lllyasviel/Fooocus · error · ValueError

You cannot specify both input_ids and inputs_embeds at the s

Error message

You cannot specify both input_ids and inputs_embeds at the same time

What it means

NLVR encoder's BertModel forward rejects calls that provide both input_ids and inputs_embeds, because they are mutually exclusive ways of specifying the input sequence; supplying both is ambiguous.

Source

Thrown at extras/BLIP/models/nlvr_encoder.py:753

            (those that don't have their past key value states given to this model) of shape :obj:`(batch_size, 1)`
            instead of all :obj:`decoder_input_ids` of shape :obj:`(batch_size, sequence_length)`.
        use_cache (:obj:`bool`, `optional`):
            If set to :obj:`True`, :obj:`past_key_values` key value states are returned and can be used to speed up
            decoding (see :obj:`past_key_values`).
        """
        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 is_decoder:
            use_cache = use_cache if use_cache is not None else self.config.use_cache
        else:
            use_cache = False

        if input_ids is not None and inputs_embeds is not None:
            raise ValueError("You cannot specify both input_ids and inputs_embeds at the same time")
        elif input_ids is not None:
            input_shape = input_ids.size()
            batch_size, seq_length = input_shape
            device = input_ids.device
        elif inputs_embeds is not None:
            input_shape = inputs_embeds.size()[:-1]
            batch_size, seq_length = input_shape
            device = inputs_embeds.device
        elif encoder_embeds is not None:    
            input_shape = encoder_embeds.size()[:-1]
            batch_size, seq_length = input_shape 
            device = encoder_embeds.device
        else:
            raise ValueError("You have to specify either input_ids or inputs_embeds or encoder_embeds")

        # past_key_values_length
        past_key_values_length = past_key_values[0][0].shape[2] if past_key_values is not None else 0

View on GitHub (pinned to ae05379cc9)

Solutions

1. Remove one of the two arguments from the call
2. Sanitize kwargs in wrappers: kwargs.pop('input_ids', None) before forwarding embeddings
3. Add a unit test asserting exactly one input source is passed

Example fix

// before
loss = model(input_ids=ids, inputs_embeds=emb)

// after
loss = model(inputs_embeds=emb)

sources = [k for k in ('input_ids', 'inputs_embeds') if locals().get(k) is not None]
assert len(sources) <= 1, f'conflicting inputs: {sources}'

When it happens

Trigger: encoder(input_ids=ids, inputs_embeds=emb) with both non-None, usually via **kwargs forwarding or incomplete refactors from ID-based to embedding-based inputs.

Common situations: Porting NLVR training code between BLIP repo versions; collators that always emit input_ids while the model code adds inputs_embeds.

Related errors

• You cannot specify both input_ids and inputs_embeds at the s
• You have to specify either input_ids or inputs_embeds or enc
• You have to specify either input_ids or inputs_embeds or enc
• checkpoint url or path is invalid
• The hidden size (%d) is not a multiple of the number of atte

AI-assisted analysis of lllyasviel/Fooocus@ae05379cc9 (2026-08-15). Data as JSON: /api/errors/07e59ec3bbbaf451. Report an issue: GitHub.