qia7ddlmZmZmZddlmZmZddlmZddl m Z m Z ddl m Z Gdded Zd Ze Gd d eZd gZy))ProcessingKwargsProcessorMixinUnpack)PreTokenizedInput TextInput) BatchFeature) ImageInputmake_flat_list_of_images)auto_docstringceZdZdddiiZy)Llama4ProcessorKwargs text_kwargs padding_sideleftN)__name__ __module__ __qualname__ _defaults^/opt/pipecat/venv/lib/python3.12/site-packages/transformers/models/llama4/processing_llama4.pyrrs F IrrF)totala>{{- bos_token }} {%- if custom_tools is defined %} {%- set tools = custom_tools %} {%- endif %} {%- if not tools_in_user_message is defined %} {%- set tools_in_user_message = true %} {%- endif %} {%- if not date_string is defined %} {%- if strftime_now is defined %} {%- set date_string = strftime_now("%d %b %Y") %} {%- else %} {%- set date_string = "26 Jul 2024" %} {%- endif %} {%- endif %} {%- if not tools is defined %} {%- set tools = none %} {%- endif %} {#- This block extracts the system message, so we can slot it into the right place. #} {%- if messages[0]['role'] == 'system' %} {%- if messages[0]['content'] is string %} {%- set system_message = messages[0]['content']|trim %} {%- else %} {#- FIXME: The processor requires an array, always. #} {%- set system_message = messages[0]['content'][0]['text']|trim %} {%- endif %} {%- set messages = messages[1:] %} {%- set user_supplied_system_message = true %} {%- else %} {%- set system_message = "" %} {%- set user_supplied_system_message = false %} {%- endif %} {#- System message if the user supplied one #} {%- if user_supplied_system_message %} {{- "<|header_start|>system<|header_end|> " }} {%- if tools is not none %} {{- "Environment: ipython " }} {%- endif %} {%- if tools is not none and not tools_in_user_message %} {{- "You have access to the following functions. To call a function, please respond with JSON for a function call." }} {{- 'Respond in the format {"name": function name, "parameters": dictionary of argument name and its value}.' }} {{- "Do not use variables. " }} {%- for t in tools %} {{- t | tojson(indent=4) }} {{- " " }} {%- endfor %} {%- endif %} {{- system_message }} {{- "<|eot|>" }} {%- endif %} {#- Custom tools are passed in a user message with some extra guidance #} {%- if tools_in_user_message and not tools is none %} {#- Extract the first user message so we can plug it in here #} {%- if messages | length != 0 %} {%- set first_user_message = messages[0]['content']|trim %} {%- set messages = messages[1:] %} {%- else %} {{- raise_exception("Cannot put tools in the first user message when there's no first user message!") }} {%- endif %} {{- '<|header_start|>user<|header_end|> ' -}} {{- "Given the following functions, please respond with a JSON for a function call " }} {{- "with its proper arguments that best answers the given prompt. " }} {{- 'Respond in the format {"name": function name, "parameters": dictionary of argument name and its value}.' }} {{- "Do not use variables. " }} {%- for t in tools %} {{- t | tojson(indent=4) }} {{- " " }} {%- endfor %} {{- first_user_message + "<|eot|>"}} {%- endif %} {%- for message in messages %} {%- if not (message.role == 'ipython' or message.role == 'tool' or 'tool_calls' in message) %} {{- '<|header_start|>' + message['role'] + '<|header_end|> ' }} {%- if message['content'] is string %} {{- message['content'] }} {%- else %} {%- for content in message['content'] %} {%- if content['type'] == 'image' %} {{- '<|image|>' }} {%- elif content['type'] == 'text' %} {{- content['text'] }} {%- endif %} {%- endfor %} {%- endif %} {{- "<|eot|>" }} {%- elif 'tool_calls' in message and message.tool_calls|length > 0 %} {{- '<|header_start|>assistant<|header_end|> ' -}} {{- '<|python_start|>' }} {%- if message['content'] is string %} {{- message['content'] }} {%- else %} {%- for content in message['content'] %} {%- if content['type'] == 'image' %} {{- '<|image|>' }} {%- elif content['type'] == 'text' %} {{- content['text'] }} {%- endif %} {%- endfor %} {%- endif %} {{- '<|python_end|>' }} {%- for tool_call in message.tool_calls %} {{- '{"name": "' + tool_call.function.name + '", ' }} {{- '"parameters": ' }} {{- tool_call.function.arguments | tojson }} {{- "}" }} {%- endfor %} {{- "<|eot|>" }} {%- elif message.role == "tool" or message.role == "ipython" %} {{- "<|header_start|>ipython<|header_end|> " }} {%- if message.content is mapping or message.content is iterable %} {{- message.content | tojson }} {%- else %} {{- message.content }} {%- endif %} {{- "<|eot|>" }} {%- endif %} {%- endfor %} {%- if add_generation_prompt %} {{- '<|header_start|>assistant<|header_end|> ' }} {%- endif %} c eZdZddddddddddd ef d ed effd Zd Ze dde dzde e ze e ze e zdzde