Add files using upload-large-folder tool

LTA_openwebtext_dualt/logs/lm1b_len128_lognormal_atoms_sde_watch/infer_lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522_step_0002000_logistic_normal_t1p45.log

@@ -0,0 +1,74 @@
+[watch-lognormal-sde] 2026-05-22_21:40:17 infer runs/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0002000.pt -> docs/lta_samples/metrics_20260522/lm1b_len128_lognormal_atoms_every1k_logistic_normal_sde_t1p45_steps128_n256/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0002000
+[load] runs/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0002000.pt
+[ckpt] step=2000
+[sde] generated 16/256
+[sde] generated 32/256
+[sde] generated 48/256
+[sde] generated 64/256
+[sde] generated 80/256
+[sde] generated 96/256
+[sde] generated 112/256
+[sde] generated 128/256
+[sde] generated 144/256
+[sde] generated 160/256
+[sde] generated 176/256
+[sde] generated 192/256
+[sde] generated 208/256
+[sde] generated 224/256
+[sde] generated 240/256
+[sde] generated 256/256
+[score] loading scorer: /e2e-data/evad-tech-vla/wanghan58/models/flowtext_scorers/gpt2-large-standard
+[summary] {
+  "type": "summary",
+  "checkpoint": "runs/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0002000.pt",
+  "step": 2000,
+  "decode": {
+    "decode_rule": "logistic_normal_resample_sde",
+    "steps": 128,
+    "model_t_mode": "const0.5",
+    "concentration_min": 1.0,
+    "concentration_max": 1024.0,
+    "endpoint_temp": 1.45,
+    "support_power": 1.0,
+    "semantic_power": 1.0,
+    "noise_init": "logistic_normal",
+    "noise_sigma": 3.0,
+    "noise_dirichlet_concentration": 1.0,
+    "sde_resample": "logistic_normal",
+    "logistic_normal_sigma_min": 0.18,
+    "logistic_normal_sigma_max": 3.0,
+    "logistic_normal_tau_min": 0.65,
+    "logistic_normal_tau_max": 1.0,
+    "final_from": "blend_0.5",
+    "n_samples": 256,
+    "seed": 20260522
+  },
+  "raw_genppl": {
+    "ppl": 49.244736454470555,
+    "nll_per_token": 3.8968024878857603,
+    "tokens": 40229,
+    "kept_samples": 256,
+    "total_samples": 256,
+    "empty_rate": 0.0,
+    "skipped_samples": 0
+  },
+  "stripped_genppl": {
+    "ppl": 71.8270066850532,
+    "nll_per_token": 4.274260543006973,
+    "tokens": 34134,
+    "kept_samples": 256,
+    "total_samples": 256,
+    "empty_rate": 0.0,
+    "skipped_samples": 0
+  },
+  "diversity": {
+    "sample_entropy": 3.7696518059031714,
+    "unique_tokens": 1626,
+    "token_count": 32768,
+    "distinct_1": 0.04962158203125,
+    "distinct_2": 0.25406003937007876,
+    "top_token_mass": 0.0968017578125
+  }
+}
+[done] docs/lta_samples/metrics_20260522/lm1b_len128_lognormal_atoms_every1k_logistic_normal_sde_t1p45_steps128_n256/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0002000/sde_steps128_samples256_scored.jsonl
+[watch-lognormal-sde] 2026-05-22_21:41:45 done step_0002000

LTA_openwebtext_dualt/logs/lm1b_len128_lognormal_atoms_sde_watch/infer_lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522_step_0007000_logistic_normal_t1p45.log

@@ -0,0 +1,74 @@
+[watch-lognormal-sde] 2026-05-22_22:20:41 infer runs/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0007000.pt -> docs/lta_samples/metrics_20260522/lm1b_len128_lognormal_atoms_every1k_logistic_normal_sde_t1p45_steps128_n256/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0007000
+[load] runs/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0007000.pt
+[ckpt] step=7000
+[sde] generated 16/256
+[sde] generated 32/256
+[sde] generated 48/256
+[sde] generated 64/256
+[sde] generated 80/256
+[sde] generated 96/256
+[sde] generated 112/256
+[sde] generated 128/256
+[sde] generated 144/256
+[sde] generated 160/256
+[sde] generated 176/256
+[sde] generated 192/256
+[sde] generated 208/256
+[sde] generated 224/256
+[sde] generated 240/256
+[sde] generated 256/256
+[score] loading scorer: /e2e-data/evad-tech-vla/wanghan58/models/flowtext_scorers/gpt2-large-standard
+[summary] {
+  "type": "summary",
+  "checkpoint": "runs/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0007000.pt",
+  "step": 7000,
+  "decode": {
+    "decode_rule": "logistic_normal_resample_sde",
+    "steps": 128,
+    "model_t_mode": "const0.5",
+    "concentration_min": 1.0,
+    "concentration_max": 1024.0,
+    "endpoint_temp": 1.45,
+    "support_power": 1.0,
+    "semantic_power": 1.0,
+    "noise_init": "logistic_normal",
+    "noise_sigma": 3.0,
+    "noise_dirichlet_concentration": 1.0,
+    "sde_resample": "logistic_normal",
+    "logistic_normal_sigma_min": 0.18,
+    "logistic_normal_sigma_max": 3.0,
+    "logistic_normal_tau_min": 0.65,
+    "logistic_normal_tau_max": 1.0,
+    "final_from": "blend_0.5",
+    "n_samples": 256,
+    "seed": 20260522
+  },
+  "raw_genppl": {
+    "ppl": 34.13540808681281,
+    "nll_per_token": 3.530335205883382,
+    "tokens": 33047,
+    "kept_samples": 256,
+    "total_samples": 256,
+    "empty_rate": 0.0,
+    "skipped_samples": 0
+  },
+  "stripped_genppl": {
+    "ppl": 44.081394070795554,
+    "nll_per_token": 3.786037790270059,
+    "tokens": 27936,
+    "kept_samples": 256,
+    "total_samples": 256,
+    "empty_rate": 0.0,
+    "skipped_samples": 0
+  },
+  "diversity": {
+    "sample_entropy": 3.233755511597012,
+    "unique_tokens": 1809,
+    "token_count": 32768,
+    "distinct_1": 0.055206298828125,
+    "distinct_2": 0.2685162401574803,
+    "top_token_mass": 0.15570068359375
+  }
+}
+[done] docs/lta_samples/metrics_20260522/lm1b_len128_lognormal_atoms_every1k_logistic_normal_sde_t1p45_steps128_n256/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0007000/sde_steps128_samples256_scored.jsonl
+[watch-lognormal-sde] 2026-05-22_22:23:22 done step_0007000

LTA_openwebtext_dualt/logs/lm1b_len128_lognormal_atoms_sde_watch/infer_lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522_step_0013000_logistic_normal_t1p45.log

@@ -0,0 +1,76 @@
+[watch-lognormal-sde] 2026-05-22_23:37:31 infer runs/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0013000.pt -> docs/lta_samples/metrics_20260522/lm1b_len128_lognormal_atoms_every1k_logistic_normal_sde_t1p45_steps128_n256/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0013000
+[load] runs/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0013000.pt
+[ckpt] step=13000
+[sde] generated 16/256
+[sde] generated 32/256
+[sde] generated 48/256
+[sde] generated 64/256
+[sde] generated 80/256
+[sde] generated 96/256
+[sde] generated 112/256
+[sde] generated 128/256
+[sde] generated 144/256
+[sde] generated 160/256
+[sde] generated 176/256
+[sde] generated 192/256
+[sde] generated 208/256
+[sde] generated 224/256
+[sde] generated 240/256
+[sde] generated 256/256
+[score] loading scorer: /e2e-data/evad-tech-vla/wanghan58/models/flowtext_scorers/gpt2-large-standard
+[summary] {
+  "type": "summary",
+  "checkpoint": "runs/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0013000.pt",
+  "step": 13000,
+  "decode": {
+    "decode_rule": "logistic_normal_resample_sde",
+    "steps": 128,
+    "model_t_mode": "const0.5",
+    "mean_mode": "anchor_semantic",
+    "endpoint_floor": 0.0,
+    "concentration_min": 1.0,
+    "concentration_max": 1024.0,
+    "endpoint_temp": 1.45,
+    "support_power": 1.0,
+    "semantic_power": 1.0,
+    "noise_init": "logistic_normal",
+    "noise_sigma": 3.0,
+    "noise_dirichlet_concentration": 1.0,
+    "sde_resample": "logistic_normal",
+    "logistic_normal_sigma_min": 0.18,
+    "logistic_normal_sigma_max": 3.0,
+    "logistic_normal_tau_min": 0.65,
+    "logistic_normal_tau_max": 1.0,
+    "final_from": "blend_0.5",
+    "n_samples": 256,
+    "seed": 20260522
+  },
+  "raw_genppl": {
+    "ppl": 42.64856193152362,
+    "nll_per_token": 3.7529935554682567,
+    "tokens": 23765,
+    "kept_samples": 256,
+    "total_samples": 256,
+    "empty_rate": 0.0,
+    "skipped_samples": 0
+  },
+  "stripped_genppl": {
+    "ppl": 55.69327778862184,
+    "nll_per_token": 4.019859453628943,
+    "tokens": 19611,
+    "kept_samples": 256,
+    "total_samples": 256,
+    "empty_rate": 0.0,
+    "skipped_samples": 0
+  },
+  "diversity": {
+    "sample_entropy": 2.4603500982206037,
+    "unique_tokens": 1359,
+    "token_count": 32768,
+    "distinct_1": 0.041473388671875,
+    "distinct_2": 0.19568159448818898,
+    "top_token_mass": 0.4425048828125
+  }
+}
+[done] docs/lta_samples/metrics_20260522/lm1b_len128_lognormal_atoms_every1k_logistic_normal_sde_t1p45_steps128_n256/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0013000/sde_steps128_samples256_scored.jsonl
+[watch-lognormal-sde] 2026-05-22_23:39:15 done step_0013000

LTA_openwebtext_dualt/logs/lm1b_len128_lognormal_atoms_sde_watch/infer_lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522_step_0031000_logistic_normal_t1p45.log

@@ -0,0 +1,76 @@
+[watch-lognormal-sde] 2026-05-23_01:27:30 infer runs/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0031000.pt -> docs/lta_samples/metrics_20260522/lm1b_len128_lognormal_atoms_every1k_logistic_normal_sde_t1p45_steps128_n256/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0031000
+[load] runs/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0031000.pt
+[ckpt] step=31000
+[sde] generated 16/256
+[sde] generated 32/256
+[sde] generated 48/256
+[sde] generated 64/256
+[sde] generated 80/256
+[sde] generated 96/256
+[sde] generated 112/256
+[sde] generated 128/256
+[sde] generated 144/256
+[sde] generated 160/256
+[sde] generated 176/256
+[sde] generated 192/256
+[sde] generated 208/256
+[sde] generated 224/256
+[sde] generated 240/256
+[sde] generated 256/256
+[score] loading scorer: /e2e-data/evad-tech-vla/wanghan58/models/flowtext_scorers/gpt2-large-standard
+[summary] {
+  "type": "summary",
+  "checkpoint": "runs/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0031000.pt",
+  "step": 31000,
+  "decode": {
+    "decode_rule": "logistic_normal_resample_sde",
+    "steps": 128,
+    "model_t_mode": "const0.5",
+    "mean_mode": "anchor_semantic",
+    "endpoint_floor": 0.0,
+    "concentration_min": 1.0,
+    "concentration_max": 1024.0,
+    "endpoint_temp": 1.45,
+    "support_power": 1.0,
+    "semantic_power": 1.0,
+    "noise_init": "logistic_normal",
+    "noise_sigma": 3.0,
+    "noise_dirichlet_concentration": 1.0,
+    "sde_resample": "logistic_normal",
+    "logistic_normal_sigma_min": 0.18,
+    "logistic_normal_sigma_max": 3.0,
+    "logistic_normal_tau_min": 0.65,
+    "logistic_normal_tau_max": 1.0,
+    "final_from": "blend_0.5",
+    "n_samples": 256,
+    "seed": 20260522
+  },
+  "raw_genppl": {
+    "ppl": 35.38284089986569,
+    "nll_per_token": 3.5662269822820702,
+    "tokens": 37231,
+    "kept_samples": 256,
+    "total_samples": 256,
+    "empty_rate": 0.0,
+    "skipped_samples": 0
+  },
+  "stripped_genppl": {
+    "ppl": 48.83814135726297,
+    "nll_per_token": 3.888511592770269,
+    "tokens": 31131,
+    "kept_samples": 256,
+    "total_samples": 256,
+    "empty_rate": 0.0,
+    "skipped_samples": 0
+  },
+  "diversity": {
+    "sample_entropy": 3.8118494156817455,
+    "unique_tokens": 2328,
+    "token_count": 32768,
+    "distinct_1": 0.071044921875,
+    "distinct_2": 0.35826771653543305,
+    "top_token_mass": 0.080902099609375
+  }
+}
+[done] docs/lta_samples/metrics_20260522/lm1b_len128_lognormal_atoms_every1k_logistic_normal_sde_t1p45_steps128_n256/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0031000/sde_steps128_samples256_scored.jsonl
+[watch-lognormal-sde] 2026-05-23_01:28:58 done step_0031000

