^j4ddlmZddlmZddlZddlmcmZddlmZddl m Z ddl m Z ddlmZdd lmZmZmZmZdd lmZdd lmZdd lmZmZmZdd lmZmZddl m!Z!ddl"m#Z#m$Z$m%Z%m&Z&m'Z'm(Z(ddl)m*Z*ddl+m,Z,ddl-m.Z.ddl/m0Z0m1Z1m2Z2e&jfe4Z5e$GddeZ6e$deGddeZ7e$deGddeZ8e$deGd d!eZ9Gd"d#ejtZ;d$ejxd%e=d&ejxfd'Z> d\d(ejtd)ejxd*ejxd+ejxd,ejxdzd-e?d.e?d/e!e#fd0Z@Gd1d2ejtZAGd3d4ejtZBGd5d6eZCGd7d8ejtZDe$Gd9d:e6ZEGd;dejtZGGd?d@ejtZHGdAdBejtZIGdCdDejtZJGdEdFejtZKGdGdHejtZLGdIdJejtZMGdKdLejtZNe$eGdMdNeZOe$dOGdPdQe6ZPGdRdSejtZQGdTdUejtZRe$dVGdWdXe6ZSGdYdZe6eZTgd[ZUy)])Callable) dataclassN)nn)initialization)ACT2FN)Cache)%ClassifierFreeGuidanceLogitsProcessorGenerationMixinGenerationModeLogitsProcessorList)GenerateDecoderOnlyOutput)GradientCheckpointingLayer)BaseModelOutputBaseModelOutputWithPooling ModelOutput)ALL_ATTENTION_FUNCTIONSPreTrainedModel)Unpack)TransformersKwargsauto_docstringcan_return_tupleloggingtorch_compilable_check torch_int)merge_with_config_defaults)capture_outputs) AutoModel) JanusConfigJanusVisionConfigJanusVQVAEConfigcReZdZUeed<dZdZdZddgZddgZ dZ dZ dZ fd Z xZS) JanusPreTrainedModelconfigmodelimagetextTLlamaDecoderLayerJanusVisionEncoderLayerpast_key_values causal_maskct||t|trZt j |j tj|j jdjdyy)Nr r0) super _init_weights isinstanceJanusVisionEmbeddingsinitcopy_ position_idstorcharangeshapeexpand)selfmodule __class__s s/var/www/ramen.bs-engineer-server.com/venv/lib/python3.12/site-packages/transformers/models/janus/modeling_janus.pyr3z"JanusPreTrainedModel._init_weights<s[ f% f3 4 JJv**ELL9L9L9R9RSU9V,W,^,^_f,g h 5)__name__ __module__ __qualname__r!__annotations__base_model_prefixinput_modalitiessupports_gradient_checkpointing_no_split_modules_skip_keys_device_placement_supports_flash_attn_supports_sdpa_can_compile_fullgraphr3 __classcell__r?s@r@r%r%/sO (&*#,.GH#4m"DN!iirAr%z9 Base class for Janus VQ-VAE mode model outputs. ) custom_introcbeZdZUdZdZej dzed<dZej dzed<y)JanusVQVAEOutputz decoded_pixel_values (`torch.FloatTensor` of shape `(batch_size, num_channels, image_size, image_size)`): Reconstructed pixel values after encoding and decoding the input. embedding_loss (`torch.FloatTensor`): Embedding loss. Ndecoded_pixel_valuesembedding_loss) rBrCrD__doc__rSr9 FloatTensorrErTrAr@rRrRBs46:%++d29/3NE%%,3rArRzy Base class for Janus model's outputs that may also contain a past key/values (to speed up sequential decoding). ceZdZUdZdZej dzed<dZe dzed<dZ e ej dzed<dZ e ej dzed<dZ e ej dzed<y)JanusBaseModelOutputWithPasta last_hidden_state (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`): Sequence of hidden-states at the output of the last layer of the model. If `past_key_values` is used only the last hidden-state of the sequences of shape `(batch_size, 1, hidden_size)` is output. image_hidden_states (`tuple(torch.FloatTensor)`, *optional*): Tuple of `torch.FloatTensor` (one for the output of the image embeddings, `(batch_size, num_images, sequence_length, hidden_size)`. image_hidden_states of the model produced