cerebras

app.py

@@ -59,9 +59,6 @@ def execute_jupyter_agent(
         api_key=HF_TOKEN,
     )
 
-    tokenizer = AutoTokenizer.from_pretrained(model)
-    # model = "meta-llama/Llama-3.1-8B-Instruct"
-
     filenames = []
     if files is not None:
         for filepath in files:
@@ -89,7 +86,7 @@ def execute_jupyter_agent(
     print("history:", message_history)
 
     for notebook_html, notebook_data, messages in run_interactive_notebook(
-        client, model, tokenizer, message_history, sbx, max_new_tokens=max_new_tokens
+        client, model, message_history, sbx, max_new_tokens=max_new_tokens
     ):
         message_history = messages
         
@@ -159,10 +156,10 @@ with gr.Blocks() as demo:
             )
 
             model = gr.Dropdown(
-                value="Qwen/Qwen3-Coder-480B-A35B-Instruct",
+                value="Qwen/Qwen3-Coder-480B-A35B-Instruct:cerebras",
                 choices=[
                     "Qwen/Qwen3-Coder-30B-A3B-Instruct",
-                    "Qwen/Qwen3-Coder-480B-A35B-Instruct",
+                    "Qwen/Qwen3-Coder-480B-A35B-Instruct:cerebras",
                 ],
                 label="Models"
             )

utils.py

@@ -82,7 +82,7 @@ def parse_exec_result_llm(execution, max_code_output=1000):
     return "\n".join(output)
 
 
-def run_interactive_notebook(client, model, tokenizer, messages, sbx, max_new_tokens=512):
+def run_interactive_notebook(client, model, messages, sbx, max_new_tokens=512):
     notebook = JupyterNotebook(messages)
     sbx_info = sbx.get_info()
     notebook.add_sandbox_countdown(sbx_info.started_at, sbx_info.end_at)