You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
@@ -53,8 +52,9 @@ The LLM has access to a single tool:
53
52
54
53
### Examples
55
54
56
-
```bash
55
+
57
56
# List files in current directory
57
+
```
58
58
./simple-function-call -m llama-3.2-1b.gguf -p "list all files in this directory" # I've had success even with this small model but I think it has trouble solving complex tasks it can still generate proper JSON for execution
./simple-function-call -m llama-2-7b.gguf -p "check the current time" # might produce some JSON
129
131
{
130
132
"type": "function",
@@ -173,8 +175,9 @@ Response:
173
175
} # The command in tried to use was probably too complex for the actual tool
174
176
}
175
177
main: decoded 134 tokens in 5.85 s, speed: 22.89 t/s # A better parser bigger model should produce better results
176
-
177
-
# You'll have less of these types of problems but you can't still nudge the smaller ones, if you have some knowledge you can always tell it to run commands directly but it defeats the purpose of it a bit, it would be good with some Speech to Text system where you wouldn't have to type but just say what you want the computer to do
178
+
```
179
+
# You'll have less of these types of problems with bigger models but you can't still nudge the smaller ones, if you have some knowledge you can always tell it to run commands directly but it defeats the purpose of it a bit, it would be good with some Speech to Text system where you wouldn't have to type but just say what you want the computer to do
180
+
```
178
181
./simple-function-call -m llama-3.2-1b-instruct.gguf -p "Check the current date" # And it will run the date command
0 commit comments