iomgaa
953fb7a456
零参数全量 900 题、N_SAMPLES smoke test、TASK_TYPES 筛选、版本覆盖。 Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
42 lines
1.3 KiB
Bash
Executable File
42 lines
1.3 KiB
Bash
Executable File
#!/usr/bin/env bash
|
|
# Video-MME 推理实验（TRM5 基线）
|
|
# 职责：在 900 道 Video-MME 题上跑全量推理，生成基线记录。
|
|
# 用法:
|
|
# bash scripts/infer_video_mme.sh # 全量 900 题
|
|
# N_SAMPLES=10 bash scripts/infer_video_mme.sh # smoke test
|
|
# TASK_TYPES="Action Recognition,Counting Problem" bash scripts/infer_video_mme.sh # 只跑指定题型
|
|
# SKILLS_VERSION=v2 PROMPTS_VERSION=v2 bash scripts/infer_video_mme.sh # 指定版本
|
|
set -euo pipefail
|
|
|
|
cd "$(dirname "$0")/.."
|
|
|
|
CUDA_VISIBLE_DEVICES="${CUDA_VISIBLE_DEVICES:-0}"
|
|
export CUDA_VISIBLE_DEVICES
|
|
|
|
# 加载环境变量（API key 等）
|
|
set -a
|
|
source .env
|
|
set +a
|
|
|
|
PYTHON="$(conda run -n Video-Tree-TRM which python)"
|
|
|
|
EXTRA_ARGS=()
|
|
|
|
if [[ -n "${TASK_TYPES:-}" ]]; then
|
|
IFS=',' read -ra TYPES <<< "${TASK_TYPES}"
|
|
EXTRA_ARGS+=(--task-types "${TYPES[@]}")
|
|
fi
|
|
|
|
"${PYTHON}" main.py \
|
|
--workspace-dir workspaces/default \
|
|
--store-dir store \
|
|
--mode infer \
|
|
--concurrency 24 \
|
|
--max-steps 40 \
|
|
--skill-mode auto \
|
|
--n-samples "${N_SAMPLES:-0}" \
|
|
--questions benchmarks/Video-MME \
|
|
--skills-version "${SKILLS_VERSION:-v1}" \
|
|
--prompts-version "${PROMPTS_VERSION:-v1}" \
|
|
"${EXTRA_ARGS[@]}"
|