LTA_openwebtext_dualt/logs/lm1b_len128_lognormal_atoms_sde_watch/infer_lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522_step_0053000_logistic_normal_t1p45.log

@@ -0,0 +1,76 @@
+[watch-lognormal-sde] 2026-05-23_03:29:49 infer runs/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0053000.pt -> docs/lta_samples/metrics_20260522/lm1b_len128_lognormal_atoms_every1k_logistic_normal_sde_t1p45_steps128_n256/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0053000
+[load] runs/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0053000.pt
+[ckpt] step=53000
+[sde] generated 16/256
+[sde] generated 32/256
+[sde] generated 48/256
+[sde] generated 64/256
+[sde] generated 80/256
+[sde] generated 96/256
+[sde] generated 112/256
+[sde] generated 128/256
+[sde] generated 144/256
+[sde] generated 160/256
+[sde] generated 176/256
+[sde] generated 192/256
+[sde] generated 208/256
+[sde] generated 224/256
+[sde] generated 240/256
+[sde] generated 256/256
+[score] loading scorer: /e2e-data/evad-tech-vla/wanghan58/models/flowtext_scorers/gpt2-large-standard
+[summary] {
+  "type": "summary",
+  "checkpoint": "runs/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0053000.pt",
+  "step": 53000,
+  "decode": {
+    "decode_rule": "logistic_normal_resample_sde",
+    "steps": 128,
+    "model_t_mode": "const0.5",
+    "mean_mode": "anchor_semantic",
+    "endpoint_floor": 0.0,
+    "concentration_min": 1.0,
+    "concentration_max": 1024.0,
+    "endpoint_temp": 1.45,
+    "support_power": 1.0,
+    "semantic_power": 1.0,
+    "noise_init": "logistic_normal",
+    "noise_sigma": 3.0,
+    "noise_dirichlet_concentration": 1.0,
+    "sde_resample": "logistic_normal",
+    "logistic_normal_sigma_min": 0.18,
+    "logistic_normal_sigma_max": 3.0,
+    "logistic_normal_tau_min": 0.65,
+    "logistic_normal_tau_max": 1.0,
+    "final_from": "blend_0.5",
+    "n_samples": 256,
+    "seed": 20260522
+  },
+  "raw_genppl": {
+    "ppl": 31.72183032653609,
+    "nll_per_token": 3.4570050977039366,
+    "tokens": 35633,
+    "kept_samples": 256,
+    "total_samples": 256,
+    "empty_rate": 0.0,
+    "skipped_samples": 0
+  },
+  "stripped_genppl": {
+    "ppl": 43.3006010160156,
+    "nll_per_token": 3.7681665151896104,
+    "tokens": 29643,
+    "kept_samples": 256,
+    "total_samples": 256,
+    "empty_rate": 0.0,
+    "skipped_samples": 0
+  },
+  "diversity": {
+    "sample_entropy": 3.664896271654086,
+    "unique_tokens": 1997,
+    "token_count": 32768,
+    "distinct_1": 0.060943603515625,
+    "distinct_2": 0.3108390748031496,
+    "top_token_mass": 0.1292724609375
+  }
+}
+[done] docs/lta_samples/metrics_20260522/lm1b_len128_lognormal_atoms_every1k_logistic_normal_sde_t1p45_steps128_n256/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0053000/sde_steps128_samples256_scored.jsonl
+[watch-lognormal-sde] 2026-05-23_03:31:17 done step_0053000

LTA_openwebtext_dualt/logs/lm1b_len128_lognormal_atoms_sde_watch/infer_lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522_step_0088000_logistic_normal_t1p45.log

@@ -0,0 +1,76 @@
+[watch-lognormal-sde] 2026-05-23_06:45:27 infer runs/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0088000.pt -> docs/lta_samples/metrics_20260522/lm1b_len128_lognormal_atoms_every1k_logistic_normal_sde_t1p45_steps128_n256/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0088000
+[load] runs/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0088000.pt
+[ckpt] step=88000
+[sde] generated 16/256
+[sde] generated 32/256
+[sde] generated 48/256
+[sde] generated 64/256
+[sde] generated 80/256
+[sde] generated 96/256
+[sde] generated 112/256
+[sde] generated 128/256
+[sde] generated 144/256
+[sde] generated 160/256
+[sde] generated 176/256
+[sde] generated 192/256
+[sde] generated 208/256
+[sde] generated 224/256
+[sde] generated 240/256
+[sde] generated 256/256
+[score] loading scorer: /e2e-data/evad-tech-vla/wanghan58/models/flowtext_scorers/gpt2-large-standard
+[summary] {
+  "type": "summary",
+  "checkpoint": "runs/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0088000.pt",
+  "step": 88000,
+  "decode": {
+    "decode_rule": "logistic_normal_resample_sde",
+    "steps": 128,
+    "model_t_mode": "const0.5",
+    "mean_mode": "anchor_semantic",
+    "endpoint_floor": 0.0,
+    "concentration_min": 1.0,
+    "concentration_max": 1024.0,
+    "endpoint_temp": 1.45,
+    "support_power": 1.0,
+    "semantic_power": 1.0,
+    "noise_init": "logistic_normal",
+    "noise_sigma": 3.0,
+    "noise_dirichlet_concentration": 1.0,
+    "sde_resample": "logistic_normal",
+    "logistic_normal_sigma_min": 0.18,
+    "logistic_normal_sigma_max": 3.0,
+    "logistic_normal_tau_min": 0.65,
+    "logistic_normal_tau_max": 1.0,
+    "final_from": "blend_0.5",
+    "n_samples": 256,
+    "seed": 20260522
+  },
+  "raw_genppl": {
+    "ppl": 33.27389293811112,
+    "nll_per_token": 3.5047730936484087,
+    "tokens": 35836,
+    "kept_samples": 256,
+    "total_samples": 256,
+    "empty_rate": 0.0,
+    "skipped_samples": 0
+  },
+  "stripped_genppl": {
+    "ppl": 45.58816518309832,
+    "nll_per_token": 3.8196481473768147,
+    "tokens": 29823,
+    "kept_samples": 256,
+    "total_samples": 256,
+    "empty_rate": 0.0,
+    "skipped_samples": 0
+  },
+  "diversity": {
+    "sample_entropy": 3.7643278889654623,
+    "unique_tokens": 2115,
+    "token_count": 32768,
+    "distinct_1": 0.064544677734375,
+    "distinct_2": 0.3441806102362205,
+    "top_token_mass": 0.098236083984375
+  }
+}
+[done] docs/lta_samples/metrics_20260522/lm1b_len128_lognormal_atoms_every1k_logistic_normal_sde_t1p45_steps128_n256/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0088000/sde_steps128_samples256_scored.jsonl
+[watch-lognormal-sde] 2026-05-23_06:46:55 done step_0088000

LTA_openwebtext_dualt/logs/lm1b_len128_lognormal_atoms_sde_watch/infer_lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522_step_0089000_logistic_normal_t1p45.log

@@ -0,0 +1,76 @@
+[watch-lognormal-sde] 2026-05-23_06:50:41 infer runs/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0089000.pt -> docs/lta_samples/metrics_20260522/lm1b_len128_lognormal_atoms_every1k_logistic_normal_sde_t1p45_steps128_n256/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0089000
+[load] runs/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0089000.pt
+[ckpt] step=89000
+[sde] generated 16/256
+[sde] generated 32/256
+[sde] generated 48/256
+[sde] generated 64/256
+[sde] generated 80/256
+[sde] generated 96/256
+[sde] generated 112/256
+[sde] generated 128/256
+[sde] generated 144/256
+[sde] generated 160/256
+[sde] generated 176/256
+[sde] generated 192/256
+[sde] generated 208/256
+[sde] generated 224/256
+[sde] generated 240/256
+[sde] generated 256/256
+[score] loading scorer: /e2e-data/evad-tech-vla/wanghan58/models/flowtext_scorers/gpt2-large-standard
+[summary] {
+  "type": "summary",
+  "checkpoint": "runs/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0089000.pt",
+  "step": 89000,
+  "decode": {
+    "decode_rule": "logistic_normal_resample_sde",
+    "steps": 128,
+    "model_t_mode": "const0.5",
+    "mean_mode": "anchor_semantic",
+    "endpoint_floor": 0.0,
+    "concentration_min": 1.0,
+    "concentration_max": 1024.0,
+    "endpoint_temp": 1.45,
+    "support_power": 1.0,
+    "semantic_power": 1.0,
+    "noise_init": "logistic_normal",
+    "noise_sigma": 3.0,
+    "noise_dirichlet_concentration": 1.0,
+    "sde_resample": "logistic_normal",
+    "logistic_normal_sigma_min": 0.18,
+    "logistic_normal_sigma_max": 3.0,
+    "logistic_normal_tau_min": 0.65,
+    "logistic_normal_tau_max": 1.0,
+    "final_from": "blend_0.5",
+    "n_samples": 256,
+    "seed": 20260522
+  },
+  "raw_genppl": {
+    "ppl": 37.451702182944835,
+    "nll_per_token": 3.6230521610783706,
+    "tokens": 25633,
+    "kept_samples": 256,
+    "total_samples": 256,
+    "empty_rate": 0.0,
+    "skipped_samples": 0
+  },
+  "stripped_genppl": {
+    "ppl": 44.449205876761276,
+    "nll_per_token": 3.794347096260644,
+    "tokens": 22137,
+    "kept_samples": 256,
+    "total_samples": 256,
+    "empty_rate": 0.0,
+    "skipped_samples": 0
+  },
+  "diversity": {
+    "sample_entropy": 2.7356215744152865,
+    "unique_tokens": 1432,
+    "token_count": 32768,
+    "distinct_1": 0.043701171875,
+    "distinct_2": 0.23071481299212598,
+    "top_token_mass": 0.36322021484375
+  }
+}
+[done] docs/lta_samples/metrics_20260522/lm1b_len128_lognormal_atoms_every1k_logistic_normal_sde_t1p45_steps128_n256/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0089000/sde_steps128_samples256_scored.jsonl
+[watch-lognormal-sde] 2026-05-23_06:52:08 done step_0089000

LTA_openwebtext_dualt/logs/lm1b_len128_lognormal_atoms_sde_watch/infer_lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522_step_0114000_logistic_normal_t1p45.log

