update sglang && support qwen3 next

[sleepcoo]

Motivation

sgl-project/sglang#14607

Modifications

Related Issues

Accuracy Test

Benchmark & Profiling

[gemini-code-assist]

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!

[ggg-s]

@sleepcoo
serve command:
CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 torchrun --nproc_per_node 8
scripts/prepare_hidden_states.py
--target-model-path /models/Qwen3-Next-80B-A3B-Instruct
--data-path dataset/sa.jsonl
--output-path cache/hidden_states/qwen3_next_80b_a3b_sa_2048
--chat-template qwen
--max-length 2048
--tp-size 8
--batch-size 1
--enable-aux-hidden-states
--sglang-mem-fraction-static 0.9
--num-samples 10
error：AttributeError: 'Qwen3NextForCausalLM' object has no attribute 'set_eagle3_layers_to_capture'

[sleepcoo]

@sleepcoo serve command: CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 torchrun --nproc_per_node 8 scripts/prepare_hidden_states.py --target-model-path /models/Qwen3-Next-80B-A3B-Instruct --data-path dataset/sa.jsonl --output-path cache/hidden_states/qwen3_next_80b_a3b_sa_2048 --chat-template qwen --max-length 2048 --tp-size 8 --batch-size 1 --enable-aux-hidden-states --sglang-mem-fraction-static 0.9 --num-samples 10 error：AttributeError: 'Qwen3NextForCausalLM' object has no attribute 'set_eagle3_layers_to_capture'

sgl-project/sglang#14607

[ggg-s]

@sleepcoo thank u. I got it