master

ascend-robottest(python): update dpo st

e531563c创建于 1 天前历史提交

文件	最后提交记录	最后更新时间
deepseek32_tp1_pp2_vpp1_ep4.sh	fix(pytorch):add ckpt-format argument to scripts	2 个月前
deepseek4_flash_mcore_tp1_pp1_ep8.sh	test(torch): add deepseekv4 st	13 天前
deepseek_500b_tp1_pp2_ep2_cp2_overlap.sh	fix(pytorch):add ckpt-format argument to scripts	2 个月前
deepseek_v2_mcore_tp1_pp1_ep8.sh	fix(pytorch):add ckpt-format argument to scripts	2 个月前
deepseek_v3_mcore_tp1_pp2_ep4.sh	fix(pytorch):add ckpt-format argument to scripts	2 个月前
dpo_qwen3_8b_tp2_pp1_cyclic_pairwise.sh	test(python): update dpo st	1 天前
mamba2_8b_tp4_pp1_cp2_recompute_4k_ptd.sh	fix(pytorch):add ckpt-format argument to scripts	2 个月前
pretrain_qwen3_8b_4K_fsdp2.yaml	feat(pytorch):ring cp support MLA/GQA	1 个月前
pretrain_qwen3_8b_4k_fsdp2.sh	[pytorch][ci]Add pretrain and finetune ST for FSDP2 backend	3 个月前
pretrain_qwen3_8b_4k_hybridcp_gqa.sh	test(megatron):replace chatglm3_gqa_cp4 testcase with qwen3-8b	2 天前
pretrain_qwen3_8b_general_cp_ring.sh	test(megatron): replace llama2_tp2_cp4_general_double_ring testcase with qwen3-8b	1 个月前
qwen3_30b_tp4_cp2_ep2.sh	feat: change ST	6 天前
qwen3_8b_tp2_pp2_vpp2_noop_layer.sh	feat(torch):add st	16 天前
qwen3_8b_tp2_pp4_vpp2.sh	feat: change ST's llama2_tp4pp2vpp2_tp2d_tpx2tpy2.sh to qwen3_8b_tp2_pp4_vpp2.sh	2 天前
qwen3_8b_tp2_pp4_vpp2_swap.sh	test(megatron): replace llama2_tp2_pp4_vpp2_swap testcase with qwen3-8b	2 天前
qwen3_next_80b_4K_A3_ptd.sh	fix(pytorch):add ckpt-format argument to scripts	2 个月前
qwen3_next_ascendc-GDN_pack_cp2.sh	test(pytorch): add st for AscendC GDN kernels, including cp and pack.	1 天前
qwen3_next_ascendc-GDN_pack_cp2.yaml	test(pytorch): add st for AscendC GDN kernels, including cp and pack.	1 天前
tune_gpt_oss_20b_a3b_4K_fsdp2.yaml	docs(megatron):update testcase develop rule	2 个月前
tune_gpt_oss_20b_a3b_4k_fsdp2.sh	[pytorch][ci]Add pretrain and finetune ST for FSDP2 backend	3 个月前
tune_qwen3_8b_lora_tp2pp1.sh	test(megatron): replace_llama3_lora_with_qwen3_8b	30 天前
tune_qwen3_8b_qlora_ptd.sh	test(megatron): replace tune_llama2_tp1_pp1_qlora testcase with qwen3-8b	29 天前