]
+            return (loss,) + output if loss is not None else output
+
+        return CausalLMOutputWithCrossAttentions(
+            loss=loss,
+            logits=logits,
+            past_key_values=outputs.past_key_values,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+            cross_attentions=outputs.cross_attentions,
+        )
+
+    def prepare_inputs_for_generation(
+        self,
+        input_ids,
+        past_key_values=None,
+        use_cache=None,
+        encoder_outputs=None,
+        attention_mask=None,
+        **kwargs,
+    ):
+        if past_key_values is not None:
+            past_length = past_key_values[0][0].shape[2]
+
+            # Some generation methods already pass only the last input ID
+            if input_ids.shape[1] > past_length:
+                remove_prefix_length = past_length
+            else:
+                # Default to old behavior: keep only final ID
+                remove_prefix_length = input_ids.shape[1] - 1
+
+            input_ids = input_ids[:, remove_prefix_length:]
+
+        return {
+            "encoder_outputs": encoder_outputs,
+            "past_key_values": past_key_values,
+            "input_ids": input_ids,
+            "use_cache": use_cache,
+            "attention_mask": attention_mask,
+        }
+
+    @staticmethod
+    def _reorder_cache(past_key_values, beam_idx):
+        reordered_past = ()
+        for layer_past in past_key_values:
+            reordered_past += (
+                tuple(past_state.index_select(0, beam_idx.to(past_state.device)) for past_state in layer_past),
+            )
+        return reordered_past
+
+
 @add_start_docstrings(
     """
     Whisper Encoder Model with a sequence classification head on top (a linear layer over the pooled output) for tasks

src/transformers/utils/dummy_pt_objects.py

@@ -8413,6 +8413,13 @@ class WhisperForAudioClassification(metaclass=DummyObject):
         requires_backends(self, ["torch"])
 
 
+class WhisperForCausalLM(metaclass=DummyObject):
+    _backends = ["torch"]
+
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["torch"])
+
+
 class WhisperForConditionalGeneration(metaclass=DummyObject):
     _backends = ["torch"]
 

tests/generation/test_utils.py

@@ -43,6 +43,7 @@ if is_torch_available():
         AutoModelForSpeechSeq2Seq,
         AutoModelForVision2Seq,
         AutoTokenizer,
+        BartForCausalLM,
         BartForConditionalGeneration,
         BartTokenizer,
         GPT2LMHeadModel,
@@ -3010,3 +3011,63 @@ class GenerationIntegrationTests(unittest.TestCase, GenerationIntegrationTestsMi
             assistant_encoder_outputs=encoder_outputs,
         )
         self.assertListEqual(outputs_assisted.tolist(), outputs_foo.tolist())
+
+    def test_assisted_decoding_encoder_decoder_shared_encoder(self):
+        # PT-only test: TF doesn't support assisted decoding yet.
+        # Bart subclass with a kwarg called foo that distorts the output
+        class FakeBart(BartForConditionalGeneration):
+            def forward(self, input_ids, foo=False, **kwargs):
+                outs = super().forward(input_ids, **kwargs)
+
+                if foo:
+                    outs["logits"][:, :, :] = 0.0
+
+                return outs
+
+            def prepare_inputs_for_generation(self, *args, foo=False, encoder_outputs=None, **kwargs):
+                kwargs["encoder_outputs"] = encoder_outputs
+                inputs = super().prepare_inputs_for_generation(*args, **kwargs)
+
+                inputs["foo"] = foo
+                return inputs
+
+        model = FakeBart.from_pretrained("hf-internal-testing/tiny-random-BartForConditionalGeneration").to(
+            torch_device
+        )
+        tokenizer = AutoTokenizer.from_pretrained("hf-internal-testing/tiny-random-BartForConditionalGeneration")
+
+        text = "Hello world"
+        tokenized_inputs = tokenizer([text], return_tensors="pt")
+        input_ids = tokenized_inputs.input_ids.to(torch_device)
+
+        # Traditional way of generating text
+        outputs_normal = model.generate(input_ids)
+        self.assertEqual(outputs_normal.shape, (1, 20))
+
+        # Should be different with foo
+        outputs_foo = model.generate(input_ids, foo=True)
+        with self.assertRaises(AssertionError):
+            self.assertListEqual(outputs_foo.tolist(), outputs_normal.tolist())
+
+        # Assistant model
+        assistant = BartForCausalLM.from_pretrained("hf-internal-testing/tiny-random-BartForConditionalGeneration").to(
+            torch_device
+        )
+
+        # If assisted generation passes model_kwargs correctly, should be same as previous
+        outputs_assisted = model.generate(
+            input_ids,
+            foo=True,
+            assistant_model=assistant,
+        )
+        self.assertListEqual(outputs_assisted.tolist(), outputs_foo.tolist())
+
+        # Check that passing encoder_outputs directly also works as expected
+        encoder_outputs = model.get_encoder()(input_ids)
+
+        outputs_assisted = model.generate(
+            foo=True,
+            assistant_model=assistant,
+            encoder_outputs=encoder_outputs,
+        )
+        self.assertListEqual(outputs_assisted.tolist(), outputs_foo.tolist())

