Model updates
Bareun automatically downloads the models and dictionaries used for morphological analysis the first time it runs.
After that, when a new model is released, you can choose and apply it yourself on the API key registration screen of the installed server.
You do not need to reinstall the server or download the package again.
Who can change it
Viewing and applying models can be done only with the API key that installed and registered this server (the key issued with the Bareun ID). With the API key of another user who shares the same server, even the list is not shown. This is so that the person who installed the server manages the models.
Opening the model update screen
- Go to the home page (built-in web server) of the installed
Bareun. The default address ishttp://<server-address>:5656/. - On the Get started (시작하기) screen, register your API key. If it is already registered, you see the
API key status(API키 상태). - In the Model update (모델 업데이트) table under
API key status, check the models you can use.
Can't see the screen?
Model update appears only after you register an API key. It is not shown in the following cases.
- You have not registered an API key yet: register one first.
- You are viewing with a key other than the one that registered this server (you do not have permission to view the list).
On a server that was installed so that it does not download models automatically, or that has a specific model pinned, only the Current model (지금 쓰는 모델) box is shown instead of the table. In that case you cannot change the model here.
The current model
Above the table is the Current model box. It shows the name of the model this server is actually using, its description (what changed in that model), the date it was downloaded, and the range of compatible server versions.
It shows even without a list
This box reads from a record the server wrote when it downloaded the model. So even in environments where the list cannot be loaded (a server with a pinned model, or a network cut off from the outside internet), you can still see what you are using now.
For an install that never downloaded a model (where the model came bundled in the installer), only the name is shown and the description is empty.
Information in the table
| Item | Description |
|---|---|
| Model | The name (identifier) of the model. The one marked Current (현재 적용) is the model in use now. |
| Description | The characteristics of that model. It says what improved. |
| Release date | The date the model was made public. |
| Compatible versions | The range of Bareun server versions that can use that model. If it does not match the current server, it is shown as Incompatible (비호환) and cannot be applied. |
| Apply | Switches to that model. If it is already in use, it shows Applied (적용됨). |
Applying a model
When you click Apply (적용), a confirmation window appears, and after you confirm, the following happens in order.
- The model and dictionary files of the chosen model are downloaded (this may take a while depending on the network).
- The
Bareunserver restarts itself and loads the new model. - After a moment, the
Currentmarker in the table moves to the new model. If it has not changed, clickRefresh(새로고침) to check.
There is a brief interruption when you apply
The new model loads only when the server restarts. Requests in progress during the apply can be cut off, so it is best to apply during a low-traffic time.
Sometimes it looks view-only
When deployed to the cloud or installed on Windows, the server cannot restart itself, so the table is view-only. You can only check which models are available, and applying one has to be done by redeploying or restarting the server.
Checking after applying
- Open
Details(상세보기) at the top of the Get started screen to see theModel ID(모델ID). If this value is the same as the new model's name, it has been applied. - Analysis results may change. The custom dictionary is kept separately from the model, so you do not need to register it again.
Installing on a network with no internet access
Bareun downloads models on first run, but if a model is already in place, it uses it as is.
So on a closed network, prepare as follows.
- Install the same version of
Bareunsomewhere with internet access and run it once so that it downloads the model. - Copy the model folder in that server's data area (
var/modelsin the install folder) as is. - Paste it in the same location on the closed-network server and start the server. Since the model is already there, it starts up without downloading.
The server version and the model must match
A model has a defined compatibility range with the Bareun server version. Move a model obtained from an install of the same version
as the closed-network server. If you put in a model of a different version, the server may refuse to use it, or
analysis quality may drop.
It is the same when updating to a new model
On a closed network you cannot change models from this screen, so move the new model in the same way and restart the server. If preparation is difficult, contact helpdesk@bareun.ai.
Frequently asked questions
Q. Does the custom dictionary disappear when I change the model? A. No. The custom dictionary is stored separately from the model, so it stays as it is.
Q. Can I go back to an earlier model?
A. Click Apply again on an earlier model that is still in the table. Models that have disappeared from the list cannot be restored.
Q. I want to use a model marked Incompatible.
A. The Bareun server version that model requires differs from your current install. Raise or lower the server to a version
within that range, and then apply it.
Q. I installed it somewhere with no internet access. A. The model list cannot be loaded, so the table does not appear and only the Current model box is shown. For how to prepare models in advance, see Installing on a network with no internet access above.