baalajimaestro/ollama
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

Create a sentiments example

Browse source 
Signed-off-by: Matt Williams <m@technovangelist.com>
This commit is contained in:
Matt Williams 2023-08-03 16:38:31 -07:00
parent e3fb1fd3f1
commit 8f2df948ab
2 changed files with 51 additions and 0 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

26
examples/sentiments/Modelfile Normal file
View file

@ -0,0 +1,26 @@
# Modelfile for creating a sentiment analyzer.
# Run `ollama create sentiments -f pathtofile` and then `ollama run sentiments` and enter a topic
FROM orca
TEMPLATE """
{{- if .First }}
### System:
You are a sentiment analyzer. You will receive text and output only one word, either POSITIVE or NEGATIVE or NEUTRAL, depending on the sentiment of the text. Here are three examples:
### User:
I hate it when my phone dies
### Response:
NEGATIVE
### User:
He is awesome
### Response:
POSITIVE
### User:
This is the link to the article
### Response:
NEUTRAL
{{- end }}
### User:
{{ .Prompt }}
### Response:
"""

25
examples/sentiments/Readme.md Normal file
View file

@ -0,0 +1,25 @@
# Sentiments Modelfile
This is a simple sentiments analyzer using the Orca model. When you pull Orca from the registry, it has a Template already defined that looks like this:
```Modelfile
{{- if .First }}
### System:
{{ .System }}
{{- end }}
### User:
{{ .Prompt }}
### Response:
```
If we just wanted to have the text:
```Plaintext
You are a sentiment analyzer. You will receive text and output only one word, either POSITIVE or NEGATIVE or NEUTRAL, depending on the sentiment of the text.
```
then we could have put this in a SYSTEM block. But we want to provide examples which require updating the full Template. Any Modelfile you create will inherit all the settings from the source model. But in this example, we are overriding the Template.
When providing examples for the input and output, you should include the way the model usually provides information. Since the Orca model expects a user prompt to appear after ### User: and the response is after ### Response, we should format our examples like that as well. If we were using the Llama 2 model, the format would be a bit different.