tests/models/whisper/test_modeling_whisper.py

@@ -51,6 +51,7 @@ if is_torch_available():
     from transformers import (
         WhisperFeatureExtractor,
         WhisperForAudioClassification,
+        WhisperForCausalLM,
         WhisperForConditionalGeneration,
         WhisperModel,
         WhisperProcessor,
@@ -1990,3 +1991,246 @@ class WhisperEncoderModelTest(ModelTesterMixin, GenerationTesterMixin, unittest.
 
                 self.assertEqual(fx_keys, pt_keys)
                 self.check_pt_flax_outputs(fx_outputs, pt_outputs_loaded, model_class)
+
+
+class WhisperStandaloneDecoderModelTester:
+    def __init__(
+        self,
+        parent,
+        batch_size=2,
+        is_training=True,
+        use_labels=False,
+        vocab_size=200,
+        hidden_size=16,
+        num_hidden_layers=2,
+        num_attention_heads=4,
+        input_channels=1,
+        hidden_act="gelu",
+        hidden_dropout_prob=0.1,
+        attention_probs_dropout_prob=0.1,
+        max_position_embeddings=20,
+        max_source_positions=30,
+        max_target_positions=40,
+        bos_token_id=98,
+        eos_token_id=98,
+        pad_token_id=0,
+        num_mel_bins=80,
+        decoder_start_token_id=85,
+        num_conv_layers=1,
+        suppress_tokens=None,
+        begin_suppress_tokens=None,
+    ):
+        self.parent = parent
+        self.batch_size = batch_size
+        self.is_training = is_training
+        self.use_labels = use_labels
+        self.vocab_size = vocab_size
+        self.hidden_size = hidden_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.input_channels = input_channels
+        self.hidden_act = hidden_act
+        self.hidden_dropout_prob = hidden_dropout_prob
+        self.attention_probs_dropout_prob = attention_probs_dropout_prob
+        self.num_mel_bins = num_mel_bins
+        self.max_position_embeddings = max_position_embeddings
+        self.max_source_positions = max_source_positions
+        self.max_target_positions = max_target_positions
+        self.eos_token_id = eos_token_id
+        self.pad_token_id = pad_token_id
+        self.bos_token_id = bos_token_id
+        self.decoder_start_token_id = decoder_start_token_id
+        self.num_conv_layers = num_conv_layers
+        self.suppress_tokens = suppress_tokens
+        self.begin_suppress_tokens = begin_suppress_tokens
+
+    def prepare_config_and_inputs(self):
+        input_features = floats_tensor([self.batch_size, self.num_mel_bins, self.seq_length], self.vocab_size)
+
+        decoder_input_ids = torch.tensor(
+            self.batch_size * [[self.decoder_start_token_id, 3, 3, 7, 2]], device=torch_device
+        )
+
+        config = self.get_config()
+        config.is_encoder_decoder = False
+        inputs_dict = prepare_whisper_inputs_dict(
+            config,
+            attention_mask=None,
+            input_features=input_features,
+            decoder_input_ids=decoder_input_ids,
+        )
+
+        inputs_dict.pop("input_features")
+        inputs_dict.pop("head_mask")
+        inputs_dict.pop("decoder_head_mask")
+        inputs_dict.pop("cross_attn_head_mask")
+
+        inputs_dict["attention_mask"] = inputs_dict.pop("decoder_attention_mask")
+        inputs_dict["input_ids"] = inputs_dict.pop("decoder_input_ids")
+        return config, inputs_dict
+
+    @property
+    def encoder_seq_length(self):
+        return 5
+
+    @property
+    def seq_length(self):
+        return 5
+
+    def get_config(self):
+        return WhisperConfig(