@@ -0,0 +1,76 @@
+[watch-lognormal-sde] 2026-05-23_09:10:28 infer runs/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0114000.pt -> docs/lta_samples/metrics_20260522/lm1b_len128_lognormal_atoms_every1k_logistic_normal_sde_t1p45_steps128_n256/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0114000
+[load] runs/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0114000.pt
+[ckpt] step=114000
+[sde] generated 16/256
+[sde] generated 32/256
+[sde] generated 48/256
+[sde] generated 64/256
+[sde] generated 80/256
+[sde] generated 96/256
+[sde] generated 112/256
+[sde] generated 128/256
+[sde] generated 144/256
+[sde] generated 160/256
+[sde] generated 176/256
+[sde] generated 192/256
+[sde] generated 208/256
+[sde] generated 224/256
+[sde] generated 240/256
+[sde] generated 256/256
+[score] loading scorer: /e2e-data/evad-tech-vla/wanghan58/models/flowtext_scorers/gpt2-large-standard
+[summary] {
+  "type": "summary",
+  "checkpoint": "runs/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0114000.pt",
+  "step": 114000,
+  "decode": {
+    "decode_rule": "logistic_normal_resample_sde",
+    "steps": 128,
+    "model_t_mode": "const0.5",
+    "mean_mode": "anchor_semantic",
+    "endpoint_floor": 0.0,
+    "concentration_min": 1.0,
+    "concentration_max": 1024.0,
+    "endpoint_temp": 1.45,
+    "support_power": 1.0,
+    "semantic_power": 1.0,
+    "noise_init": "logistic_normal",
+    "noise_sigma": 3.0,
+    "noise_dirichlet_concentration": 1.0,
+    "sde_resample": "logistic_normal",
+    "logistic_normal_sigma_min": 0.18,
+    "logistic_normal_sigma_max": 3.0,
+    "logistic_normal_tau_min": 0.65,
+    "logistic_normal_tau_max": 1.0,
+    "final_from": "blend_0.5",
+    "n_samples": 256,
+    "seed": 20260522
+  },
+  "raw_genppl": {
+    "ppl": 33.59211599032921,
+    "nll_per_token": 3.514291396291185,
+    "tokens": 35890,
+    "kept_samples": 256,
+    "total_samples": 256,
+    "empty_rate": 0.0,
+    "skipped_samples": 0
+  },
+  "stripped_genppl": {
+    "ppl": 42.07712923710024,
+    "nll_per_token": 3.739504344550458,
+    "tokens": 30627,
+    "kept_samples": 256,
+    "total_samples": 256,
+    "empty_rate": 0.0,
+    "skipped_samples": 0
+  },
+  "diversity": {
+    "sample_entropy": 3.6871935630110833,
+    "unique_tokens": 2141,
+    "token_count": 32768,
+    "distinct_1": 0.065338134765625,
+    "distinct_2": 0.3384596456692913,
+    "top_token_mass": 0.078521728515625
+  }
+}
+[done] docs/lta_samples/metrics_20260522/lm1b_len128_lognormal_atoms_every1k_logistic_normal_sde_t1p45_steps128_n256/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0114000/sde_steps128_samples256_scored.jsonl
+[watch-lognormal-sde] 2026-05-23_09:11:55 done step_0114000

LTA_openwebtext_dualt/logs/lm1b_len128_lognormal_atoms_sde_watch/infer_lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522_step_0122000_logistic_normal_t1p45.log

@@ -0,0 +1,76 @@
+[watch-lognormal-sde] 2026-05-23_09:54:32 infer runs/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0122000.pt -> docs/lta_samples/metrics_20260522/lm1b_len128_lognormal_atoms_every1k_logistic_normal_sde_t1p45_steps128_n256/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0122000
+[load] runs/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0122000.pt
+[ckpt] step=122000
+[sde] generated 16/256
+[sde] generated 32/256
+[sde] generated 48/256
+[sde] generated 64/256
+[sde] generated 80/256
+[sde] generated 96/256
+[sde] generated 112/256
+[sde] generated 128/256
+[sde] generated 144/256
+[sde] generated 160/256
+[sde] generated 176/256
+[sde] generated 192/256
+[sde] generated 208/256
+[sde] generated 224/256
+[sde] generated 240/256
+[sde] generated 256/256
+[score] loading scorer: /e2e-data/evad-tech-vla/wanghan58/models/flowtext_scorers/gpt2-large-standard
+[summary] {
+  "type": "summary",
+  "checkpoint": "runs/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0122000.pt",
+  "step": 122000,
+  "decode": {
+    "decode_rule": "logistic_normal_resample_sde",
+    "steps": 128,
+    "model_t_mode": "const0.5",
+    "mean_mode": "anchor_semantic",
+    "endpoint_floor": 0.0,
+    "concentration_min": 1.0,
+    "concentration_max": 1024.0,
+    "endpoint_temp": 1.45,
+    "support_power": 1.0,
+    "semantic_power": 1.0,
+    "noise_init": "logistic_normal",
+    "noise_sigma": 3.0,
+    "noise_dirichlet_concentration": 1.0,
+    "sde_resample": "logistic_normal",
+    "logistic_normal_sigma_min": 0.18,
+    "logistic_normal_sigma_max": 3.0,
+    "logistic_normal_tau_min": 0.65,
+    "logistic_normal_tau_max": 1.0,
+    "final_from": "blend_0.5",
+    "n_samples": 256,
+    "seed": 20260522
+  },
+  "raw_genppl": {
+    "ppl": 35.1050136466682,
+    "nll_per_token": 3.5583439592077264,
+    "tokens": 33833,
+    "kept_samples": 256,
+    "total_samples": 256,
+    "empty_rate": 0.0,
+    "skipped_samples": 0
+  },
+  "stripped_genppl": {
+    "ppl": 44.71879769307101,
+    "nll_per_token": 3.8003939432678764,
+    "tokens": 28753,
+    "kept_samples": 256,
+    "total_samples": 256,
+    "empty_rate": 0.0,
+    "skipped_samples": 0
+  },
+  "diversity": {
+    "sample_entropy": 3.532846984384408,
+    "unique_tokens": 2248,
+    "token_count": 32768,
+    "distinct_1": 0.068603515625,
+    "distinct_2": 0.34171998031496065,
+    "top_token_mass": 0.150146484375
+  }
+}
+[done] docs/lta_samples/metrics_20260522/lm1b_len128_lognormal_atoms_every1k_logistic_normal_sde_t1p45_steps128_n256/lta_lm1b_classic_c1024_len128_lognormalatoms_gbs512_4gpu_20260522/step_0122000/sde_steps128_samples256_scored.jsonl
+[watch-lognormal-sde] 2026-05-23_09:56:00 done step_0122000

LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/transformers/models/vit_mae/__init__.py

@@ -0,0 +1,27 @@
+# Copyright 2024 The HuggingFace Team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+from typing import TYPE_CHECKING
+
+from ...utils import _LazyModule
+from ...utils.import_utils import define_import_structure
+
+
+if TYPE_CHECKING:
+    from .configuration_vit_mae import *
+    from .modeling_vit_mae import *
+else:
+    import sys
+
+    _file = globals()["__file__"]
+    sys.modules[__name__] = _LazyModule(__name__, _file, define_import_structure(_file), module_spec=__spec__)

LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/transformers/models/vit_mae/configuration_vit_mae.py

@@ -0,0 +1,72 @@
+# Copyright 2022 Facebook AI and The HuggingFace Inc. team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""ViT MAE model configuration"""
+
+from huggingface_hub.dataclasses import strict
+
+from ...configuration_utils import PreTrainedConfig
+from ...utils import auto_docstring
+
+
+@auto_docstring(checkpoint="facebook/vit-mae-base")
+@strict
+class ViTMAEConfig(PreTrainedConfig):
+    r"""
+    decoder_num_hidden_layers (`int`, *optional*, defaults to 8):
+        Number of hidden layers in the decoder.
+    mask_ratio (`float`, *optional*, defaults to 0.75):
+        The ratio of the number of masked tokens in the input sequence.
+    norm_pix_loss (`bool`, *optional*, defaults to `False`):
+        Whether or not to train with normalized pixels (see Table 3 in the paper). Using normalized pixels improved
+        representation quality in the experiments of the authors.
+
+    Example:
+
+    ```python
+    >>> from transformers import ViTMAEConfig, ViTMAEModel
+
+    >>> # Initializing a ViT MAE vit-mae-base style configuration
+    >>> configuration = ViTMAEConfig()
+
+    >>> # Initializing a model (with random weights) from the vit-mae-base style configuration
+    >>> model = ViTMAEModel(configuration)
+
+    >>> # Accessing the model configuration
+    >>> configuration = model.config
+    ```"""
+
+    model_type = "vit_mae"
+
+    hidden_size: int = 768
+    num_hidden_layers: int = 12
+    num_attention_heads: int = 12
+    intermediate_size: int = 3072
+    hidden_act: str = "gelu"
+    hidden_dropout_prob: float | int = 0.0
+    attention_probs_dropout_prob: float | int = 0.0
+    initializer_range: float = 0.02
+    layer_norm_eps: float = 1e-12
+    image_size: int | list[int] | tuple[int, int] = 224
+    patch_size: int | list[int] | tuple[int, int] = 16
+    num_channels: int = 3
+    qkv_bias: bool = True
+    decoder_num_attention_heads: int = 16
+    decoder_hidden_size: int = 512
+    decoder_num_hidden_layers: int = 8
+    decoder_intermediate_size: int = 2048
+    mask_ratio: float = 0.75
+    norm_pix_loss: bool = False
+
+
+__all__ = ["ViTMAEConfig"]

LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/transformers/models/vit_mae/modular_vit_mae.py

