Describe

The translate task in Whisper only supports translating other languages into English. OpenAI does not guarantee translations between arbitrary languages. In such cases, you can opt to use the Translation Model for translation tasks. However, it's important to note that the Translation Model runs very slowly on CPU, and the completion time may be twice as long as usual. It is recommended to run the Translation Model on devices with GPUs for better performance.

The larger the parameters of the Translation model, the better its translation capability is expected. However, this also requires higher computational resources and slower running speed.

Currently, when the Highlight Words timestamps option is enabled in the Whisper Word Timestamps options, it cannot be used simultaneously with the Translation Model. This is because Highlight Words splits the source text, and after translation, it becomes a non-word-level string.

Translation Model

The required VRAM is provided for reference and may not apply to everyone. If the model's VRAM requirement exceeds the available capacity of the system, the model will operate on the CPU, resulting in significantly longer execution times.

CTranslate2 is a C++ and Python library for efficient inference with Transformer models. Models converted from CTranslate2 can run with lower resources and faster speed. Encoder-decoder models currently supported: Transformer base/big, M2M-100, NLLB, BART, mBART, Pegasus, T5, Whisper.

M2M100

2M100 is a multilingual translation model introduced by Facebook AI in October 2020. It supports arbitrary translation among 101 languages. The paper is titled "Beyond English-Centric Multilingual Machine Translation" (arXiv:2010.11125).

Name	Parameters	Size	type/quantize	Required VRAM
facebook/m2m100_418M	480M	1.94 GB	float32	≈2 GB
facebook/m2m100_1.2B	1.2B	4.96 GB	float32	≈5 GB
facebook/m2m100-12B-last-ckpt	12B	47.2 GB	float32	N/A

M2M100-CTranslate2

Name	Parameters	Size	type/quantize	Required VRAM
michaelfeil/ct2fast-m2m100_418M	480M	970 MB	float16	≈0.6 GB
michaelfeil/ct2fast-m2m100_1.2B	1.2B	2.48 GB	float16	≈1.3 GB
michaelfeil/ct2fast-m2m100-12B-last-ckpt	12B	23.6 GB	float16	N/A

NLLB-200

NLLB-200 is a multilingual translation model introduced by Meta AI in July 2022. It supports arbitrary translation among 202 languages. The paper is titled "No Language Left Behind: Scaling Human-Centered Machine Translation" (arXiv:2207.04672).

Name	Parameters	Size	type/quantize	Required VRAM
facebook/nllb-200-distilled-600M	600M	2.46 GB	float32	≈2.5 GB
facebook/nllb-200-distilled-1.3B	1.3B	5.48 GB	float32	≈5.9 GB
facebook/nllb-200-1.3B	1.3B	5.48 GB	float32	5.8 GB
facebook/nllb-200-3.3B	3.3B	17.58 GB	float32	13.4 GB

NLLB-200-CTranslate2

Name	Parameters	Size	type/quantize	Required VRAM
michaelfeil/ct2fast-nllb-200-distilled-1.3B	1.3B	1.38 GB	int8_float16	≈1.3 GB
michaelfeil/ct2fast-nllb-200-3.3B	3.3B	3.36 GB	int8_float16	≈3.2 GB
JustFrederik/nllb-200-1.3B-ct2-int8	1.3B	1.38 GB	int8	≈1.3 GB
JustFrederik/nllb-200-1.3B-ct2-float16	1.3B	2.74 GB	float16	≈1.3 GB
JustFrederik/nllb-200-distilled-600M-ct2	600M	2.46 GB	float32	≈0.6 GB
JustFrederik/nllb-200-distilled-600M-ct2-float16	600M	1.23 GB	float16	≈0.6 GB
JustFrederik/nllb-200-distilled-600M-ct2-int8	600M	623 MB	int8	≈0.6 GB
JustFrederik/nllb-200-distilled-1.3B-ct2-float16	1.3B	2.74 GB	float16	≈1.3 GB
JustFrederik/nllb-200-distilled-1.3B-ct2-int8	1.3B	1.38 GB	int8	≈1.3 GB
JustFrederik/nllb-200-distilled-1.3B-ct2	1.3B	5.49 GB	float32	≈1.3 GB
JustFrederik/nllb-200-1.3B-ct2	1.3B	5.49 GB	float32	≈1.3 GB
JustFrederik/nllb-200-3.3B-ct2-float16	3.3B	6.69 GB	float16	≈3.2 GB

MT5

mT5 is a multilingual pre-trained Text-to-Text Transformer introduced by Google Research in October 2020. It is a multilingual variant of the T5 model, pre-trained on datasets in 101 languages. Further fine-tuning is required to transform it into a translation model. The paper is titled "mT5: A Massively Multilingual Pre-trained Text-to-Text Transformer" (arXiv:2010.11934).
The 'mt5-zh-ja-en-trimmed' model is finetuned from Google's 'mt5-base' model. This model has a relatively good translation speed, but it only supports three languages: Chinese, Japanese, and English.

Name	Parameters	Size	type/quantize	Required VRAM
mt5-base	N/A	2.33 GB	float32	N/A
K024/mt5-zh-ja-en-trimmed	N/A	1.32 GB	float32	≈1.4 GB
engmatic-earth/mt5-zh-ja-en-trimmed-fine-tuned-v1	N/A	1.32 GB	float32	≈1.4 GB

ALMA

ALMA is a many-to-many LLM-based translation model introduced by Haoran Xu and colleagues in September 2023. It is based on the fine-tuning of a large language model (LLaMA-2). The approach used for this model is referred to as Advanced Language Model-based trAnslator (ALMA). The paper is titled "A Paradigm Shift in Machine Translation: Boosting Translation Performance of Large Language Models" (arXiv:2309.11674).
The official support for ALMA currently includes 10 language directions: English↔German, English↔Czech, English↔Icelandic, English↔Chinese, and English↔Russian. However, the author hints that there might be surprises in other directions, so there are currently no restrictions on the languages that ALMA can be chosen for in the web UI.

Name	Parameters	Size	type/quantize	Required VRAM
haoranxu/ALMA-7B	7B	26.95 GB	float32	N/A
haoranxu/ALMA-13B	13B	52.07 GB	float32	N/A

ALMA-GPTQ

GPTQ is a technique used to quantize the parameters of large language models into integer formats such as int8 or int4. Although the quantization process may lead to a loss in model performance, it significantly reduces both file size and the required VRAM.

Name	Parameters	Size	type/quantize	Required VRAM
TheBloke/ALMA-7B-GPTQ	7B	3.9 GB	4 Bits	≈4.3 GB
TheBloke/ALMA-13B-GPTQ	13B	7.26 GB	4 Bits	≈8.1

Options

Translation - Batch Size

transformers: batch_size
When the pipeline will use DataLoader (when passing a dataset, on GPU for a Pytorch model), the size of the batch to use, for inference this is not always beneficial.
ctranslate2: max_batch_size
The maximum batch size.

Translation - No Repeat Ngram Size

transformers: no_repeat_ngram_size
Value that will be used by default in the generate method of the model for no_repeat_ngram_size. If set to int > 0, all ngrams of that size can only occur once.
ctranslate2: no_repeat_ngram_size
Prevent repetitions of ngrams with this size (set 0 to disable).

Translation - Num Beams

transformers: num_beams
Number of beams for beam search that will be used by default in the generate method of the model. 1 means no beam search.
ctranslate2: beam_size
Beam size (1 for greedy search).