by the vision encoder, and optionally by the perceiver Nlast_hidden_stater- hidden_states attentionsimage_hidden_states)rBrCrDrUrZr9rVrEr-r r[tupler\r]rWrAr@rYrYTs| 37u((4/6$(OUT\(59M5**+d2926Je''(4/6;?u001D8?rArYzQ Base class for Janus causal language model (or autoregressive) outputs. ceZdZUdZdZej dzed<dZej dzed<dZ e dzed<dZ e ej dzed<dZ e ej dzed<dZe ej dzed<y) JanusCausalLMOutputWithPastae loss (`torch.FloatTensor` of shape `(1,)`, *optional*, returned when `labels` is provided): Language modeling loss (for next-token prediction). logits (`torch.FloatTensor` of shape `(batch_size, sequence_length, config.vocab_size)`): Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax). past_key_values (`Cache`, *optional*, returned when `use_cache=True` is passed or when `config.use_cache=True`): It is a [`~cache_utils.Cache`] instance. For more details, see our [kv cache guide](https://huggingface.co/docs/transformers/en/kv_cache). Contains pre-computed hidden-states (key and values in the self-attention blocks) that can be used (see `past_key_values` input) to speed up sequential decoding. image_hidden_states (`tuple(torch.FloatTensor)`, *optional*): Tuple of `torch.FloatTensor` (one for the output of the image embeddings, `(batch_size, num_images, sequence_length, hidden_size)`. image_hidden_states of the model produced by the vision encoder, and optionally by the perceiver Nlosslogitsr-r[r\r])rBrCrDrUrar9rVrErbr-r r[r^r\r]rWrAr@r`r`os"&*D%  d ")'+FE   $+$(OUT\(59M5**+d2926Je''(4/6;?u001D8?rAr`ceZdZdeffd Zdej dededej fdZd dej d e dej fd Z xZ S) r5r&cft|||_|j|_|j |_|j |_tj|j|j|j |j d|_ |j |j zdz|_ |j|_ tj|j|j|_|jdt!j"|jj%ddy)Nvalid) in_channels out_channels kernel_sizestridepaddingrr8r1F) persistent)r2__init__r& hidden_size embed_dim image_size patch_sizerConv2d num_channelspatch_embedding num_patches num_positions Embeddingposition_embeddingregister_bufferr9r:r<r=r&r?s r@rlzJanusVisionEmbeddings.__init__s  ++ ++ ++!yy++??  !OOt>1D!--"$,,t/A/A4>>"R ^U\\$:L:L-M-T-TU\-]joprA embeddingsheightwidthreturnc|jd}|jjjd}tjj s%||k(r ||k(r|j|j S|jjjd}|jd}||jz}||jz} t|dz} |jd| | |}|jdddd}tjj||| fdd }|jddddjdd|}|S) a This method allows to interpolate the pre-trained position encodings, to be able to use the model on higher resolution images. This method is also adapted to support torch.jit tracing and no class embeddings. Adapted from: - https://github.com/facebookresearch/dino/blob/de9ee3df6cf39fac952ab558447af1fa1365362a/vision_transformer.py#L174-L194, and - https://github.com/facebookresearch/dinov2/blob/e1277af2ba9496fbadf7aec6eba56e8d882d1e35/dinov2/models/vision_transformer.py#L179-L211 r rr0g?rrbicubicF)sizemode align_corners)r;rwweightr9jit is_tracingr8 unsqueezerprreshapepermuter functional interpolateview) r=rzr{r|rtrupatch_pos_embeddim new_height new_widthsqrt_num_positionss r@interpolate_pos_encodingz.JanusVisionEmbeddings.interpolate_pos_encodingsE!&&q) //66<