From acef4af21aacb8680a27d2431e8d81640ed72b77 Mon Sep 17 00:00:00 2001
From: Davidzhangyuanhan <704464079@qq.com>
Date: Fri, 11 Oct 2024 16:05:01 +0800
Subject: [PATCH] chore: update checkpoint for llava-onevision-qwen2-72b-ov and
 llava-onevision-qwen2-7b-ov

---
 scripts/video/train/SO400M_Qwen2_72B_ov_to_video_am9.sh | 2 +-
 scripts/video/train/SO400M_Qwen2_7B_ov_to_video_am9.sh  | 4 ++--
 2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/scripts/video/train/SO400M_Qwen2_72B_ov_to_video_am9.sh b/scripts/video/train/SO400M_Qwen2_72B_ov_to_video_am9.sh
index 2b74faeff..9dbfdae05 100755
--- a/scripts/video/train/SO400M_Qwen2_72B_ov_to_video_am9.sh
+++ b/scripts/video/train/SO400M_Qwen2_72B_ov_to_video_am9.sh
@@ -26,7 +26,7 @@ echo "BASE_RUN_NAME: ${BASE_RUN_NAME}"
 # Stage 2
 PROMPT_VERSION="qwen_1_5"
 MID_RUN_NAME="llavanext-${VISION_MODEL_VERSION_CLEAN}-${LLM_VERSION_CLEAN}-ov_to_video_am9"
-PREV_STAGE_CHECKPOINT="lmms-lab/llava-onevision-qwen2-72b-ov"
+PREV_STAGE_CHECKPOINT="lmms-lab/llava-onevision-qwen2-72b-ov-si"
 echo "PREV_STAGE_CHECKPOINT: ${PREV_STAGE_CHECKPOINT}"
 echo "MID_RUN_NAME: ${MID_RUN_NAME}"
 
diff --git a/scripts/video/train/SO400M_Qwen2_7B_ov_to_video_am9.sh b/scripts/video/train/SO400M_Qwen2_7B_ov_to_video_am9.sh
index 9f5f48989..0a72bc134 100755
--- a/scripts/video/train/SO400M_Qwen2_7B_ov_to_video_am9.sh
+++ b/scripts/video/train/SO400M_Qwen2_7B_ov_to_video_am9.sh
@@ -26,7 +26,7 @@ echo "BASE_RUN_NAME: ${BASE_RUN_NAME}"
 # Stage 2
 PROMPT_VERSION="qwen_1_5"
 MID_RUN_NAME="llavanext-${VISION_MODEL_VERSION_CLEAN}-${LLM_VERSION_CLEAN}-ov_to_video_am9"
-PREV_STAGE_CHECKPOINT="lmms-lab/llava-onevision-qwen2-7b-ov"
+PREV_STAGE_CHECKPOINT="lmms-lab/llava-onevision-qwen2-7b-ov-si"
 echo "PREV_STAGE_CHECKPOINT: ${PREV_STAGE_CHECKPOINT}"
 echo "MID_RUN_NAME: ${MID_RUN_NAME}"
 
@@ -75,7 +75,7 @@ ACCELERATE_CPU_AFFINITY=1 torchrun --nproc_per_node="${ARNOLD_WORKER_GPU}" --nno
     --torch_compile True \
     --torch_compile_backend "inductor" \
     --dataloader_drop_last True \
-    --frames_upbound 110 \
+    --frames_upbound 64 \
     --mm_newline_position grid \
     --add_time_instruction True \
     --force_sample True \