@@ -0,0 +1,682 @@
+# Copyright 2022 Facebook AI and The HuggingFace Inc. team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""PyTorch ViT MAE (masked autoencoder) model - modular implementation."""
+
+from copy import deepcopy
+from dataclasses import dataclass
+
+import torch
+from torch import nn
+
+from ... import initialization as init
+from ...masking_utils import create_bidirectional_mask
+from ...processing_utils import Unpack
+from ...utils import ModelOutput, TransformersKwargs, auto_docstring, logging
+from ...utils.generic import can_return_tuple, merge_with_config_defaults
+from ...utils.output_capturing import capture_outputs
+from ..vit.modeling_vit import (
+    PreTrainedModel,
+    ViTAttention,
+    ViTEmbeddings,
+    ViTLayer,
+    ViTMLP,
+    ViTModel,
+    ViTPatchEmbeddings,
+    ViTPreTrainedModel,
+)
+from .configuration_vit_mae import ViTMAEConfig
+
+
+logger = logging.get_logger(__name__)
+
+
+def build_2d_sinusoidal_position_embedding(
+    height: int,
+    width: int,
+    embed_dim: int = 256,
+    temperature: float = 10000.0,
+    cls_token: bool = False,
+    device: torch.device | None = None,
+    dtype: torch.dtype = torch.float32,
+) -> torch.Tensor:
+    """2D sinusoidal position embeddings for an image patch grid.
+
+    Each (h, w) position gets an ``embed_dim``-dimensional vector laid out as
+    ``[sin_h | cos_h | sin_w | cos_w]``, with row-major (H-outer) patch ordering.
+
+    Args:
+        height: Grid height in patches.
+        width: Grid width in patches.
+        embed_dim: Total embedding dimension; must be divisible by 4.
+        temperature: Base for the frequency decay.
+        cls_token: If `True`, prepend a zero row for a CLS token.
+        device: Target device; defaults to CPU.
+        dtype: Output dtype; frequency arithmetic uses float64 internally.
+
+    Returns:
+        Tensor of shape ``(height * width [+1], embed_dim)``.
+    """
+    if embed_dim % 4 != 0:
+        raise ValueError(f"`embed_dim` must be divisible by 4, got {embed_dim}")
+
+    pos_dim = embed_dim // 4
+    omega = torch.arange(pos_dim, dtype=torch.float64, device=device) / pos_dim
+    omega = 1.0 / temperature**omega  # (D/4,)
+
+    grid_h = torch.arange(height, dtype=torch.float64, device=device)
+    grid_w = torch.arange(width, dtype=torch.float64, device=device)
+    grid_h, grid_w = torch.meshgrid(grid_h, grid_w, indexing="ij")  # (H, W) each
+
+    emb_h = grid_h.flatten().outer(omega)  # (H*W, D/4)
+    emb_w = grid_w.flatten().outer(omega)  # (H*W, D/4)
+
+    pos_embed = torch.cat([emb_h.sin(), emb_h.cos(), emb_w.sin(), emb_w.cos()], dim=1)
+
+    if cls_token:
+        pos_embed = torch.cat([torch.zeros(1, embed_dim, dtype=torch.float64, device=device), pos_embed], dim=0)
+
+    return pos_embed.to(dtype)
+
+
+@auto_docstring(
+    custom_intro="""
+    Class for ViTMAEModel's outputs, with potential hidden states and attentions.
+    """
+)
+@dataclass
+class ViTMAEModelOutput(ModelOutput):
+    r"""
+    mask (`torch.FloatTensor` of shape `(batch_size, sequence_length)`):
+        Tensor indicating which patches are masked (1) and which are not (0).
+    ids_restore (`torch.LongTensor` of shape `(batch_size, sequence_length)`):
+        Tensor containing the original index of the (shuffled) masked patches.
+    """
+
+    last_hidden_state: torch.FloatTensor | None = None
+    mask: torch.LongTensor | None = None
+    ids_restore: torch.LongTensor | None = None
+    hidden_states: tuple[torch.FloatTensor] | None = None
+    attentions: tuple[torch.FloatTensor] | None = None
+
+
+@auto_docstring(
+    custom_intro="""
+    Class for ViTMAEDecoder's outputs, with potential hidden states and attentions.
+    """
+)
+@dataclass
+class ViTMAEDecoderOutput(ModelOutput):
+    r"""
+    logits (`torch.FloatTensor` of shape `(batch_size, sequence_length, patch_size ** 2 * num_channels)`):
+        Pixel reconstruction logits.
+    """
+
+    logits: torch.FloatTensor | None = None
+    hidden_states: tuple[torch.FloatTensor] | None = None
+    attentions: tuple[torch.FloatTensor] | None = None
+
+
+@auto_docstring(
+    custom_intro="""
+    Class for ViTMAEForPreTraining's outputs, with potential hidden states and attentions.
+    """
+)
+@dataclass
+class ViTMAEForPreTrainingOutput(ModelOutput):
+    r"""
+    loss (`torch.FloatTensor` of shape `(1,)`):
+        Pixel reconstruction loss.
+    logits (`torch.FloatTensor` of shape `(batch_size, sequence_length, patch_size ** 2 * num_channels)`):
+        Pixel reconstruction logits.
+    mask (`torch.FloatTensor` of shape `(batch_size, sequence_length)`):
+        Tensor indicating which patches are masked (1) and which are not (0).
+    ids_restore (`torch.LongTensor` of shape `(batch_size, sequence_length)`):
+        Tensor containing the original index of the (shuffled) masked patches.
+    """
+
+    loss: torch.FloatTensor | None = None
+    logits: torch.FloatTensor | None = None
+    mask: torch.LongTensor | None = None
+    ids_restore: torch.LongTensor | None = None
+    hidden_states: tuple[torch.FloatTensor] | None = None
+    attentions: tuple[torch.FloatTensor] | None = None
+
+
+class ViTMAEPatchEmbeddings(ViTPatchEmbeddings):
+    pass
+
+
+class ViTMAEEmbeddings(ViTEmbeddings):
+    """
+    Construct the CLS token, position and patch embeddings for MAE.
+    """
+
+    def __init__(self, config: ViTMAEConfig):
+        super().__init__()
+        self.cls_token = nn.Parameter(torch.zeros(1, 1, config.hidden_size))
+        num_patches = self.patch_embeddings.num_patches
+        # fixed sin-cos embedding
+        self.position_embeddings = nn.Parameter(
+            torch.zeros(1, num_patches + 1, config.hidden_size), requires_grad=False
+        )
+        self.config = config
+        del self.mask_token
+        del self.dropout
+
+    def initialize_weights(self):
+        if getattr(self.patch_embeddings.projection, "_is_hf_initialized", False):
+            return
+        # initialize (and freeze) position embeddings by sin-cos embedding
+        grid_size = int(self.patch_embeddings.num_patches**0.5)
+        pos_embed = build_2d_sinusoidal_position_embedding(
+            height=grid_size,
+            width=grid_size,
+            embed_dim=self.position_embeddings.shape[-1],
+            cls_token=True,
+        )
+        # The original ViT-MAE implementation had a variable naming bug that
+        # swapped h and w, producing [sin_w|cos_w|sin_h|cos_h] instead of the
+        # canonical [sin_h|cos_h|sin_w|cos_w]. Pretrained weights rely on this
+        # layout, so we rotate the two halves to match.
+        half = pos_embed.shape[-1] // 2
+        pos_embed = torch.cat([pos_embed[..., half:], pos_embed[..., :half]], dim=-1)
+        init.copy_(self.position_embeddings, pos_embed.unsqueeze(0))
+
+        # initialize patch_embeddings like nn.Linear (instead of nn.Conv2d)
+        w = self.patch_embeddings.projection.weight
+        init.xavier_uniform_(w.view([w.shape[0], -1]))
+
+        # timm's trunc_normal_(std=.02) is effectively normal_(std=0.02) as cutoff is too big (2.)
+        init.normal_(self.cls_token, std=self.config.initializer_range)
+
+    def random_masking(self, sequence, noise=None):
+        """
+        Perform per-sample random masking by per-sample shuffling. Per-sample shuffling is done by argsort random
+        noise.
+
+        Args:
+            sequence (`torch.LongTensor` of shape `(batch_size, sequence_length, dim)`)
+            noise (`torch.FloatTensor` of shape `(batch_size, sequence_length)`, *optional*) which is
+                mainly used for testing purposes to control randomness and maintain the reproducibility
+        """
+        batch_size, seq_length, dim = sequence.shape
+        len_keep = int(seq_length * (1 - self.config.mask_ratio))
+
+        if noise is None:
+            noise = torch.rand(batch_size, seq_length, device=sequence.device)
+
+        # sort noise for each sample
+        ids_shuffle = torch.argsort(noise, dim=1).to(sequence.device)
+        ids_restore = torch.argsort(ids_shuffle, dim=1).to(sequence.device)
+
+        # keep the first subset
+        ids_keep = ids_shuffle[:, :len_keep]
+        sequence_unmasked = torch.gather(sequence, dim=1, index=ids_keep.unsqueeze(-1).repeat(1, 1, dim))
+
+        # generate the binary mask: 0 is keep, 1 is remove
+        mask = torch.ones([batch_size, seq_length], device=sequence.device)
+        mask[:, :len_keep] = 0
+        # unshuffle to get the binary mask
+        mask = torch.gather(mask, dim=1, index=ids_restore)
+
+        return sequence_unmasked, mask, ids_restore
+
+    def forward(
+        self,
+        pixel_values: torch.Tensor,
+        noise: torch.Tensor | None = None,
+        interpolate_pos_encoding: bool = False,
+    ):
+        height, width = pixel_values.shape[2:]
+        embeddings = self.patch_embeddings(pixel_values)
+        if interpolate_pos_encoding:
+            position_embeddings = self.interpolate_pos_encoding(embeddings, height, width)
+        else:
+            if height != self.image_size[0] or width != self.image_size[1]:
+                raise ValueError(
+                    f"Input image size ({height}*{width}) doesn't match model"
+                    f" ({self.image_size[0]}*{self.image_size[1]})."
+                )
+            position_embeddings = self.position_embeddings
+
+        # add position embeddings w/o cls token
+        embeddings = embeddings + position_embeddings[:, 1:, :]
+
+        # masking: length -> length * config.mask_ratio
+        embeddings, mask, ids_restore = self.random_masking(embeddings, noise)
+
+        # prepend cls token
+        cls_token = self.cls_token + position_embeddings[:, :1, :]
+        cls_tokens = cls_token.expand(embeddings.shape[0], -1, -1)
+        embeddings = torch.cat((cls_tokens, embeddings), dim=1)
+
+        return embeddings, mask, ids_restore
+
+
+# Pass-through: ViT MAE encoder uses same architecture as ViT
+class ViTMAEAttention(ViTAttention):
+    pass
+
+
+class ViTMAEMLP(ViTMLP):
+    pass
+
+
+class ViTMAELayer(ViTLayer):
+    pass
+
+
+class ViTMAEDecoder(nn.Module):
+    def __init__(self, config: ViTMAEConfig, num_patches: int):
+        super().__init__()
+        self.decoder_embed = nn.Linear(config.hidden_size, config.decoder_hidden_size, bias=True)
+        self.mask_token = nn.Parameter(torch.zeros(1, 1, config.decoder_hidden_size))
+        self.decoder_pos_embed = nn.Parameter(
+            torch.zeros(1, num_patches + 1, config.decoder_hidden_size),
+            requires_grad=False,
+        )
+
+        decoder_config = deepcopy(config)
+        decoder_config.hidden_size = config.decoder_hidden_size
+        decoder_config.num_hidden_layers = config.decoder_num_hidden_layers
+        decoder_config.num_attention_heads = config.decoder_num_attention_heads
+        decoder_config.intermediate_size = config.decoder_intermediate_size
+        self.decoder_layers = nn.ModuleList(
+            [ViTMAELayer(decoder_config) for _ in range(config.decoder_num_hidden_layers)]
+        )
+
+        self.decoder_norm = nn.LayerNorm(config.decoder_hidden_size, eps=config.layer_norm_eps)
+        self.decoder_pred = nn.Linear(
+            config.decoder_hidden_size,
+            config.patch_size**2 * config.num_channels,
+            bias=True,
+        )
+        self.gradient_checkpointing = False
+        self.config = decoder_config
+        self.initialize_weights(num_patches)
+
+    def interpolate_pos_encoding(self, embeddings: torch.Tensor) -> torch.Tensor:
+        """
+        This method is a modified version of the interpolation function for ViT-mae model at the decoder, that
+        allows to interpolate the pre-trained decoder position encodings, to be able to use the model on higher
+        resolution images.
+
+        Adapted from:
+        https://github.com/facebookresearch/dino/blob/de9ee3df6cf39fac952ab558447af1fa1365362a/vision_transformer.py#L174
+        """
+
+        # -1 removes the class dimension since we later append it without interpolation
+        embeddings_positions = embeddings.shape[1] - 1
+
+        # Separation of class token and patch tokens
+        class_pos_embed = self.decoder_pos_embed[:, :1]
+        patch_pos_embed = self.decoder_pos_embed[:, 1:]
+
+        dim = self.decoder_pos_embed.shape[-1]
+
+        # Increasing a dimension to enable bicubic interpolation
+        patch_pos_embed = patch_pos_embed.reshape(1, 1, -1, dim)
+
+        # permute to bring the dimension to be interpolated, to the last
+        patch_pos_embed = patch_pos_embed.permute(0, 3, 1, 2)
+
+        # Interpolating the decoder position embeddings shape wrt embeddings shape i.e (x).
+        patch_pos_embed = nn.functional.interpolate(
+            patch_pos_embed,
+            size=(patch_pos_embed.shape[-2], embeddings_positions),
+            mode="bicubic",
+            align_corners=False,
+        )
+
+        # Converting back to the original shape
+        patch_pos_embed = patch_pos_embed.permute(0, 2, 3, 1).view(1, -1, dim)
+        return torch.cat((class_pos_embed, patch_pos_embed), dim=1)