+            vocab_size=self.vocab_size,
+            d_model=self.hidden_size,
+            encoder_layers=self.num_hidden_layers,
+            decoder_layers=self.num_hidden_layers,
+            encoder_attention_heads=self.num_attention_heads,
+            decoder_attention_heads=self.num_attention_heads,
+            input_channels=self.input_channels,
+            dropout=self.hidden_dropout_prob,
+            attention_dropout=self.attention_probs_dropout_prob,
+            max_position_embeddings=self.max_position_embeddings,
+            max_source_positions=self.max_source_positions,
+            max_target_positions=self.max_target_positions,
+            eos_token_id=self.eos_token_id,
+            bos_token_id=self.bos_token_id,
+            pad_token_id=self.pad_token_id,
+            decoder_ffn_dim=self.hidden_size,
+            encoder_ffn_dim=self.hidden_size,
+            decoder_start_token_id=self.decoder_start_token_id,
+            suppress_tokens=self.suppress_tokens,
+            begin_suppress_tokens=self.begin_suppress_tokens,
+        )
+
+    def prepare_config_and_inputs_for_common(self):
+        config, inputs_dict = self.prepare_config_and_inputs()
+
+        inputs_dict["input_ids"][:, -1] = self.pad_token_id
+
+        return config, inputs_dict
+
+    def prepare_config_and_inputs_for_decoder(self):
+        config, input_features = self.prepare_config_and_inputs()
+        input_ids = input_features["input_ids"]
+        encoder_hidden_states = floats_tensor([self.batch_size, self.decoder_seq_length, self.hidden_size])
+
+        return (config, input_ids, encoder_hidden_states)
+
+    def create_and_check_decoder_model_past(self, config, input_ids):
+        config.use_cache = True
+        model = WhisperDecoder(config=config).to(torch_device).eval()
+        # first forward pass
+        outputs = model(input_ids, use_cache=True)
+        outputs_use_cache_conf = model(input_ids)
+        outputs_no_past = model(input_ids, use_cache=False)
+
+        self.parent.assertTrue(len(outputs) == len(outputs_use_cache_conf))
+        self.parent.assertTrue(len(outputs) == len(outputs_no_past) + 1)
+
+        past_key_values = outputs["past_key_values"]
+
+        # create hypothetical next token and extent to next_input_ids
+        next_tokens = ids_tensor((self.batch_size, 1), config.vocab_size)
+
+        # append to next input_ids and
+        next_input_ids = torch.cat([input_ids, next_tokens], dim=-1)
+
+        output_from_no_past = model(next_input_ids)["last_hidden_state"]
+        output_from_past = model(next_tokens, past_key_values=past_key_values)["last_hidden_state"]
+
+        # select random slice
+        random_slice_idx = ids_tensor((1,), output_from_past.shape[-1]).item()
+        output_from_no_past_slice = output_from_no_past[:, next_input_ids.shape[-1] - 1, random_slice_idx].detach()
+        output_from_past_slice = output_from_past[:, 0, random_slice_idx].detach()
+
+        # test that outputs are equal for slice
+        assert torch.allclose(output_from_past_slice, output_from_no_past_slice, atol=1e-3)
+
+    def create_and_check_decoder_model_attention_mask_past(self, config, input_ids):
+        model = WhisperDecoder(config=config).to(torch_device).eval()
+
+        # create attention mask
+        attn_mask = torch.ones(input_ids.shape, dtype=torch.long, device=torch_device)
+
+        half_seq_length = input_ids.shape[-1] // 2
+        attn_mask[:, half_seq_length:] = 0
+
+        # first forward pass
