This section lists the languages, embedding models, and large language models (LLMs) that HeatWave GenAI supports.
In HeatWave 9.0.0
and later versions,
HeatWave GenAI supports natural-language communication, ingesting
documents, as well as generating text-based content in English
(en
).
In HeatWave 9.0.1-u1
and later versions,
HeatWave GenAI supports natural-language communication, ingesting
documents, as well as generating text-based content in
multiple languages. The quality of the generated text outputs
depends on the training and ability of the LLM to work with
the language.
Following is a combined list of languages supported by HeatWave In-Database LLMs and OCI Generative AI Service LLMs:
Arabic (
ar
)Bengali (
bn
)Brazilian Portuguese (
pt-br
)Burmese (
my
)Chinese (
zh
)Czech (
cs
)Dutch (
nl
)English (
en
)French (
fr
)German (
de
)Hebrew (
he
)Hindi (
hi
)Indonesian (
id
)Italian (
it
)Japanese (
ja
)Khmer (
km
)Korean (
ko
)Lao (
lo
)Malay (
ms
)Persian (
fa
)Polish (
pl
)Portuguese (
pt
)Spanish (
es
)Tagalog (
tl
)Thai (
th
)Turkish (
tr
)Urdu (
ur
)Vietnamese (
vi
)
However, not all LLMs support all the languages. To learn which LLM supports which language, see HeatWave In-Database LLMs and OCI Generative AI Service LLMs.
To set the value of the language
parameter in HeatWave GenAI
routines that support this parameter, do not use the
language name to specify the language. Use the two-letter
ISO 639-1
code for the language instead.
For example, to use French, use the ISO
639-1
code for French, which is
fr
.
HeatWave GenAI supports the following in-database embedding models for generating vector embeddings:
all_minilm_l12_v2
: in HeatWave9.0.0
and later versions, HeatWave GenAI uses this embedding model, by default, for encoding English documents.multilingual-e5-small
: in HeatWave9.0.1-u1
and later versions, HeatWave GenAI uses this embedding model, by default, for encoding documents in all supported languages other than English.
In HeatWave 9.0.1-u1
and later versions,
HeatWave GenAI supports Oracle Cloud Infrastructure (OCI) Generative AI service embedding
models for
generating vector embeddings.
These models are available only in the regions supported by the OCI Generative AI service. For more information, see Pretrained Foundational Models in Generative AI.
However, to use the OCI Generative AI service with HeatWave GenAI, you need to enable the database system to access OCI services. For more information, see Section 4.2.3, “Authenticating OCI Generative AI Service”.
HeatWave GenAI supports the following OCI Generative AI service embedding models:
cohere.embed-english-v3.0
: for encoding English text.cohere.embed-multilingual-v3.0
: for encoding text in supported languages other than English.
HeatWave GenAI supports the following in-database LLMs:
-
mistral-7b-instruct-v1
In HeatWave
9.0.0
and later versions, supports English.-
In HeatWave
9.0.1-u1
and later versions, supports the following languages:English (
en
)French (
fr
)German (
de
)Spanish (
es
)
-
llama2-7b-v1
In HeatWave
9.0.0
, supports English (en
).In HeatWave
9.0.1-u1
and later versions, support forllama2-7b-v1
has been deprecated.
-
llama3-8b-instruct-v1
This LLM is available in HeatWave
9.0.1-u1
and later versions.-
It supports the following languages:
English (
en
)French (
fr
)German (
de
)Hindi (
hi
)Italian (
it
)Portuguese (
pt
)Spanish (
es
)Thai (
th
)
In HeatWave 9.0.1-u1
and later versions,
HeatWave GenAI supports OCI Generative AI service LLMs.
These LLMs are available only in the regions supported by the OCI Generative AI service. For more information, see Pretrained Foundational Models in Generative AI.
However, to use the OCI Generative AI service with HeatWave GenAI, you need to enable the database system to access OCI services. For more information, see Section 4.2.3, “Authenticating OCI Generative AI Service”.
HeatWave GenAI supports the following OCI Generative AI service LLMs: