Ctrl+K
- ablation_visuals
- dememwm_long_horizon
- dememwm_sourcefree_long_horizon
- diagnostics
- outputs
- results
- slurm_cache
- slurm_logs
- train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k
- train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k
- train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k
- train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k
- train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k
- train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k
- train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k
- train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k
- train_dememwm_sep_mem_fulldit_mem_route_anchor_high_dynamic_revisit_low_a2d4r2_ntok8_8h200_bs16_global128_175k
- train_dememwm_separate_memory_gatefix_4layers512_valid500_berzelius_8a100_bs16_global128_350k
- train_dememwm_separate_memory_gatefix_alllayers1024_valid500_berzelius_8a100_bs16_global128_350k
- train_dememwm_sourcefree_age_count_off_constant_8h200_bs16_global128_seed42_200k
- train_dememwm_sourcefree_age_count_off_cosine_8h200_bs16_global128_seed42_200k
- train_dememwm_sourcefree_age_off_8h200_bs16_global128_seed42_200k
- train_dememwm_sourcefree_noise_off_8h200_bs16_global128_seed42_200k
- train_dememwm_sourcefree_ref_8h200_bs16_global128_seed42_200k
- train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k
- train_dememwm_unified_dense_memory_ddp_selection_geometry_8h200_bs16_global128_200k
- train_dememwm_worldmem_additive_p0_worldmem_homogeneous_8h200_bs16_global128_seed42_train200k_eval198k_gita7a6a844c5ac
- train_dememwm_worldmem_additive_p1_anchor_revisit_8h200_bs8_accum2_global128_seed42_train200k_eval198k_git75d5aa1fef52_job85135
- train_dememwm_worldmem_additive_p2_complementary_revisit_8h200_bs8_accum2_global128_seed42_train200k_eval198k_git75d5aa1fef52_job85135
- train_dememwm_worldmem_additive_p4_unified_role_reader_8h200_bs8_accum2_global128_seed42_train200k_eval198k_git75d5aa1fef52_job85135
- wandb_validation
- worldmem_long_horizon
- worldmem_native_p0_8h200_bs8_global64_seed42_175k
- worldmem_native_p1_8h200_bs8_global64_seed42_175k
- 2.97 MB