Translate text from India’s 22 scheduled languages into English with AI4Bharat’s 1.1B-parameter IndicTrans2 checkpoint. Use the playground to give IndicTrans2 Indic-English 1B a representative input and review its output. Adjust the available settings to shape the result for your task.
Model
ai4bharat/indictrans2-indic-en-1B
Documentation
Use this reference to understand how every field in the playground changes the request sent to ai4bharat/indictrans2-indic-en-1B.
Required
Type: Text
Provides one or more sentences to translate from the selected Indian language into English. Keep each sentence in the script associated with the selected language code; longer batches take more memory and each result follows the corresponding input order.
Required
Type: String
Default: hin_Deva
Values: Assamese, Bengali, Bodo, Dogri, Konkani, Gujarati, Hindi, Kannada, Kashmiri (Arabic), Kashmiri (Devanagari), Maithili, Malayalam, Manipuri (Bengali), Manipuri (Meitei), Marathi, Nepali, Odia, Punjabi, Sanskrit, Santali, Sindhi (Arabic), Sindhi (Devanagari), Tamil, Telugu, Urdu
Identifies the language and writing script of the source sentences using an IndicTrans2 code. Choose the exact language-script combination—for example, Kashmiri in Arabic versus Devanagari—because it controls preprocessing and an incorrect value can produce a poor translation.
Optional
Type: Integer
Default: 256
Range: 1–Unbounded
Sets the maximum total token length of each generated translation. The model may stop earlier, but a value that is too low can truncate the result; increasing it permits longer output at the cost of additional generation time.
Optional
Type: Integer
Default: 0
Range: 0–Unbounded
Prevents the model from ending a generated translation or sequence before this many tokens have been produced. Leave it at zero for natural stopping, and raise it cautiously because forcing extra length can add unnecessary or inaccurate text.
Optional
Type: Integer
Default: 5
Range: 1–Unbounded
Sets how many candidate continuations the model explores in parallel at each generation step. A value of 1 uses ordinary generation, while larger values can find a better overall sequence at the cost of additional time and memory.
Optional
Type: Integer
Default: 1
Range: 1–Unbounded
Requests this many separate candidate outputs for the same input. More candidates provide alternatives to compare but increase generation time and memory use, and sampled decoding is normally needed for the candidates to differ.
Optional
Type: Boolean
Default: On
Caches attention calculations from previously generated tokens so each new token does not recompute the entire sequence. Keep it enabled for ordinary generation; disable it mainly for debugging, memory experiments, or models whose cache implementation is incompatible.