+
+    def initialize_weights(self, num_patches: int):
+        grid_size = int(num_patches**0.5)
+        decoder_pos_embed = build_2d_sinusoidal_position_embedding(
+            height=grid_size,
+            width=grid_size,
+            embed_dim=self.decoder_pos_embed.shape[-1],
+            cls_token=True,
+        )
+        # See comment in initialize_weights above: rotate h/w blocks to match pretrained layout.
+        half = decoder_pos_embed.shape[-1] // 2
+        decoder_pos_embed = torch.cat([decoder_pos_embed[..., half:], decoder_pos_embed[..., :half]], dim=-1)
+        init.copy_(self.decoder_pos_embed, decoder_pos_embed.unsqueeze(0))
+
+        init.normal_(self.mask_token, std=self.config.initializer_range)
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        ids_restore: torch.Tensor,
+        interpolate_pos_encoding: bool = False,
+    ) -> ViTMAEDecoderOutput:
+        # Embed tokens
+        hidden_states = self.decoder_embed(hidden_states)
+
+        # Append mask tokens to sequence
+        mask_tokens = self.mask_token.repeat(
+            hidden_states.shape[0], ids_restore.shape[1] + 1 - hidden_states.shape[1], 1
+        )
+        unmasked_tokens = torch.cat([hidden_states[:, 1:, :], mask_tokens], dim=1)
+
+        # Unshuffle
+        unmasked_tokens = torch.gather(
+            unmasked_tokens,
+            dim=1,
+            index=ids_restore.unsqueeze(-1).repeat(1, 1, hidden_states.shape[2]).to(unmasked_tokens.device),
+        )
+        hidden_states = torch.cat([hidden_states[:, :1, :], unmasked_tokens], dim=1)
+
+        # Add pos embed
+        if interpolate_pos_encoding:
+            decoder_pos_embed = self.interpolate_pos_encoding(hidden_states)
+        else:
+            decoder_pos_embed = self.decoder_pos_embed
+        hidden_states = hidden_states + decoder_pos_embed
+
+        # Apply Transformer layers (blocks)
+        for layer_module in self.decoder_layers:
+            hidden_states = layer_module(hidden_states)
+
+        hidden_states = self.decoder_norm(hidden_states)
+
+        # Predictor projection
+        logits = self.decoder_pred(hidden_states)
+
+        # Remove cls token
+        logits = logits[:, 1:, :]
+
+        return ViTMAEDecoderOutput(logits=logits)
+
+
+@auto_docstring
+class ViTMAEPreTrainedModel(ViTPreTrainedModel):
+    base_model_prefix = "vit"
+    _no_split_modules = ["ViTMAEEmbeddings", "ViTMAELayer", "ViTMAEDecoder"]
+
+    @torch.no_grad()
+    def _init_weights(self, module):
+        """Initialize the weights"""
+        PreTrainedModel._init_weights(self, module)
+        if isinstance(module, ViTMAEEmbeddings):
+            module.initialize_weights()
+        elif isinstance(module, ViTMAEDecoder):
+            init.zeros_(module.mask_token)
+            init.zeros_(module.decoder_pos_embed)
+
+
+@auto_docstring
+class ViTMAEModel(ViTModel):
+    def __init__(self, config: ViTMAEConfig):
+        r"""
+        config (`ViTMAEConfig`):
+            Configuration for the model.
+        """
+        super().__init__(config)
+        self.embeddings = ViTMAEEmbeddings(config)
+        del self.pooler
+
+    @merge_with_config_defaults
+    @capture_outputs(tie_last_hidden_states=False)
+    @auto_docstring
+    def forward(
+        self,
+        pixel_values: torch.Tensor | None = None,
+        noise: torch.Tensor | None = None,
+        interpolate_pos_encoding: bool | None = None,
+        attention_mask: torch.Tensor | None = None,
+        **kwargs: Unpack[TransformersKwargs],
+    ) -> ViTMAEModelOutput:
+        r"""
+        noise (`torch.FloatTensor` of shape `(batch_size, sequence_length)`, *optional*):
+            Mainly used for testing purposes to control randomness and maintain the reproducibility
+        interpolate_pos_encoding (`bool`, *optional*, default `False`):
+            Whether to interpolate the pre-trained position encodings. This is mainly used to use the model on higher
+            resolution images.
+
+        Examples:
+
+        ```python
+        >>> from transformers import AutoImageProcessor, ViTMAEModel
+        >>> from PIL import Image
+        >>> import httpx
+        >>> from io import BytesIO
+
+        >>> url = "http://images.cocodataset.org/val2017/000000039769.jpg"
+        >>> with httpx.stream("GET", url) as response:
+        ...     image = Image.open(BytesIO(response.read()))
+
+        >>> image_processor = AutoImageProcessor.from_pretrained("facebook/vit-mae-base")
+        >>> model = ViTMAEModel.from_pretrained("facebook/vit-mae-base")
+
+        >>> inputs = image_processor(images=image, return_tensors="pt")
+        >>> outputs = model(**inputs)
+        >>> last_hidden_states = outputs.last_hidden_state
+        ```"""
+
+        embedding_output, mask, ids_restore = self.embeddings(
+            pixel_values,
+            noise=noise,
+            interpolate_pos_encoding=interpolate_pos_encoding,
+        )
+
+        attention_mask = create_bidirectional_mask(
+            config=self.config,
+            inputs_embeds=embedding_output,
+            attention_mask=attention_mask,
+        )
+
+        hidden_states = embedding_output
+        for layer in self.layers:
+            hidden_states = layer(hidden_states, attention_mask, **kwargs)
+        sequence_output = self.layernorm(hidden_states)
+
+        return ViTMAEModelOutput(last_hidden_state=sequence_output, mask=mask, ids_restore=ids_restore)
+
+
+@auto_docstring(
+    custom_intro="""
+    The ViTMAE Model transformer with the decoder on top for self-supervised pre-training.
+
+    <Tip>
+
+    Note that we provide a script to pre-train this model on custom data in our [examples
+    directory](https://github.com/huggingface/transformers/tree/main/examples/pytorch/image-pretraining).
+
+    </Tip>
+    """
+)
+class ViTMAEForPreTraining(ViTMAEPreTrainedModel):
+    def __init__(self, config: ViTMAEConfig):
+        super().__init__(config)
+        self.config = config
+
+        self.vit = ViTMAEModel(config)
+        self.decoder = ViTMAEDecoder(config, num_patches=self.vit.embeddings.patch_embeddings.num_patches)
+
+        self.post_init()
+
+    def patchify(self, pixel_values: torch.Tensor, interpolate_pos_encoding: bool = False) -> torch.Tensor:
+        """
+        Args:
+            pixel_values (`torch.FloatTensor` of shape `(batch_size, num_channels, height, width)`):
+                Pixel values.
+            interpolate_pos_encoding (`bool`, *optional*, default `False`):
+                interpolation flag passed during the forward pass.
+
+        Returns:
+            `torch.FloatTensor` of shape `(batch_size, num_patches, patch_size**2 * num_channels)`:
+                Patchified pixel values.
+        """
+        patch_size, num_channels = self.config.patch_size, self.config.num_channels
+        if not interpolate_pos_encoding and (
+            pixel_values.shape[2] != pixel_values.shape[3] or pixel_values.shape[2] % patch_size != 0
+        ):
+            raise ValueError("Make sure the pixel values have a squared size that is divisible by the patch size")
+        if pixel_values.shape[1] != num_channels:
+            raise ValueError(
+                "Make sure the number of channels of the pixel values is equal to the one set in the configuration"
+            )
+
+        batch_size = pixel_values.shape[0]
+        num_patches_h = pixel_values.shape[2] // patch_size
+        num_patches_w = pixel_values.shape[3] // patch_size
+        patchified_pixel_values = pixel_values.reshape(
+            batch_size,
+            num_channels,
+            num_patches_h,
+            patch_size,
+            num_patches_w,
+            patch_size,
+        )
+        patchified_pixel_values = patchified_pixel_values.permute(0, 2, 4, 3, 5, 1)
+        patchified_pixel_values = patchified_pixel_values.reshape(
+            batch_size,
+            num_patches_h * num_patches_w,
+            patch_size**2 * num_channels,
+        )
+        return patchified_pixel_values
+
+    def unpatchify(
+        self,
+        patchified_pixel_values: torch.Tensor,
+        original_image_size: tuple[int, int] | None = None,
+    ) -> torch.Tensor:
+        """
+        Args:
+            patchified_pixel_values (`torch.FloatTensor` of shape `(batch_size, num_patches, patch_size**2 * num_channels)`):
+                Patchified pixel values.
+            original_image_size (`tuple[int, int]`, *optional*):
+                Original image size.
+
+        Returns:
+            `torch.FloatTensor` of shape `(batch_size, num_channels, height, width)`:
+                Pixel values.
+        """
+        patch_size, num_channels = self.config.patch_size, self.config.num_channels
+        original_image_size = (
+            original_image_size
+            if original_image_size is not None
+            else (self.config.image_size, self.config.image_size)
+        )
+        original_height, original_width = original_image_size
+        num_patches_h = original_height // patch_size
+        num_patches_w = original_width // patch_size
+        if num_patches_h * num_patches_w != patchified_pixel_values.shape[1]:
+            raise ValueError(
+                f"The number of patches in the patchified pixel values {patchified_pixel_values.shape[1]}, does not match the number of patches on original image {num_patches_h}*{num_patches_w}"
+            )
+
+        batch_size = patchified_pixel_values.shape[0]
+        patchified_pixel_values = patchified_pixel_values.reshape(
+            batch_size,
+            num_patches_h,
+            num_patches_w,
+            patch_size,
+            patch_size,
+            num_channels,
+        )
+        patchified_pixel_values = patchified_pixel_values.permute(0, 5, 1, 3, 2, 4)
+        pixel_values = patchified_pixel_values.reshape(
+            batch_size,
+            num_channels,
+            num_patches_h * patch_size,
+            num_patches_w * patch_size,
+        )
+        return pixel_values
+
+    @can_return_tuple
+    @auto_docstring
+    def forward(
+        self,
+        pixel_values: torch.Tensor | None = None,
+        noise: torch.Tensor | None = None,
+        interpolate_pos_encoding: bool | None = None,
+        attention_mask: torch.Tensor | None = None,
+        **kwargs: Unpack[TransformersKwargs],
+    ) -> ViTMAEForPreTrainingOutput:
+        r"""
+        noise (`torch.FloatTensor` of shape `(batch_size, sequence_length)`, *optional*):
+            Mainly used for testing purposes to control randomness and maintain the reproducibility
+        interpolate_pos_encoding (`bool`, *optional*, default `False`):
+            Whether to interpolate the pre-trained position encodings. This is mainly used to use the model on higher
+            resolution images.
+
+        Examples:
+
+        ```python
+        >>> from transformers import AutoImageProcessor, ViTMAEForPreTraining
+        >>> from PIL import Image
+        >>> import httpx
+        >>> from io import BytesIO
+
+        >>> url = "http://images.cocodataset.org/val2017/000000039769.jpg"
+        >>> with httpx.stream("GET", url) as response:
+        ...     image = Image.open(BytesIO(response.read())).convert("RGB")
+
+        >>> image_processor = AutoImageProcessor.from_pretrained("facebook/vit-mae-base")
+        >>> model = ViTMAEForPreTraining.from_pretrained("facebook/vit-mae-base")
+
+        >>> inputs = image_processor(images=image, return_tensors="pt")
+        >>> outputs = model(**inputs)
+        >>> loss = outputs.loss
+        >>> mask = outputs.mask
+        >>> ids_restore = outputs.ids_restore
+        ```"""
+
+        outputs: ViTMAEModelOutput = self.vit(
+            pixel_values,
+            noise=noise,
+            interpolate_pos_encoding=interpolate_pos_encoding,
+            attention_mask=attention_mask,
+            **kwargs,
+        )
+
+        latent = outputs.last_hidden_state
+        ids_restore = outputs.ids_restore
+        mask = outputs.mask
+
+        decoder_outputs: ViTMAEDecoderOutput = self.decoder(
+            latent, ids_restore, interpolate_pos_encoding=interpolate_pos_encoding
+        )
+        logits = decoder_outputs.logits
+
+        # Pixel reconstruction loss: MSE between predicted and ground-truth patch pixels (optionally per-patch normalized), averaged only over masked locations.
+        target = self.patchify(pixel_values, interpolate_pos_encoding=interpolate_pos_encoding)
+        if self.config.norm_pix_loss:
+            mean = target.mean(dim=-1, keepdim=True)
+            var = target.var(dim=-1, keepdim=True)
+            target = (target - mean) / (var + 1.0e-6) ** 0.5
+        loss = (logits - target) ** 2
+        loss = loss.mean(dim=-1)
+        loss = (loss * mask).sum() / mask.sum()
+
+        return ViTMAEForPreTrainingOutput(
+            loss=loss,
+            logits=logits,
+            mask=mask,
+            ids_restore=ids_restore,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )
+
+
+__all__ = [
+    "ViTMAEForPreTraining",
+    "ViTMAEModel",
+    "ViTMAEPreTrainedModel",
+]

LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/transformers/models/vivit/__init__.py

@@ -0,0 +1,28 @@
+# Copyright 2024 The HuggingFace Team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+from typing import TYPE_CHECKING
+
+from ...utils import _LazyModule
+from ...utils.import_utils import define_import_structure
+
+
+if TYPE_CHECKING:
+    from .configuration_vivit import *
+    from .image_processing_vivit import *
+    from .modeling_vivit import *
+else:
+    import sys
+
+    _file = globals()["__file__"]
+    sys.modules[__name__] = _LazyModule(__name__, _file, define_import_structure(_file), module_spec=__spec__)

LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/transformers/models/vivit/configuration_vivit.py

@@ -0,0 +1,74 @@
+# Copyright 2023 The HuggingFace Inc. team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""ViViT model configuration"""
+
+from huggingface_hub.dataclasses import strict
+
+from ...configuration_utils import PreTrainedConfig
+from ...utils import auto_docstring
+
+
+@auto_docstring(checkpoint="google/vivit-b-16x2-kinetics400")
+@strict
+class VivitConfig(PreTrainedConfig):
+    r"""
+    num_frames (`int`, *optional*, defaults to 32):
+        The number of frames in each video.
+    tubelet_size (`list[int]`, *optional*, defaults to `[2, 16, 16]`):
+        The size (resolution) of each tubelet.
+    pooler_output_size (`int`, *optional*):
+        Dimensionality of the pooler layer. If None, defaults to `hidden_size`.
+    pooler_act (`str`, *optional*, defaults to `"tanh"`):
+        The activation function to be used by the pooler.
+
+    Example:
+
+    ```python
+    >>> from transformers import VivitConfig, VivitModel
+
+    >>> # Initializing a ViViT google/vivit-b-16x2-kinetics400 style configuration
+    >>> configuration = VivitConfig()
+
+    >>> # Initializing a model (with random weights) from the google/vivit-b-16x2-kinetics400 style configuration
+    >>> model = VivitModel(configuration)
+
+    >>> # Accessing the model configuration
+    >>> configuration = model.config
+    ```"""
+
+    model_type = "vivit"
+
+    image_size: int | list[int] | tuple[int, int] = 224
+    num_frames: int = 32
+    tubelet_size: list[int] | tuple[int, ...] = (2, 16, 16)
+    num_channels: int = 3
+    hidden_size: int = 768
+    num_hidden_layers: int = 12
+    num_attention_heads: int = 12
+    intermediate_size: int = 3072
+    hidden_act: str = "gelu_fast"
+    hidden_dropout_prob: float | int = 0.0
+    attention_probs_dropout_prob: float | int = 0.0
+    initializer_range: float = 0.02
+    layer_norm_eps: float = 1e-06
+    qkv_bias: bool = True
+    pooler_output_size: int | None = None
+    pooler_act: str = "tanh"
+
+    def __post_init__(self, **kwargs):
+        self.pooler_output_size = self.pooler_output_size if self.pooler_output_size else self.hidden_size
+        super().__post_init__(**kwargs)
+
+
+__all__ = ["VivitConfig"]

LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/transformers/models/vivit/image_processing_vivit.py

@@ -0,0 +1,398 @@
+# Copyright 2023 The HuggingFace Inc. team.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""Image processor class for Vivit."""
+
+import numpy as np
+
+from transformers.utils import is_vision_available
+from transformers.utils.generic import TensorType
+
+from ...image_processing_utils import BaseImageProcessor, BatchFeature, get_size_dict
+from ...image_transforms import (
+    get_resize_output_image_size,
+    rescale,
+    resize,
+    to_channel_dimension_format,
+)
+from ...image_utils import (
+    IMAGENET_STANDARD_MEAN,
+    IMAGENET_STANDARD_STD,
+    ChannelDimension,
+    ImageInput,
+    PILImageResampling,
+    infer_channel_dimension_format,
+    is_scaled_image,
+    is_valid_image,
+    to_numpy_array,
+    valid_images,
+    validate_preprocess_arguments,
+)
+from ...utils import filter_out_non_signature_kwargs, logging
+
+
+if is_vision_available():
+    import PIL
+
+logger = logging.get_logger(__name__)
+
+
+def make_batched(videos) -> list[list[ImageInput]]:
+    if isinstance(videos, (list, tuple)) and isinstance(videos[0], (list, tuple)) and is_valid_image(videos[0][0]):
+        return videos
+
+    elif isinstance(videos, (list, tuple)) and is_valid_image(videos[0]):
+        return [videos]
+
+    elif is_valid_image(videos):
+        return [[videos]]
+
+    raise ValueError(f"Could not make batched video from {videos}")
+
+
+class VivitImageProcessor(BaseImageProcessor):
+    r"""
+    Constructs a Vivit image processor.
+
+    Args:
+        do_resize (`bool`, *optional*, defaults to `True`):
+            Whether to resize the image's (height, width) dimensions to the specified `size`. Can be overridden by the
+            `do_resize` parameter in the `preprocess` method.
+        size (`dict[str, int]` *optional*, defaults to `{"shortest_edge": 256}`):
+            Size of the output image after resizing. The shortest edge of the image will be resized to
+            `size["shortest_edge"]` while maintaining the aspect ratio of the original image. Can be overridden by
+            `size` in the `preprocess` method.
+        resample (`PILImageResampling`, *optional*, defaults to `Resampling.BILINEAR`):
+            Resampling filter to use if resizing the image. Can be overridden by the `resample` parameter in the
+            `preprocess` method.
+        do_center_crop (`bool`, *optional*, defaults to `True`):
+            Whether to center crop the image to the specified `crop_size`. Can be overridden by the `do_center_crop`
+            parameter in the `preprocess` method.
+        crop_size (`dict[str, int]`, *optional*, defaults to `{"height": 224, "width": 224}`):
+            Size of the image after applying the center crop. Can be overridden by the `crop_size` parameter in the
+            `preprocess` method.
+        do_rescale (`bool`, *optional*, defaults to `True`):
+            Whether to rescale the image by the specified scale `rescale_factor`. Can be overridden by the `do_rescale`
+            parameter in the `preprocess` method.
+        rescale_factor (`int` or `float`, *optional*, defaults to `1/127.5`):
+            Defines the scale factor to use if rescaling the image. Can be overridden by the `rescale_factor` parameter
+            in the `preprocess` method.
+        offset (`bool`, *optional*, defaults to `True`):
+            Whether to scale the image in both negative and positive directions. Can be overridden by the `offset` in
+            the `preprocess` method.
+        do_normalize (`bool`, *optional*, defaults to `True`):
+            Whether to normalize the image. Can be overridden by the `do_normalize` parameter in the `preprocess`
+            method.
+        image_mean (`float` or `list[float]`, *optional*, defaults to `IMAGENET_STANDARD_MEAN`):
+            Mean to use if normalizing the image. This is a float or list of floats the length of the number of
+            channels in the image. Can be overridden by the `image_mean` parameter in the `preprocess` method.
+        image_std (`float` or `list[float]`, *optional*, defaults to `IMAGENET_STANDARD_STD`):
+            Standard deviation to use if normalizing the image. This is a float or list of floats the length of the
+            number of channels in the image. Can be overridden by the `image_std` parameter in the `preprocess` method.
+    """
+
+    model_input_names = ["pixel_values"]
+
+    def __init__(
+        self,
+        do_resize: bool = True,
+        size: dict[str, int] | None = None,
+        resample: PILImageResampling = PILImageResampling.BILINEAR,
+        do_center_crop: bool = True,
+        crop_size: dict[str, int] | None = None,
+        do_rescale: bool = True,
+        rescale_factor: int | float = 1 / 127.5,
+        offset: bool = True,
+        do_normalize: bool = True,
+        image_mean: float | list[float] | None = None,
+        image_std: float | list[float] | None = None,
+        **kwargs,
+    ) -> None:
+        super().__init__(**kwargs)
+        size = size if size is not None else {"shortest_edge": 256}
+        size = get_size_dict(size, default_to_square=False)
+        crop_size = crop_size if crop_size is not None else {"height": 224, "width": 224}
+        crop_size = get_size_dict(crop_size, param_name="crop_size")
+
+        self.do_resize = do_resize
+        self.size = size
+        self.do_center_crop = do_center_crop
+        self.crop_size = crop_size
+        self.resample = resample
+        self.do_rescale = do_rescale
+        self.rescale_factor = rescale_factor
+        self.offset = offset
+        self.do_normalize = do_normalize
+        self.image_mean = image_mean if image_mean is not None else IMAGENET_STANDARD_MEAN
+        self.image_std = image_std if image_std is not None else IMAGENET_STANDARD_STD
+
+    def resize(
+        self,
+        image: np.ndarray,
+        size: dict[str, int],
+        resample: PILImageResampling = PILImageResampling.BILINEAR,
+        data_format: str | ChannelDimension | None = None,
+        input_data_format: str | ChannelDimension | None = None,
+        **kwargs,
+    ) -> np.ndarray:
+        """
+        Resize an image.
+
+        Args:
+            image (`np.ndarray`):
+                Image to resize.
+            size (`dict[str, int]`):
+                Size of the output image. If `size` is of the form `{"height": h, "width": w}`, the output image will
+                have the size `(h, w)`. If `size` is of the form `{"shortest_edge": s}`, the output image will have its
+                shortest edge of length `s` while keeping the aspect ratio of the original image.
+            resample (`PILImageResampling`, *optional*, defaults to `PILImageResampling.BILINEAR`):
+                Resampling filter to use when resiizing the image.
+            data_format (`str` or `ChannelDimension`, *optional*):
+                The channel dimension format of the image. If not provided, it will be the same as the input image.
+            input_data_format (`str` or `ChannelDimension`, *optional*):
+                The channel dimension format of the input image. If not provided, it will be inferred.
+        """
+        size = get_size_dict(size, default_to_square=False)
+        if "shortest_edge" in size:
+            output_size = get_resize_output_image_size(
+                image, size["shortest_edge"], default_to_square=False, input_data_format=input_data_format
+            )
+        elif "height" in size and "width" in size:
+            output_size = (size["height"], size["width"])
+        else:
+            raise ValueError(f"Size must have 'height' and 'width' or 'shortest_edge' as keys. Got {size.keys()}")
