Lanxiang Hu's picture

Lanxiang Hu PRO

Snyhlxde

·

https://snyhlxde1.github.io/

snyhlxde1

AI & ML interests

Machine learning systems, distributed learning, efficient machine learning on edge devices

Recent Activity

upvoted a paper 11 days ago

Efficient Long-context Language Model Training by Core Attention Disaggregation

updated a model 11 days ago

multi-token/1020-lx-math-4b-math-n16w16

published a model 13 days ago

multi-token/1020-lx-math-4b-math-n16w16

View all activity

Organizations

Snyhlxde 's models 41

Snyhlxde/shiftedattn-9-21-qwen25coder-instr-from-scratch-n64w32-lr1-e6

Updated Sep 23 • 2

Snyhlxde/math_8b_tool

4B • Updated Sep 21 • 1

Snyhlxde/code-8b-single

4B • Updated Sep 21 • 1

Snyhlxde/math-8b-reasoning

4B • Updated Sep 21 • 1

Snyhlxde/sdk_1.7B_single_policy

Updated Sep 20 • 2

Snyhlxde/sdk_1.7B_single_agent

Updated Sep 20 • 2

Snyhlxde/code_baseline

4B • Updated Sep 20 • 2

Snyhlxde/code_two_policy_generation

4B • Updated Sep 20 • 2

Snyhlxde/code_two_policy_test

4B • Updated Sep 20 • 2

Snyhlxde/openthought2_mathsubset

Snyhlxde/aligned-8-24-7B-ntok64_ce_soft_loss_length_20k_flexattn_data_v1_4k_smp_ar_10_lr5-6

8B • Updated Aug 24

Snyhlxde/aligned-8-24-7B-ntok64_ce_soft_loss_length_20k_flexattn_data_v1_4k_smp_ar_1_only_lr5-6

8B • Updated Aug 24

Snyhlxde/8-23-openthinker2-7B-ntok64_ce_soft_loss_length_capped_16k_flexattn_data_v1_4kexample_ar_10

8B • Updated Aug 24

Snyhlxde/8-23-openthinker2-7B-ntok64_ce_soft_loss_length_capped_16k_flexattn_data_v1_4kexample_ar_only

8B • Updated Aug 24 • 1

Snyhlxde/openthinker2-7B-ntok64_soft_ce_loss_length_20k_flexattn_data_v1_32k_sample_ar_10_c_d

8B • Updated Aug 22

Snyhlxde/openthinker2-7B-ntok64_soft_ce_loss_length_20k_flexattn_data_v1_4k_sample_ar_10_c_d

8B • Updated Aug 21

Snyhlxde/soft_loss_length_capped_16k_flexattn_data_v1_4k4samples_8_16_no_prompt_boundary

Snyhlxde/openthoughts13k-keep_ratio0.15_soft_cllm_loss_downsampling_training

8B • Updated Aug 5 • 3

Snyhlxde/without_think_format_tetris_3by3_0514

Snyhlxde/with_think_format_tetris_3by3_0514

Snyhlxde/without_think_format_tetris_0513

Snyhlxde/with_think_format_tetris_0513

Snyhlxde/with_think_format_tetris_0512

Snyhlxde/without_think_format_tetris_0511

Snyhlxde/deprecated_tetris_dev

Snyhlxde/DeepSeek-R1-Distill-Llama-8B-openthought-vllm

Snyhlxde/base_207M_v1_decay0p99_2

Snyhlxde/base_207M_v1_fliprob0.03_decay1p0

Snyhlxde/base_207M_v1_decay0p99

Snyhlxde/large_682M_v1_decay0p98_2