+        past_key_values = model(input_ids, attention_mask=attn_mask, use_cache=True)["past_key_values"]
+
+        # create hypothetical next token and extent to next_input_ids
+        next_tokens = ids_tensor((self.batch_size, 1), config.vocab_size)
+
+        # change a random masked slice from input_ids
+        random_seq_idx_to_change = ids_tensor((1,), half_seq_length).item() + 1
+        random_other_next_tokens = ids_tensor((self.batch_size, 1), config.vocab_size).squeeze(-1)
+        input_ids[:, -random_seq_idx_to_change] = random_other_next_tokens
+
+        # append to next input_ids and attn_mask
+        next_input_ids = torch.cat([input_ids, next_tokens], dim=-1)
+        attn_mask = torch.cat(
+            [attn_mask, torch.ones((attn_mask.shape[0], 1), dtype=torch.long, device=torch_device)],
+            dim=1,
+        )
+
+        # get two different outputs
+        output_from_no_past = model(next_input_ids, attention_mask=attn_mask)["last_hidden_state"]
+        output_from_past = model(next_tokens, attention_mask=attn_mask, past_key_values=past_key_values)[
+            "last_hidden_state"
+        ]
+
+        # select random slice
+        random_slice_idx = ids_tensor((1,), output_from_past.shape[-1]).item()
+        output_from_no_past_slice = output_from_no_past[:, next_input_ids.shape[-1] - 1, random_slice_idx].detach()
+        output_from_past_slice = output_from_past[:, 0, random_slice_idx].detach()
+
+        # test that outputs are equal for slice
+        assert torch.allclose(output_from_past_slice, output_from_no_past_slice, atol=1e-3)
+
+
+@require_torch
+class WhisperStandaloneDecoderModelTest(ModelTesterMixin, GenerationTesterMixin, unittest.TestCase):
+    all_model_classes = (WhisperDecoder, WhisperForCausalLM) if is_torch_available() else ()
+    all_generative_model_classes = (WhisperForCausalLM,) if is_torch_available() else ()
+    fx_comptatible = False
+    test_pruning = False
+    is_encoder_decoder = False
+    test_missing_keys = False
+
+    def setUp(self):
+        self.model_tester = WhisperStandaloneDecoderModelTester(self, is_training=False)
+        self.config_tester = ConfigTester(self, config_class=WhisperConfig)
+
+    def test_config(self):
+        self.config_tester.run_common_tests()
+
+    def test_decoder_model_past(self):
+        config_and_inputs = self.model_tester.prepare_config_and_inputs()
+        config, inputs_dict = config_and_inputs
+
+        self.model_tester.create_and_check_decoder_model_past(config=config, input_ids=inputs_dict["input_ids"])
+
+    def test_decoder_model_attn_mask_past(self):
+        config_and_inputs = self.model_tester.prepare_config_and_inputs()
+        config, inputs_dict = config_and_inputs
+
+        self.model_tester.create_and_check_decoder_model_attention_mask_past(
+            config=config, input_ids=inputs_dict["input_ids"]
+        )
+
+    @unittest.skip("Generate needs input ids")
+    def test_generate_without_input_ids(self):
+        # generate only works with input ids for whisper
+        pass
+
+    @unittest.skip("Decoder can't keep attention grads")
+    def test_retain_grad_hidden_states_attentions(self):
+        # decoder cannot keep gradients
+        return
+
+    @unittest.skip("The model doesn't support fast init from base")
+    def test_save_load_fast_init_from_base(self):
+        pass
+
+    @unittest.skip("The model doesn't support left padding")  # and it's not used enough to be worth fixing :)
+    def test_left_padding_compatibility(self):
+        pass