Skip to content

[VLM multi-turn] delete fsdp (#1141) #1463

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
gxlvera wants to merge 1 commit into
base: main
Choose a base branch
from
+3 −13
Open

[VLM multi-turn] delete fsdp (#1141) #1463

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion examples/geo3k_vlm_multi_turn/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -30,7 +30,7 @@ The reward model is the default math RM.
export WANDB_API_KEY=...
export SLIME_SCRIPT_MODEL_NAME=Qwen3-VL-2B-Instruct
export SLIME_SCRIPT_NUM_GPUS=4
export SLIME_SCRIPT_TRAIN_BACKEND=fsdp
export SLIME_SCRIPT_TRAIN_BACKEND=megatron

# 2) Download the dataset
hf download --repo-type dataset VeraIsHere/geo3k_imgurl_processed --local-dir /root/datasets/geo3k_imgurl_processed
Expand Down
Binary file modified examples/geo3k_vlm_multi_turn/geo3k_vlm_multi_turn_reward.png
View file
Open in desktop
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Binary file modified examples/geo3k_vlm_multi_turn/rollout_experiment_result_megatron.png
View file
Open in desktop
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
14 changes: 2 additions & 12 deletions examples/geo3k_vlm_multi_turn/run_geo3k_vlm_multi_turn.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -15,8 +15,8 @@

NUM_GPUS = int(os.environ.get("SLIME_SCRIPT_NUM_GPUS", "4"))
EXTERNAL_RAY = int(os.environ.get("SLIME_SCRIPT_EXTERNAL_RAY", "0"))
TRAIN_BACKEND = os.environ.get("SLIME_SCRIPT_TRAIN_BACKEND", "fsdp").lower()
assert TRAIN_BACKEND in {"fsdp", "megatron"}
TRAIN_BACKEND = os.environ.get("SLIME_SCRIPT_TRAIN_BACKEND", "megatron").lower()
assert TRAIN_BACKEND in {"megatron"}

DATASET_NAME = "VeraIsHere/geo3k_imgurl_processed"
DATA_ROOT = "/root/datasets/geo3k_imgurl_processed"
Expand Down Expand Up @@ -104,13 +104,6 @@ def execute():
f"--sglang-cuda-graph-bs {' '.join(map(str, [1, 2, 4, 8] + list(range(16, 257, 8))))} "
)

fsdp_args = (
"--train-backend fsdp "
"--gradient-checkpointing "
"--sglang-attention-backend fa3 "
"--attn-implementation flash_attention_3 "
"--update-weight-buffer-size 536870912 "
)

megatron_args = (
"--train-backend megatron "
Expand Down Expand Up @@ -142,9 +135,6 @@ def execute():
backend_args = megatron_args
megatron_model_type = get_megatron_model_type(MODEL_NAME)
os.environ["MODEL_ARGS_ROTARY_BASE"] = "5000000"
else:
backend_args = fsdp_args
megatron_model_type = None

train_args = (
f"{ckpt_args} "
Expand Down