edefdZxZS)Llama4ProcessorNg? <|image|><|image_start|> <|image_end|> <|patch|><|tile_x_separator|><|tile_y_separator|> patch_sizepixel_shuffle_ratioc  t|||| ttd|dzz |_||_||_||_|j|j|_ ||_ ||_ | |_ | |_ | |_y)a~ patch_size (`int`, *optional*, defaults to 28): The size of image patches for tokenization. pixel_shuffle_ratio (`float`, *optional*, defaults to `0.5`): The ratio used for pixel shuffling when processing images. This controls the downsampling factor applied to image patches. The actual downsampling ratio is calculated as `1 / (pixel_shuffle_ratio^2)`. fake_image_token (`str`, *optional*, defaults to `"<|image|>"`): The placeholder token in the text that will be replaced with actual image tokens. This token serves as a marker indicating where images should be inserted in the text sequence. image_token (`str`, *optional*, defaults to `"<|image|>"`): The token to be used to represent an image in the text. start_of_image_token (`str`, *optional*, defaults to `"<|image_start|>"`): The special token that marks the beginning of an image sequence in the text. This token is prepended to image token sequences to delimit image boundaries. end_of_image_token (`str`, *optional*, defaults to `"<|image_end|>"`): The special token that marks the end of an image sequence in the text. This token is appended to image token sequences to delimit image boundaries. patch_token (`str`, *optional*, defaults to `"<|patch|>"`): The token used to represent individual image patches. Multiple patch tokens are used to represent the full image, with the number depending on the image size and patch configuration. tile_x_separator_token (`str`, *optional*, defaults to `"<|tile_x_separator|>"`): The token used to separate tiles (patches) horizontally within an image. This token is inserted between patches in the same row when images are split into multiple tiles. tile_y_separator_token (`str`, *optional*, defaults to `"<|tile_y_separator|>"`): The token used to separate tiles (patches) vertically within an image. This token is inserted between rows of patches when images are split into multiple tiles. ) chat_templateg?N)super__init__introunddownsample_ratior#fake_image_token image_tokenconvert_tokens_to_idsimage_token_idstart_of_img_tokenend_of_img_tokenimg_patch_token tile_tokentile_global_token)selfimage_processor tokenizerr#r$r-r.start_of_image_tokenend_of_image_token patch_tokentile_x_separator_tokentile_y_separator_tokenr&kwargs __class__s rr)zLlama4Processor.__init__%sV )=Q #E#1Da1G*H$I J$ 0&'==d>N>NO"6 2*0!7rcd}|\}}||zdkDr;t|D]-}t|D]}|d|zz }||dz ks|dz }|dz }/|dz }|d|zz }|dz }|S)z Create a structured string representation of image tokens Args: num_patches: Number of patches in the image Returns: String with appropriate image tokens rr r!r"rr)range)r6 aspect_rationum_patches_per_chunk img_stringratio_hratio_wyyxxs r_prompt_split_imagez#Llama4Processor._prompt_split_image^s' ' W q Gn 5.=B+0E"EEJGaK'"&<< = 44  5 k! k$999 o% rimagestextr>returnc | tdjtfdjji|}t |t tfs|g}i}|‰jj|}t|}jdd|i|d}|ddjdd\}}t|jz|jzzjz}|jd } t!fd |D} | t#|k7rtd | d t#|d d} g} |D]} | j%j&}|dk(r| j)| 5| j+j&}g}t-|D]G\}}|j)|||ksj/| | |}| dz } |j)|I| j)dj1|| t#|k7r td| }|djdd}j|fi|d}j3||dgt5i|||S)aA Returns: [`BatchFeature`]: A [`BatchFeature`] with the following fields: - **input_ids** -- List of token ids to be fed to a model. Returned when `text` is not `None`. - **attention_mask** -- List of indices specifying which tokens should be attended to by the model (when `return_attention_mask=True` or if *"attention_mask"* is in `self.model_input_names` and if `text` is not `None`). - **pixel_values** -- Pixel values to be fed to a model. Returned when `images` is not `None`. NzYou have to specify text.tokenizer_init_kwargsrK images_kwargs pixel_valuesr aspect_ratiosc3TK|]}|jj!yw)N)countr-).0promptr6s r z+Llama4Processor.__call__..s $\VV\\$2G2G%H$\s%(zFound z) placeholders across the batch, but have z flattened images.rAzONumber of image placeholders in the prompt does not match the number of images.rreturn_tensorsimage) modalities)data tensor_typer) ValueError _merge_kwargsrr8 init_kwargs isinstancelisttupler7 fetch_imagesr shaper*r#r,popsumlenrUr-appendsplit enumeraterJjoin_check_special_mm_tokensr )r6rKrLr> output_kwargs image_inputs image_height image_widthrDrStotal_placeholders image_indexprocessed_textrWplaceholder_count prompt_splits new_promptlocal_image_index split_parttokens_for_this_imagerZ text_inputss` r__call__zLlama4Processor.__call__ys" <89 9*** ! "&.."<"<  $u .6D  ))66v>F-f5F/4//`v`A_`L(4^(DQ(G(M(Mbc(R %L+$'0[DOO5STX\XmXmm% !),,_=M!$$\W[$\!\ !S[0 /01 #F },>@ KN ;$*LL1F1F$G!$)"))&1 & T-B-B C  5>}5MA1%z%%j1(+<<040H0H)+68M1-$q( "))*?@A%%bggj&9:! ;$c&k) !rss!D&}599:JDQ$dnnTJ]=-IJ  %%dKWI%N!@K!@rs^ SRM2?#,E]P b^nb^b^J  r