o }oiG @sddlmZddlZddlmZddlmZmZddlm Z ddl m Z ddl m Z mZddlmZmZmZmZmZGd d d eZd eeefd eeeffd dZe eededed eeejffddZdS))AnyN)MonoCut)CutMixedCut)ifnone)registered_prompt_format_fn)ModalityPromptFormatter) CANARY_BOS CANARY_EOS CANARY_NOPNC CANARY_PNCCANARY_SPECIAL_TOKENIZERcseZdZdZdZdedejeddddd d d ejed d dddddddddd ddede dejidiZ de e efde e e fddffd d! Zd"e d#e d$e deeffd%d& ZZS)'CanaryPromptFormattercanary assistantuserz%|source_lang||task||target_lang||pnc|asrast translate transcribes2t_translation<|transcribe|> <|translate|>yesnotrueTruefalseFalse10pncnopncz<|pnc|>z <|nopnc|>) source_langtask target_langr")templateslotsz|text|textexpectedreceivedreturnNcs.d|vrd|vr|d|d<tj||dS)Ntasknamer%)r*r+)popsuper_validate_slot_values)selfr*r+ __class__Z/home/ubuntu/.local/lib/python3.10/site-packages/nemo/collections/common/prompts/canary.pyr0:sz+CanaryPromptFormatter._validate_slot_valuesprompt_templateexpected_slots slot_valuescst|}tj|||dS)N)r6r7r8)%map_manifest_values_to_special_tokensr/ encode_turn)r1r6r7r8r2r4r5r:?sz!CanaryPromptFormatter.encode_turn)__name__ __module__ __qualname__NAME OUTPUT_ROLEr rText TextLiteralr TEMPLATEdictstrrr0listintr: __classcell__r4r4r2r5r!s. **rr8r,cCs |}d}dD] }||vr(||}dr|ds(d||d||<d}qd}||vrE||ttfvrE||dvr?tnt||<d}dD]0}||vrw||d vrw||d vr^d |d <n||d vrid|d <n Jd||d|d}qG|rtj|vrt|tj<|S)NF)r$r&z<|z|>Tr")rr rrr")r%r-)rr>rrrrr%>rrrzTask z invalid task for slot )copy startswithendswithr r r PROMPT_LANGUAGE_SLOTr)r8any_special_token_presentkvr4r4r5r9Ls.$     r9cutpromptc sLttrjttstddt|d}|tj}d|vr9djvr9jdjd<|d|rGt dj d|dt difd d |D|j t id g}d d djD}|t dd||j tjdjjdid ||}|dd|jjksJd|d|ddd|d<|S)a` Prepend and append control tokens to the token sequence as per Canary format. We use the following special tokens: * <|startoftranscript|> * <|transcribe|> * <|translate|> * <|nopnc|> * <|pnc|> * <|endoftext|> * <|LANG|> - for each supported language. * <|nospeech|> The prompt format syntax is as follows: <|startoftranscript|> [ <|nospeech|> | <|LANG|> [ <|transcribe|> | <|translate|> ] <|LANG|> [ <|pnc|> | <|nopnc|> ] TEXT <|endoftext|> ] Where expression ``[ a | b ]`` denotes expression ``a`` or expression ``b``, and can be nested. Note that ``<|LANG|>`` appears twice: the first occurrence is for the "source" language (i.e., spoken language in the recording) and the second occurrence is for the "target" language (i.e., the language in which we are going to output the text). z_Expected input audio to have a single channel (required MonoCut/MixedCut, but we received: cut=)rr%r-zWe found cut with ID z% that is missing the following keys: zNPlease ensure that every utterance in the input manifests contains these keys.csi|]}|j|qSr4)custom).0slotrOr4r5 szcanary..)roler( css |] }|jdur|jVqdS)N)r))rSsr4r4r5 szcanary..rr)rr& answer_idszs     "+"&