allenai / allenai/OLMo-core

OLMo evaluation tasks not found

未關閉
#638 0 則留言 0 個 reaction 已指派 0 人 在 GitHub 檢視
主要語言
Python
星號
1.5k
分支
315
平均合併
1 天 9 小時
30 天內合併 PR
11

描述

Hi,

Thank you for your great work. I am using your Docker image (docker://ghcr.io/allenai/olmo-core:tch270cu128-v2.1.0) and trying to launch a pretraining run for the 7B model (src/scripts/official/OLMo3/OLMo-3-1025-7B-pretrain-1.py). However, I encountered errors during the downstream evaluation step. The error message says that the tasks do not exist, so I checked list_tasks in the olmo_eval package — but none of the listed tasks have a corresponding implementation available.

Could this be a configuration issue?

Below is my code for checking task availability:

- Checking code:
```python
from olmo_eval import list_tasks

all_tasks = list_tasks()
print(f"Total tasks available: {len(all_tasks)}\n")

needed_tasks = [
"arc_challenge_test_bpb_5shot",
"arc_challenge_test_mc_5shot_fast",
"arc_easy_test_bpb_5shot",
"arc_easy_test_mc_5shot_fast",
"hellaswag_bpb_5shot",
"mmlu_humanities_test_bpb_5shot",
"mmlu_humanities_test_mc_5shot_fast",
"mmlu_other_test_bpb_5shot",
"mmlu_other_test_mc_5shot_fast",
"mmlu_social_sciences_test_bpb_5shot",
"mmlu_social_sciences_test_mc_5shot_fast",
"mmlu_stem_test_bpb_5shot",
"mmlu_stem_test_mc_5shot_fast",
# Basic Skills
"basic_skills_arithmetic_rc_5shot",
"basic_skills_coding_rc_5shot",
"basic_skills_common_knowledge_rc_5shot",
"basic_skills_logical_reasoning_rc_5shot",
"basic_skills_pattern_rc_5shot",
"basic_skills_string_operations_rc_5shot",
# Gen tasks BPB
"codex_humaneval_gold_bpb_3shot",
"codex_mbpp_gold_bpb_3shot",
"minerva_math_500_gold_bpb_0shot",
"mt_mbpp_cpp_gold_bpb_3shot",
"mt_mbpp_java_gold_bpb_3shot",
"mt_mbpp_rust_gold_bpb_3shot",
# Sanity check for MCQA ability
"copycolors_10way_fast",
]

print("Checking for needed tasks:")
for task in needed_tasks:
status = "✓ FOUND" if task in all_tasks else "✗ MISSING"
print(f" {status}: {task}")
```

- Results:
```bash
Total tasks available: 167

Checking for needed tasks:
✗ MISSING: arc_challenge_test_bpb_5shot
✗ MISSING: arc_challenge_test_mc_5shot_fast
✗ MISSING: arc_easy_test_bpb_5shot
✗ MISSING: arc_easy_test_mc_5shot_fast
✗ MISSING: hellaswag_bpb_5shot
✗ MISSING: mmlu_humanities_test_bpb_5shot
✗ MISSING: mmlu_humanities_test_mc_5shot_fast
✗ MISSING: mmlu_other_test_bpb_5shot
✗ MISSING: mmlu_other_test_mc_5shot_fast
✗ MISSING: mmlu_social_sciences_test_bpb_5shot
✗ MISSING: mmlu_social_sciences_test_mc_5shot_fast
✗ MISSING: mmlu_stem_test_bpb_5shot
✗ MISSING: mmlu_stem_test_mc_5shot_fast
✗ MISSING: basic_skills_arithmetic_rc_5shot
✗ MISSING: basic_skills_coding_rc_5shot
✗ MISSING: basic_skills_common_knowledge_rc_5shot
✗ MISSING: basic_skills_logical_reasoning_rc_5shot
✗ MISSING: basic_skills_pattern_rc_5shot
✗ MISSING: basic_skills_string_operations_rc_5shot
✗ MISSING: codex_humaneval_gold_bpb_3shot
✗ MISSING: codex_mbpp_gold_bpb_3shot
✗ MISSING: minerva_math_500_gold_bpb_0shot
✗ MISSING: mt_mbpp_cpp_gold_bpb_3shot
✗ MISSING: mt_mbpp_java_gold_bpb_3shot
✗ MISSING: mt_mbpp_rust_gold_bpb_3shot
✗ MISSING: copycolors_10way_fast
```

貢獻指南

開啟貢獻指南

評估

這個 Issue 還沒有評估資料。

把新 issue 寄到你的電子郵件信箱

精選適合新手參與的 GitHub issue 摘要。