| [{"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt100000.pt", "type": "ablation", "layer": 0, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000/ablation_layer0.npz", "name": "plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000_ablation_L0", "itr": 100000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt100000.pt", "type": "ablation", "layer": 1, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000/ablation_layer1.npz", "name": "plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000_ablation_L1", "itr": 100000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt100000.pt", "type": "baseline", "layer": 0, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000/baseline.npz", "name": "plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000_baseline", "itr": 100000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt100000.pt", "type": "cinclogits", "layer": 0, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000/cinclogits_layer0.npz", "name": "plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000_cinclogits_L0", "itr": 100000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt100000.pt", "type": "cinclogits", "layer": 1, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000/cinclogits_layer1.npz", "name": "plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000_cinclogits_L1", "itr": 100000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt100000.pt", "type": "hijack", "layer": 0, "trials": 2000, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000/hijack.npz", "name": "plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000_hijack", "itr": 100000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt100000.pt", "type": "intensity", "layer": 0, "ub": 5, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000/intensity_layer0.npz", "name": "plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000_intensity_ub5_L0", "itr": 100000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt100000.pt", "type": "intensity", "layer": 0, "ub": 10, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000/intensity_layer0_ub10.npz", "name": "plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000_intensity_ub10_L0", "itr": 100000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt100000.pt", "type": "intensity", "layer": 0, "ub": 15, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000/intensity_layer0_ub15.npz", "name": "plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000_intensity_ub15_L0", "itr": 100000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt100000.pt", "type": "intensity", "layer": 0, "ub": 20, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000/intensity_layer0_ub20.npz", "name": "plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000_intensity_ub20_L0", "itr": 100000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt100000.pt", "type": "intensity", "layer": 0, "ub": 30, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000/intensity_layer0_ub30.npz", "name": "plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000_intensity_ub30_L0", "itr": 100000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt100000.pt", "type": "intensity", "layer": 0, "ub": 50, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000/intensity_layer0_ub50.npz", "name": "plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000_intensity_ub50_L0", "itr": 100000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt100000.pt", "type": "intensity", "layer": 0, "ub": 60, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000/intensity_layer0_ub60.npz", "name": "plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000_intensity_ub60_L0", "itr": 100000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt100000.pt", "type": "intensity", "layer": 1, "ub": 5, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000/intensity_layer1.npz", "name": "plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000_intensity_ub5_L1", "itr": 100000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt100000.pt", "type": "intensity", "layer": 1, "ub": 10, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000/intensity_layer1_ub10.npz", "name": "plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000_intensity_ub10_L1", "itr": 100000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt100000.pt", "type": "intensity", "layer": 1, "ub": 15, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000/intensity_layer1_ub15.npz", "name": "plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000_intensity_ub15_L1", "itr": 100000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt100000.pt", "type": "intensity", "layer": 1, "ub": 20, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000/intensity_layer1_ub20.npz", "name": "plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000_intensity_ub20_L1", "itr": 100000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt100000.pt", "type": "intensity", "layer": 1, "ub": 30, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000/intensity_layer1_ub30.npz", "name": "plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000_intensity_ub30_L1", "itr": 100000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt100000.pt", "type": "intensity", "layer": 1, "ub": 50, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000/intensity_layer1_ub50.npz", "name": "plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000_intensity_ub50_L1", "itr": 100000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt100000.pt", "type": "intensity", "layer": 1, "ub": 60, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000/intensity_layer1_ub60.npz", "name": "plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000_intensity_ub60_L1", "itr": 100000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt100000.pt", "type": "intensity_asym", "layer": 0, "unsorted_ub": 60, "unsorted_lb": 0, "unsorted_ub_num": 1, "unsorted_lb_num": 0, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000/intensity_layer0_ub60_lb0.npz", "name": "plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000_asym_ub60_lb0", "itr": 100000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt100000.pt", "type": "intensity_asym", "layer": 0, "unsorted_ub": 0, "unsorted_lb": 60, "unsorted_ub_num": 0, "unsorted_lb_num": 1, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000/intensity_layer0_ub0_lb60.npz", "name": "plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000_asym_ub0_lb60", "itr": 100000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt100000.pt", "type": "separator_random", "layer": 0, "trials": 1000, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000/separator_random.npz", "name": "plots_V256_B16_LR3e-2_MI100000_E64_H1_L2_ds1337_is1337_ckpt100000_sep_rand", "itr": 100000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt500000.pt", "type": "ablation", "layer": 0, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000/ablation_layer0.npz", "name": "plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000_ablation_L0", "itr": 500000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt500000.pt", "type": "ablation", "layer": 1, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000/ablation_layer1.npz", "name": "plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000_ablation_L1", "itr": 500000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt500000.pt", "type": "baseline", "layer": 0, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000/baseline.npz", "name": "plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000_baseline", "itr": 500000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt500000.pt", "type": "cinclogits", "layer": 0, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000/cinclogits_layer0.npz", "name": "plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000_cinclogits_L0", "itr": 500000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt500000.pt", "type": "cinclogits", "layer": 1, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000/cinclogits_layer1.npz", "name": "plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000_cinclogits_L1", "itr": 500000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt500000.pt", "type": "hijack", "layer": 0, "trials": 2000, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000/hijack.npz", "name": "plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000_hijack", "itr": 500000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt500000.pt", "type": "intensity", "layer": 0, "ub": 5, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000/intensity_layer0.npz", "name": "plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000_intensity_ub5_L0", "itr": 500000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt500000.pt", "type": "intensity", "layer": 0, "ub": 10, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000/intensity_layer0_ub10.npz", "name": "plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000_intensity_ub10_L0", "itr": 500000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt500000.pt", "type": "intensity", "layer": 0, "ub": 15, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000/intensity_layer0_ub15.npz", "name": "plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000_intensity_ub15_L0", "itr": 500000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt500000.pt", "type": "intensity", "layer": 0, "ub": 20, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000/intensity_layer0_ub20.npz", "name": "plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000_intensity_ub20_L0", "itr": 500000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt500000.pt", "type": "intensity", "layer": 0, "ub": 30, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000/intensity_layer0_ub30.npz", "name": "plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000_intensity_ub30_L0", "itr": 500000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt500000.pt", "type": "intensity", "layer": 0, "ub": 50, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000/intensity_layer0_ub50.npz", "name": "plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000_intensity_ub50_L0", "itr": 500000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt500000.pt", "type": "intensity", "layer": 0, "ub": 60, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000/intensity_layer0_ub60.npz", "name": "plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000_intensity_ub60_L0", "itr": 500000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt500000.pt", "type": "intensity", "layer": 1, "ub": 5, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000/intensity_layer1.npz", "name": "plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000_intensity_ub5_L1", "itr": 500000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt500000.pt", "type": "intensity", "layer": 1, "ub": 10, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000/intensity_layer1_ub10.npz", "name": "plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000_intensity_ub10_L1", "itr": 500000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt500000.pt", "type": "intensity", "layer": 1, "ub": 15, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000/intensity_layer1_ub15.npz", "name": "plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000_intensity_ub15_L1", "itr": 500000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt500000.pt", "type": "intensity", "layer": 1, "ub": 20, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000/intensity_layer1_ub20.npz", "name": "plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000_intensity_ub20_L1", "itr": 500000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt500000.pt", "type": "intensity", "layer": 1, "ub": 30, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000/intensity_layer1_ub30.npz", "name": "plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000_intensity_ub30_L1", "itr": 500000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt500000.pt", "type": "intensity", "layer": 1, "ub": 50, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000/intensity_layer1_ub50.npz", "name": "plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000_intensity_ub50_L1", "itr": 500000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt500000.pt", "type": "intensity", "layer": 1, "ub": 60, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000/intensity_layer1_ub60.npz", "name": "plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000_intensity_ub60_L1", "itr": 500000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt500000.pt", "type": "intensity_asym", "layer": 0, "unsorted_ub": 60, "unsorted_lb": 0, "unsorted_ub_num": 1, "unsorted_lb_num": 0, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000/intensity_layer0_ub60_lb0.npz", "name": "plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000_asym_ub60_lb0", "itr": 500000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt500000.pt", "type": "intensity_asym", "layer": 0, "unsorted_ub": 0, "unsorted_lb": 60, "unsorted_ub_num": 0, "unsorted_lb_num": 1, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000/intensity_layer0_ub0_lb60.npz", "name": "plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000_asym_ub0_lb60", "itr": 500000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt500000.pt", "type": "separator_random", "layer": 0, "trials": 1000, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000/separator_random.npz", "name": "plots_V256_B16_LR3e-2_MI500000_E64_H1_L2_ds1337_is1337_ckpt500000_sep_rand", "itr": 500000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt900000.pt", "type": "ablation", "layer": 0, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000/ablation_layer0.npz", "name": "plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000_ablation_L0", "itr": 900000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt900000.pt", "type": "ablation", "layer": 1, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000/ablation_layer1.npz", "name": "plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000_ablation_L1", "itr": 900000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt900000.pt", "type": "baseline", "layer": 0, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000/baseline.npz", "name": "plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000_baseline", "itr": 900000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt900000.pt", "type": "cinclogits", "layer": 0, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000/cinclogits_layer0.npz", "name": "plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000_cinclogits_L0", "itr": 900000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt900000.pt", "type": "cinclogits", "layer": 1, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000/cinclogits_layer1.npz", "name": "plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000_cinclogits_L1", "itr": 900000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt900000.pt", "type": "hijack", "layer": 0, "trials": 2000, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000/hijack.npz", "name": "plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000_hijack", "itr": 900000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt900000.pt", "type": "intensity", "layer": 0, "ub": 5, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000/intensity_layer0.npz", "name": "plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000_intensity_ub5_L0", "itr": 900000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt900000.pt", "type": "intensity", "layer": 0, "ub": 10, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000/intensity_layer0_ub10.npz", "name": "plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000_intensity_ub10_L0", "itr": 900000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt900000.pt", "type": "intensity", "layer": 0, "ub": 15, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000/intensity_layer0_ub15.npz", "name": "plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000_intensity_ub15_L0", "itr": 900000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt900000.pt", "type": "intensity", "layer": 0, "ub": 20, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000/intensity_layer0_ub20.npz", "name": "plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000_intensity_ub20_L0", "itr": 900000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt900000.pt", "type": "intensity", "layer": 0, "ub": 30, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000/intensity_layer0_ub30.npz", "name": "plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000_intensity_ub30_L0", "itr": 900000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt900000.pt", "type": "intensity", "layer": 0, "ub": 50, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000/intensity_layer0_ub50.npz", "name": "plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000_intensity_ub50_L0", "itr": 900000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt900000.pt", "type": "intensity", "layer": 0, "ub": 60, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000/intensity_layer0_ub60.npz", "name": "plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000_intensity_ub60_L0", "itr": 900000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt900000.pt", "type": "intensity", "layer": 1, "ub": 5, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000/intensity_layer1.npz", "name": "plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000_intensity_ub5_L1", "itr": 900000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt900000.pt", "type": "intensity", "layer": 1, "ub": 10, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000/intensity_layer1_ub10.npz", "name": "plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000_intensity_ub10_L1", "itr": 900000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt900000.pt", "type": "intensity", "layer": 1, "ub": 15, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000/intensity_layer1_ub15.npz", "name": "plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000_intensity_ub15_L1", "itr": 900000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt900000.pt", "type": "intensity", "layer": 1, "ub": 20, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000/intensity_layer1_ub20.npz", "name": "plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000_intensity_ub20_L1", "itr": 900000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt900000.pt", "type": "intensity", "layer": 1, "ub": 30, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000/intensity_layer1_ub30.npz", "name": "plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000_intensity_ub30_L1", "itr": 900000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt900000.pt", "type": "intensity", "layer": 1, "ub": 50, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000/intensity_layer1_ub50.npz", "name": "plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000_intensity_ub50_L1", "itr": 900000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt900000.pt", "type": "intensity", "layer": 1, "ub": 60, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000/intensity_layer1_ub60.npz", "name": "plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000_intensity_ub60_L1", "itr": 900000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt900000.pt", "type": "intensity_asym", "layer": 0, "unsorted_ub": 60, "unsorted_lb": 0, "unsorted_ub_num": 1, "unsorted_lb_num": 0, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000/intensity_layer0_ub60_lb0.npz", "name": "plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000_asym_ub60_lb0", "itr": 900000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt900000.pt", "type": "intensity_asym", "layer": 0, "unsorted_ub": 0, "unsorted_lb": 60, "unsorted_ub_num": 0, "unsorted_lb_num": 1, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000/intensity_layer0_ub0_lb60.npz", "name": "plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000_asym_ub0_lb60", "itr": 900000}, {"ckpt_path": "/mnt/task_runtime/sort-llm/1000k-checkpoints/sortgpt_k16_methfixed_mlp1_L2_N256_E64_pos0_fln1_wd0p0_lr0p03_dseed1337_iseed1337__ckpt900000.pt", "type": "separator_random", "layer": 0, "trials": 1000, "out": "/mnt/task_runtime/sort-llm/1000k-checkpoints/outputs/tmp_results/plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000/separator_random.npz", "name": "plots_V256_B16_LR3e-2_MI900000_E64_H1_L2_ds1337_is1337_ckpt900000_sep_rand", "itr": 900000}] |