commit bdb5c09d1ad64dcf8a2db0266ae04f4b065defc2
parent d195ecc2648dc74c89e96ed15110ef3dffe19724
Author: Oliver Lowe <o@olowe.co>
Date: Sun, 29 Sep 2024 14:39:17 +1000
llama: try out llama 3.2 3B by default
Diffstat:
2 files changed, 9 insertions(+), 9 deletions(-)
diff --git a/bin/llama b/bin/llama
@@ -5,13 +5,13 @@ import os
import sys
import urllib.request
-# these models hallucinate
-# model = "llama3.1-8b-instant"
-# model = "llama-3.1-70b-versatile"
-
url = "https://api.groq.com/openai/v1/chat/completions"
-model = "llama3-8b-8192"
-big = "llama3-70b-8192"
+# small models include:
+# llama-3.1-8b-instant
+# llama-3.2-3b-preview
+# llama-3.2-1b-preview
+model = "llama-3.2-3b-preview"
+big = "llama-3.1-70b-versatile"
def read_token(name):
with open(name) as f:
@@ -28,7 +28,7 @@ message = {"messages": [{"role": "user","content": prompt}], "model": model}
req = urllib.request.Request(url, json.dumps(message).encode())
req.add_header("Content-Type", "application/json")
req.add_header("Authorization", "Bearer "+token)
-# groq blocks urllib's user agent?!
+# groq blocks urllib's user agent
req.add_header("User-Agent", "curl/8.9.0")
with urllib.request.urlopen(req) as resp:
diff --git a/man/llama.1 b/man/llama.1
@@ -11,7 +11,7 @@
reads a prompt from the standard input
and sends it to a large language model hosted by Groq.
The reply is written to the standard output.
-The default model is Llama 3 8B.
+The default model is Llama 3.2 3B.
.Pp
A Groq API token must be written to
.Pa $HOME/.config/groq/token .
@@ -19,7 +19,7 @@ A Groq API token must be written to
The following flags are understood:
.Bl -tag -width Ds
.It Fl b
-Prompt the "bigger" 70B model.
+Prompt the "bigger" LLama 3.1 70B model.
.Sh EXAMPLE
.Dl echo 'What is LLM slop?' | llama
.Sh EXIT STATUS
