[PR #723] Remove hard-coded trust_remote_code=True #294

New issue

Open

opened 2026-10-26 21:05:30 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-10-26 21:05:30 +00:00

Owner

📋 Pull Request Information

Original PR: https://github.com/deepseek-ai/DeepSeek-Coder/pull/723
Author: @Dextheking1
Created: 10/23/2026
Status: 🔄 Open

Base: main ← Head: remove-hardcoded-trust-remote-code

📝 Commits (1)

5699e12 remove hard-coded trust_remote_code=True from model loading

📊 Changes

8 files changed (+18 additions, -20 deletions)

View changed files

📝 Evaluation/HumanEval/eval_pal.py (+1 -1)
📝 Evaluation/HumanEval/humaneval.py (+1 -1)
📝 Evaluation/LeetCode/vllm_inference.py (+1 -2)
📝 Evaluation/MBPP/eval_pal.py (+1 -1)
📝 Evaluation/MBPP/mbpp.py (+1 -1)
📝 Evaluation/PAL-Math/run.py (+2 -2)
📝 README.md (+10 -10)
📝 finetune/finetune_deepseekcoder.py (+1 -2)

📄 Description

Loading a checkpoint should not silently run code from the checkpoint repo.

This removes trust_remote_code=True from the eval scripts, the finetune script, and the README examples. Model and tokenizer loading now uses the default, which does not execute remote code.

Anyone with a checkpoint that needs remote code can still pass the flag in their own call.

Changed files:

Evaluation/HumanEval/eval_pal.py
Evaluation/HumanEval/humaneval.py
Evaluation/MBPP/eval_pal.py
Evaluation/MBPP/mbpp.py
Evaluation/PAL-Math/run.py
Evaluation/LeetCode/vllm_inference.py
finetune/finetune_deepseekcoder.py
README.md

Syntax checked the changed python files before pushing.

_{🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.}

## 📋 Pull Request Information **Original PR:** https://github.com/deepseek-ai/DeepSeek-Coder/pull/723 **Author:** [@Dextheking1](https://github.com/Dextheking1) **Created:** 10/23/2026 **Status:** 🔄 Open **Base:** `main` ← **Head:** `remove-hardcoded-trust-remote-code` --- ### 📝 Commits (1) - [`5699e12`](https://github.com/deepseek-ai/DeepSeek-Coder/commit/5699e12430e0a2d62a5234249edb7088c564376d) remove hard-coded trust_remote_code=True from model loading ### 📊 Changes **8 files changed** (+18 additions, -20 deletions) <details> <summary>View changed files</summary> 📝 `Evaluation/HumanEval/eval_pal.py` (+1 -1) 📝 `Evaluation/HumanEval/humaneval.py` (+1 -1) 📝 `Evaluation/LeetCode/vllm_inference.py` (+1 -2) 📝 `Evaluation/MBPP/eval_pal.py` (+1 -1) 📝 `Evaluation/MBPP/mbpp.py` (+1 -1) 📝 `Evaluation/PAL-Math/run.py` (+2 -2) 📝 `README.md` (+10 -10) 📝 `finetune/finetune_deepseekcoder.py` (+1 -2) </details> ### 📄 Description Loading a checkpoint should not silently run code from the checkpoint repo. This removes trust_remote_code=True from the eval scripts, the finetune script, and the README examples. Model and tokenizer loading now uses the default, which does not execute remote code. Anyone with a checkpoint that needs remote code can still pass the flag in their own call. Changed files: - Evaluation/HumanEval/eval_pal.py - Evaluation/HumanEval/humaneval.py - Evaluation/MBPP/eval_pal.py - Evaluation/MBPP/mbpp.py - Evaluation/PAL-Math/run.py - Evaluation/LeetCode/vllm_inference.py - finetune/finetune_deepseekcoder.py - README.md Syntax checked the changed python files before pushing. --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

bsakaguchi added the

pull-request

label

2026-10-26 21:05:30 +00:00