h[:SSKrSSKJrJr SSKJr SSKJr SSKJ r SSK J r \"SS55r S r S rS r\ "S 5r\S :Xa\ "\ 5r\R'5Sr\"SS\R*S9r\R/\ SS/\R*S9r\R1\\R*S9r\R/\S/\R*S9r\R2(a?\R3\R4\R*S9 \R3\R4SS9 ggg)N) dataclassfield)Optional) load_dataset) ModelCard)HfArgumentParserct\rSrSr%Sr\"SSS0S9r\\S'\"SSS 0S9r \ \S '\"S SS 0S9r \ \ \S 'Srg )ScriptArgumentsa Arguments for the script. Args: push_to_hub (`bool`, *optional*, defaults to `False`): Whether to push the dataset to the Hugging Face Hub. repo_id (`str`, *optional*, defaults to `"trl-lib/llava-instruct-mix"`): Hugging Face repository ID to push the dataset to. dataset_num_proc (`int` or `None`, *optional*, defaults to `None`): Number of workers to use for dataset processing. Fhelpz4Whether to push the dataset to the Hugging Face Hub.)defaultmetadata push_to_hubztrl-lib/llava-instruct-mixz2Hugging Face repository ID to push the dataset to.repo_idNz0Number of workers to use for dataset processing.dataset_num_proc)__name__ __module__ __qualname____firstlineno____doc__rrbool__annotations__rstrrrint__static_attributes__r^/home/james-whalen/.local/lib/python3.13/site-packages/examples/datasets/llava_instruct_mix.pyr r si PQK,NOGS',LM'hsmrr c/n[R"US5HHnUSnURSS5R5nUSS:XaSOSnUR XCS .5 MJ XS /S .$) N conversationsvaluezfromhumanuser assistant)rolecontentimage)messagesimages)ast literal_evalreplacestripappend)exampler*messager(r's rprocess_exampler34sxH##GO$<='"//)R0668 G3v:; > !W-=,> ??rc6[SUS55nUS:*$)Nc3># UHn[US5v M g7f)r(Nlen).0msgs r 'filter_long_examples..?sJ6Iss3y>**6Isr*i)sum)r1 total_lengths rfilter_long_examplesr>>s"Jgj6IJJL 4 rcZ[US5S:deUSSSUS'USSSUS'U$)zd Splits the messages into a prompt and a completion. The last message is considered the completion. r*Nprompt completionr6)r1s rsplit_prompt_completionrDCsM wz" #a '' ' +CR0GH#J/4GL Nra --- tags: [trl] --- # LLaVA Instruct Mix ## Summary The LLaVA Instruct Mix dataset is a processed version of [LLaVA Instruct Mix](https://huggingface.co/datasets/theblackcat102/llava-instruct-mix). ## Data Structure - **Format**: [Conversational](https://huggingface.co/docs/trl/main/dataset_formats#conversational) - **Type**: [Language-modeling](https://huggingface.co/docs/trl/main/dataset_formats#language-modeling) Columns: - `"images"`: The image associated with the text. - `"prompt"`: A list of messages that form the context for the conversation. - `"completion"`: The last message in the conversation, which is the model's response. This structure allows models to learn from the context of the conversation, enhancing their understanding of how to generate descriptive text based on visual inputs. ## Generation script The script used to generate this dataset can be found [here](https://github.com/huggingface/trl/blob/main/examples/datasets/llava_instruct_mix.py). __main__z!theblackcat102/llava-instruct-mixtrain)splitnum_procr r))remove_columnsrH)rHr*dataset) repo_type)r, dataclassesrrtypingrdatasetsrhuggingface_hubr transformersrr r3r>rD model_cardrparserparse_args_into_dataclasses script_argsrrJmapfilterrrrrrrWs9 (!%)  6@  8 z o .F446q9K>gXcXtXtuGkk'(B[MiMiGnn1K