+        return resize(
+            image,
+            size=output_size,
+            resample=resample,
+            data_format=data_format,
+            input_data_format=input_data_format,
+            **kwargs,
+        )
+
+    def rescale(
+        self,
+        image: np.ndarray,
+        scale: int | float,
+        offset: bool = True,
+        data_format: str | ChannelDimension | None = None,
+        input_data_format: str | ChannelDimension | None = None,
+        **kwargs,
+    ):
+        """
+        Rescale an image by a scale factor.
+
+        If `offset` is `True`, the image has its values rescaled by `scale` and then offset by 1. If `scale` is
+        1/127.5, the image is rescaled between [-1, 1].
+            image = image * scale - 1
+
+        If `offset` is `False`, and `scale` is 1/255, the image is rescaled between [0, 1].
+            image = image * scale
+
+        Args:
+            image (`np.ndarray`):
+                Image to rescale.
+            scale (`int` or `float`):
+                Scale to apply to the image.
+            offset (`bool`, *optional*):
+                Whether to scale the image in both negative and positive directions.
+            data_format (`str` or `ChannelDimension`, *optional*):
+                The channel dimension format of the image. If not provided, it will be the same as the input image.
+            input_data_format (`ChannelDimension` or `str`, *optional*):
+                The channel dimension format of the input image. If not provided, it will be inferred.
+        """
+        rescaled_image = rescale(
+            image, scale=scale, data_format=data_format, input_data_format=input_data_format, **kwargs
+        )
+
+        if offset:
+            rescaled_image = rescaled_image - 1
+
+        return rescaled_image
+
+    def _preprocess_image(
+        self,
+        image: ImageInput,
+        do_resize: bool | None = None,
+        size: dict[str, int] | None = None,
+        resample: PILImageResampling | None = None,
+        do_center_crop: bool | None = None,
+        crop_size: dict[str, int] | None = None,
+        do_rescale: bool | None = None,
+        rescale_factor: float | None = None,
+        offset: bool | None = None,
+        do_normalize: bool | None = None,
+        image_mean: float | list[float] | None = None,
+        image_std: float | list[float] | None = None,
+        data_format: ChannelDimension | None = ChannelDimension.FIRST,
+        input_data_format: str | ChannelDimension | None = None,
+    ) -> np.ndarray:
+        """Preprocesses a single image."""
+
+        validate_preprocess_arguments(
+            do_rescale=do_rescale,
+            rescale_factor=rescale_factor,
+            do_normalize=do_normalize,
+            image_mean=image_mean,
+            image_std=image_std,
+            do_center_crop=do_center_crop,
+            crop_size=crop_size,
+            do_resize=do_resize,
+            size=size,
+            resample=resample,
+        )
+
+        if offset and not do_rescale:
+            raise ValueError("For offset, do_rescale must also be set to True.")
+
+        # All transformations expect numpy arrays.
+        image = to_numpy_array(image)
+
+        if do_rescale and is_scaled_image(image):
+            logger.warning_once(
+                "It looks like you are trying to rescale already rescaled images. If the input"
+                " images have pixel values between 0 and 1, set `do_rescale=False` to avoid rescaling them again."
+            )
+
+        if input_data_format is None:
+            input_data_format = infer_channel_dimension_format(image)
+
+        if do_resize:
+            image = self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format)
+
+        if do_center_crop:
+            image = self.center_crop(image, size=crop_size, input_data_format=input_data_format)
+
+        if do_rescale:
+            image = self.rescale(image=image, scale=rescale_factor, offset=offset, input_data_format=input_data_format)
+
+        if do_normalize:
+            image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format)
+
+        image = to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format)
+        return image
+
+    @filter_out_non_signature_kwargs()
+    def preprocess(
+        self,
+        videos: ImageInput,
+        do_resize: bool | None = None,
+        size: dict[str, int] | None = None,
+        resample: PILImageResampling | None = None,
+        do_center_crop: bool | None = None,
+        crop_size: dict[str, int] | None = None,
+        do_rescale: bool | None = None,
+        rescale_factor: float | None = None,
+        offset: bool | None = None,
+        do_normalize: bool | None = None,
+        image_mean: float | list[float] | None = None,
+        image_std: float | list[float] | None = None,
+        return_tensors: str | TensorType | None = None,
+        data_format: ChannelDimension = ChannelDimension.FIRST,
+        input_data_format: str | ChannelDimension | None = None,
+    ) -> PIL.Image.Image:
+        """
+        Preprocess an image or batch of images.
+
+        Args:
+            videos (`ImageInput`):
+                Video frames to preprocess. Expects a single or batch of video frames with pixel values ranging from 0
+                to 255. If passing in frames with pixel values between 0 and 1, set `do_rescale=False`.
+            do_resize (`bool`, *optional*, defaults to `self.do_resize`):
+                Whether to resize the image.
+            size (`dict[str, int]`, *optional*, defaults to `self.size`):
+                Size of the image after applying resize.
+            resample (`PILImageResampling`, *optional*, defaults to `self.resample`):
+                Resampling filter to use if resizing the image. This can be one of the enum `PILImageResampling`, Only
+                has an effect if `do_resize` is set to `True`.
+            do_center_crop (`bool`, *optional*, defaults to `self.do_centre_crop`):
+                Whether to centre crop the image.
+            crop_size (`dict[str, int]`, *optional*, defaults to `self.crop_size`):
+                Size of the image after applying the centre crop.
+            do_rescale (`bool`, *optional*, defaults to `self.do_rescale`):
+                Whether to rescale the image values between `[-1 - 1]` if `offset` is `True`, `[0, 1]` otherwise.
+            rescale_factor (`float`, *optional*, defaults to `self.rescale_factor`):
+                Rescale factor to rescale the image by if `do_rescale` is set to `True`.
+            offset (`bool`, *optional*, defaults to `self.offset`):
+                Whether to scale the image in both negative and positive directions.
+            do_normalize (`bool`, *optional*, defaults to `self.do_normalize`):
+                Whether to normalize the image.
+            image_mean (`float` or `list[float]`, *optional*, defaults to `self.image_mean`):
+                Image mean.
+            image_std (`float` or `list[float]`, *optional*, defaults to `self.image_std`):
+                Image standard deviation.
+            return_tensors (`str` or `TensorType`, *optional*):
+                The type of tensors to return. Can be one of:
+                    - Unset: Return a list of `np.ndarray`.
+                    - `TensorType.PYTORCH` or `'pt'`: Return a batch of type `torch.Tensor`.
+                    - `TensorType.NUMPY` or `'np'`: Return a batch of type `np.ndarray`.
+            data_format (`ChannelDimension` or `str`, *optional*, defaults to `ChannelDimension.FIRST`):
+                The channel dimension format for the output image. Can be one of:
+                    - `ChannelDimension.FIRST`: image in (num_channels, height, width) format.
+                    - `ChannelDimension.LAST`: image in (height, width, num_channels) format.
+                    - Unset: Use the inferred channel dimension format of the input image.
+            input_data_format (`ChannelDimension` or `str`, *optional*):
+                The channel dimension format for the input image. If unset, the channel dimension format is inferred
+                from the input image. Can be one of:
+                - `"channels_first"` or `ChannelDimension.FIRST`: image in (num_channels, height, width) format.
+                - `"channels_last"` or `ChannelDimension.LAST`: image in (height, width, num_channels) format.
+                - `"none"` or `ChannelDimension.NONE`: image in (height, width) format.
+        """
+        do_resize = do_resize if do_resize is not None else self.do_resize
+        resample = resample if resample is not None else self.resample
+        do_center_crop = do_center_crop if do_center_crop is not None else self.do_center_crop
+        do_rescale = do_rescale if do_rescale is not None else self.do_rescale
+        rescale_factor = rescale_factor if rescale_factor is not None else self.rescale_factor
+        offset = offset if offset is not None else self.offset
+        do_normalize = do_normalize if do_normalize is not None else self.do_normalize
+        image_mean = image_mean if image_mean is not None else self.image_mean
+        image_std = image_std if image_std is not None else self.image_std
+
+        size = size if size is not None else self.size
+        size = get_size_dict(size, default_to_square=False)
+        crop_size = crop_size if crop_size is not None else self.crop_size
+        crop_size = get_size_dict(crop_size, param_name="crop_size")
+
+        if not valid_images(videos):
+            raise ValueError("Invalid image type. Must be of type PIL.Image.Image, numpy.ndarray, or torch.Tensor")
+
+        videos = make_batched(videos)
+
+        videos = [
+            [
+                self._preprocess_image(
+                    image=img,
+                    do_resize=do_resize,
+                    size=size,
+                    resample=resample,
+                    do_center_crop=do_center_crop,
+                    crop_size=crop_size,
+                    do_rescale=do_rescale,
+                    rescale_factor=rescale_factor,
+                    offset=offset,
+                    do_normalize=do_normalize,
+                    image_mean=image_mean,
+                    image_std=image_std,
+                    data_format=data_format,
+                    input_data_format=input_data_format,
+                )
+                for img in video
+            ]
+            for video in videos
+        ]
+
+        data = {"pixel_values": videos}
+        return BatchFeature(data=data, tensor_type=return_tensors)
+
+
+__all__ = ["VivitImageProcessor"]

LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/transformers/models/vivit/modeling_vivit.py

@@ -0,0 +1,570 @@
+#                🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
+#           This file was automatically generated from src/transformers/models/vivit/modular_vivit.py.
+#               Do NOT edit this file manually as any edits will be overwritten by the generation of
+#             the file from the modular. If any change should be done, please apply the change to the
+#                          modular_vivit.py file directly. One of our CI enforces this.
+#                🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
+# Copyright 2023 Google AI and The HuggingFace Inc. team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from collections.abc import Callable, Iterable
+
+import torch
+from torch import nn
+
+from ... import initialization as init
+from ...activations import ACT2FN
+from ...masking_utils import create_bidirectional_mask
+from ...modeling_layers import GradientCheckpointingLayer
+from ...modeling_outputs import BaseModelOutput, BaseModelOutputWithPooling, ImageClassifierOutput
+from ...modeling_utils import ALL_ATTENTION_FUNCTIONS, PreTrainedModel
+from ...processing_utils import Unpack
+from ...utils import TransformersKwargs, auto_docstring, torch_int
+from ...utils.generic import can_return_tuple, merge_with_config_defaults
+from ...utils.output_capturing import capture_outputs
+from .configuration_vivit import VivitConfig
+
+
+class VivitTubeletEmbeddings(nn.Module):
+    """
+    This class turns `pixel_values` of shape `(batch_size, num_frames, num_channels, height, width)` into the initial
+    `hidden_states` (tubelet embeddings) of shape `(batch_size, seq_length, hidden_size)` to be consumed by a
+    Transformer encoder.
+
+    The seq_length equals (num_frames // tubelet_size[0]) * (height // tubelet_size[1]) * (width // tubelet_size[2]).
+    """
+
+    def __init__(self, config: VivitConfig):
+        super().__init__()
+        tubelet_size = config.tubelet_size
+        image_size = config.image_size
+        image_size = image_size if isinstance(image_size, Iterable) else (image_size, image_size)
+
+        self.num_patches = (
+            (config.num_frames // tubelet_size[0])
+            * (image_size[0] // tubelet_size[1])
+            * (image_size[1] // tubelet_size[2])
+        )
+        self.image_size = image_size
+        self.projection = nn.Conv3d(
+            config.num_channels, config.hidden_size, kernel_size=tubelet_size, stride=tubelet_size
+        )
+
+    def forward(self, pixel_values: torch.Tensor) -> torch.Tensor:
+        # transpose (batch_size, num_channels, num_frames, height, width) for Conv3d
+        pixel_values = pixel_values.transpose(1, 2)
+        return self.projection(pixel_values).flatten(2).transpose(1, 2)
+
+
+class VivitEmbeddings(nn.Module):
+    """
+    Construct the CLS token, position and tubelet patch embeddings for video input.
+    """
+
+    def __init__(self, config: VivitConfig):
+        super().__init__()
+
+        self.cls_token = nn.Parameter(torch.zeros(1, 1, config.hidden_size))
+        self.patch_embeddings = VivitTubeletEmbeddings(config)
+        num_patches = self.patch_embeddings.num_patches
+        self.position_embeddings = nn.Parameter(torch.zeros(1, num_patches + 1, config.hidden_size))
+        self.dropout = nn.Dropout(config.hidden_dropout_prob)
+        # patch_size is the spatial (height, width) part of the tubelet for pos encoding interpolation
+        self.patch_size = config.tubelet_size[1:]
+        self.image_size = self.patch_embeddings.image_size
+
+    def interpolate_pos_encoding(self, embeddings: torch.Tensor, height: int, width: int) -> torch.Tensor:
+        """
+        This method allows to interpolate the pre-trained position encodings, to be able to use the model on higher resolution
+        images. This method is also adapted to support torch.jit tracing.
+
+        Adapted from:
+        - https://github.com/facebookresearch/dino/blob/de9ee3df6cf39fac952ab558447af1fa1365362a/vision_transformer.py#L174-L194, and
+        - https://github.com/facebookresearch/dinov2/blob/e1277af2ba9496fbadf7aec6eba56e8d882d1e35/dinov2/models/vision_transformer.py#L179-L211
+        """
+        num_patches = self.patch_embeddings.num_patches
+        num_positions = self.position_embeddings.shape[1] - 1
+
+        # always interpolate when tracing to ensure the exported model works for dynamic input shapes
+        if not torch.jit.is_tracing() and num_patches == num_positions and height == width:
+            return self.position_embeddings
+
+        class_pos_embed = self.position_embeddings[:, :1]
+        patch_pos_embed = self.position_embeddings[:, 1:]
+
+        dim = embeddings.shape[-1]
+        # patch_size is a 2-tuple (height, width) for the spatial tubelet dimensions
+        new_height = height // self.patch_size[0]  # noqa: F841
+        new_width = width // self.patch_size[1]  # noqa: F841
+
+        sqrt_num_positions = torch_int(num_positions**0.5)
+        patch_pos_embed = patch_pos_embed.reshape(1, sqrt_num_positions, sqrt_num_positions, dim)
+        patch_pos_embed = patch_pos_embed.permute(0, 3, 1, 2)
+
+        patch_pos_embed = nn.functional.interpolate(
+            patch_pos_embed,
+            size=(new_height, new_width),
+            mode="bicubic",
+            align_corners=False,
+        )
+
+        patch_pos_embed = patch_pos_embed.permute(0, 2, 3, 1).view(1, -1, dim)
+
+        return torch.cat((class_pos_embed, patch_pos_embed), dim=1)
+
+    def forward(self, pixel_values: torch.Tensor, interpolate_pos_encoding: bool = False) -> torch.Tensor:
+        batch_size, num_frames, num_channels, height, width = pixel_values.shape
+        embeddings = self.patch_embeddings(pixel_values)
+
+        # add the [CLS] token to the embedded patch tokens
+        cls_tokens = self.cls_token.expand(batch_size, -1, -1)
+        embeddings = torch.cat((cls_tokens, embeddings), dim=1)
+
+        if interpolate_pos_encoding:
+            embeddings = embeddings + self.interpolate_pos_encoding(embeddings, height, width)
+        else:
+            if height != self.image_size[0] or width != self.image_size[1]:
+                raise ValueError(
+                    f"Input image size ({height}*{width}) doesn't match model"
+                    f" ({self.image_size[0]}*{self.image_size[1]})."
+                )
+            embeddings = embeddings + self.position_embeddings
+
+        embeddings = self.dropout(embeddings)
+
+        return embeddings
+
+
+def eager_attention_forward(
+    module: nn.Module,
+    query: torch.Tensor,
+    key: torch.Tensor,
+    value: torch.Tensor,
+    attention_mask: torch.Tensor | None,
+    scaling: float | None = None,
+    dropout: float = 0.0,
+    **kwargs: Unpack[TransformersKwargs],
+):
+    if scaling is None:
+        scaling = query.size(-1) ** -0.5
+
+    # Take the dot product between "query" and "key" to get the raw attention scores.
+    attn_weights = torch.matmul(query, key.transpose(2, 3)) * scaling
+
+    if attention_mask is not None:
+        attn_weights = attn_weights + attention_mask
+
+    attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query.dtype)
+    attn_weights = nn.functional.dropout(attn_weights, p=dropout, training=module.training)
+
+    attn_output = torch.matmul(attn_weights, value)
+    attn_output = attn_output.transpose(1, 2).contiguous()
+
+    return attn_output, attn_weights
+
+
+class VivitAttention(nn.Module):
+    def __init__(self, config: VivitConfig):
+        super().__init__()
+        self.config = config
+        self.num_attention_heads = config.num_attention_heads
+        self.head_dim = getattr(config, "head_dim", config.hidden_size // config.num_attention_heads)
+        self.attention_dropout = config.attention_probs_dropout_prob
+        self.scaling = self.head_dim**-0.5
+        self.is_causal = False
+
+        self.q_proj = nn.Linear(config.hidden_size, config.num_attention_heads * self.head_dim, bias=config.qkv_bias)
+        self.k_proj = nn.Linear(config.hidden_size, config.num_attention_heads * self.head_dim, bias=config.qkv_bias)
+        self.v_proj = nn.Linear(config.hidden_size, config.num_attention_heads * self.head_dim, bias=config.qkv_bias)
+        self.o_proj = nn.Linear(config.num_attention_heads * self.head_dim, config.hidden_size, bias=True)
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: torch.Tensor | None = None,
+        **kwargs: Unpack[TransformersKwargs],
+    ) -> tuple[torch.Tensor, torch.Tensor]:
+        input_shape = hidden_states.shape[:-1]
+        hidden_shape = (*input_shape, -1, self.head_dim)
+
+        query_states = self.q_proj(hidden_states).view(hidden_shape).transpose(1, 2)
+        key_states = self.k_proj(hidden_states).view(hidden_shape).transpose(1, 2)
+        value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)
+
+        attention_interface: Callable = ALL_ATTENTION_FUNCTIONS.get_interface(
+            self.config._attn_implementation, eager_attention_forward
+        )
+
+        attn_output, attn_weights = attention_interface(
+            self,
+            query_states,
+            key_states,
+            value_states,
+            attention_mask,
+            dropout=0.0 if not self.training else self.attention_dropout,
+            scaling=self.scaling,
+            **kwargs,
+        )
+
+        attn_output = attn_output.reshape(*input_shape, -1).contiguous()
+        attn_output = self.o_proj(attn_output)
+
+        return attn_output, attn_weights
+
+
+class VivitMLP(nn.Module):
+    def __init__(self, config: VivitConfig):
+        super().__init__()
+        self.config = config
+        self.activation_fn = ACT2FN[config.hidden_act]
+        self.fc1 = nn.Linear(config.hidden_size, config.intermediate_size)
+        self.fc2 = nn.Linear(config.intermediate_size, config.hidden_size)
+
+    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
+        hidden_states = self.fc1(hidden_states)
+        hidden_states = self.activation_fn(hidden_states)
+        hidden_states = self.fc2(hidden_states)
+
+        return hidden_states
+
+
+class VivitLayer(GradientCheckpointingLayer):
+    def __init__(self, config: VivitConfig):
+        super().__init__()
+        self.attention = VivitAttention(config)
+        self.layernorm_before = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
+        self.layernorm_after = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
+        self.mlp = VivitMLP(config)
+        self.dropout = nn.Dropout(config.hidden_dropout_prob)
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: torch.Tensor | None = None,
+        **kwargs: Unpack[TransformersKwargs],
+    ) -> torch.Tensor:
+        # Self Attention
+        residual = hidden_states
+        hidden_states = self.layernorm_before(hidden_states)
+        hidden_states, _ = self.attention(hidden_states, attention_mask, **kwargs)
+        hidden_states = self.dropout(hidden_states)
+        hidden_states = hidden_states + residual
+
+        # Fully Connected
+        residual = hidden_states
+        hidden_states = self.layernorm_after(hidden_states)
+        hidden_states = self.mlp(hidden_states)
+        hidden_states = self.dropout(hidden_states)
+        hidden_states = hidden_states + residual
+
+        return hidden_states
+
+
+class VivitPooler(nn.Module):
+    def __init__(self, config: VivitConfig):
+        super().__init__()
+        self.dense = nn.Linear(config.hidden_size, config.pooler_output_size)
+        self.activation = ACT2FN[config.pooler_act]
+
+    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
+        # We "pool" the model by simply taking the hidden state corresponding
+        # to the first token.
+        first_token_tensor = hidden_states[:, 0]
+        pooled_output = self.dense(first_token_tensor)
+        pooled_output = self.activation(pooled_output)
+        return pooled_output
+
+
+@auto_docstring
+class VivitPreTrainedModel(PreTrainedModel):
+    config: VivitConfig
+    base_model_prefix = "vivit"
+    main_input_name = "pixel_values"
+    input_modalities = ("video",)
+    supports_gradient_checkpointing = True
+    _no_split_modules = ["VivitEmbeddings", "VivitLayer"]
+    _supports_sdpa = True
+    _supports_flash_attn = True
+    _supports_flex_attn = True
+    _supports_attention_backend = True
+    _can_compile_fullgraph = True
+    _can_record_outputs = {
+        "hidden_states": VivitLayer,
+        "attentions": VivitAttention,
+    }
+    _input_embed_layer = "patch_embeddings"
+
+    @torch.no_grad()
+    def _init_weights(self, module):
+        """Initialize the weights"""
+        super()._init_weights(module)
+        if isinstance(module, VivitEmbeddings):
+            init.zeros_(module.cls_token)
+            init.zeros_(module.position_embeddings)
+
+
+@auto_docstring
+class VivitModel(VivitPreTrainedModel):
+    def __init__(self, config: VivitConfig, add_pooling_layer: bool = True):
+        r"""
+        add_pooling_layer (bool, *optional*, defaults to `True`):
+            Whether to add a pooling layer
+        """
+        super().__init__(config)
+        self.config = config
+        self.embeddings = VivitEmbeddings(config)
+        self.layers = nn.ModuleList([VivitLayer(config) for _ in range(config.num_hidden_layers)])
+        self.layernorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
+        self.pooler = VivitPooler(config) if add_pooling_layer else None
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    @merge_with_config_defaults
+    @capture_outputs(tie_last_hidden_states=False)
+    @auto_docstring
+    def forward(
+        self,
+        pixel_values: torch.FloatTensor | None = None,
+        interpolate_pos_encoding: bool = False,
+        attention_mask: torch.Tensor | None = None,
+        **kwargs: Unpack[TransformersKwargs],
+    ) -> BaseModelOutputWithPooling:
+        r"""
+        Examples:
+
+        ```python
+        >>> import av
+        >>> import numpy as np
+
+        >>> from transformers import VivitImageProcessor, VivitModel
+        >>> from huggingface_hub import hf_hub_download
+
+        >>> np.random.seed(0)
+
+
+        >>> def read_video_pyav(container, indices):
+        ...     '''
+        ...     Decode the video with PyAV decoder.
+        ...     Args:
+        ...         container (`av.container.input.InputContainer`): PyAV container.
+        ...         indices (`list[int]`): List of frame indices to decode.
+        ...     Returns:
+        ...         result (np.ndarray): np array of decoded frames of shape (num_frames, height, width, 3).
+        ...     '''
+        ...     frames = []
+        ...     container.seek(0)
+        ...     start_index = indices[0]
+        ...     end_index = indices[-1]
+        ...     for i, frame in enumerate(container.decode(video=0)):
+        ...         if i > end_index:
+        ...             break
+        ...         if i >= start_index and i in indices:
+        ...             frames.append(frame)
+        ...     return np.stack([x.to_ndarray(format="rgb24") for x in frames])
+
+
+        >>> def sample_frame_indices(clip_len, frame_sample_rate, seg_len):
+        ...     '''
+        ...     Sample a given number of frame indices from the video.
+        ...     Args:
+        ...         clip_len (`int`): Total number of frames to sample.
+        ...         frame_sample_rate (`int`): Sample every n-th frame.
+        ...         seg_len (`int`): Maximum allowed index of sample's last frame.
+        ...     Returns:
+        ...         indices (`list[int]`): List of sampled frame indices
+        ...     '''
+        ...     converted_len = int(clip_len * frame_sample_rate)
+        ...     end_idx = np.random.randint(converted_len, seg_len)
+        ...     start_idx = end_idx - converted_len
+        ...     indices = np.linspace(start_idx, end_idx, num=clip_len)
+        ...     indices = np.clip(indices, start_idx, end_idx - 1).astype(np.int64)
+        ...     return indices
+
+
+        >>> # video clip consists of 300 frames (10 seconds at 30 FPS)
+        >>> file_path = hf_hub_download(
+        ...     repo_id="nielsr/video-demo", filename="eating_spaghetti.mp4", repo_type="dataset"
+        ... )
+        >>> container = av.open(file_path)
+
+        >>> # sample 32 frames
+        >>> indices = sample_frame_indices(clip_len=32, frame_sample_rate=1, seg_len=container.streams.video[0].frames)
+        >>> video = read_video_pyav(container=container, indices=indices)
+
+        >>> image_processor = VivitImageProcessor.from_pretrained("google/vivit-b-16x2-kinetics400")
+        >>> model = VivitModel.from_pretrained("google/vivit-b-16x2-kinetics400")
+
+        >>> # prepare video for the model
+        >>> inputs = image_processor(list(video), return_tensors="pt")
+
+        >>> # forward pass
+        >>> outputs = model(**inputs)
+        >>> last_hidden_states = outputs.last_hidden_state
+        >>> list(last_hidden_states.shape)
+        [1, 3137, 768]
+        ```"""
+
+        embedding_output = self.embeddings(pixel_values, interpolate_pos_encoding=interpolate_pos_encoding)
+        attention_mask = create_bidirectional_mask(
+            config=self.config,
+            inputs_embeds=embedding_output,
+            attention_mask=attention_mask,
+        )
+        hidden_states = embedding_output
+        for layer in self.layers:
+            hidden_states = layer(hidden_states, attention_mask, **kwargs)
+        sequence_output = self.layernorm(hidden_states)
+        pooled_output = self.pooler(sequence_output) if self.pooler is not None else None
+
+        return BaseModelOutputWithPooling(last_hidden_state=sequence_output, pooler_output=pooled_output)
+
+
+@auto_docstring(
+    custom_intro="""
+        ViViT Transformer model with a video classification head on top (a linear layer on top of the final hidden state of the
+    [CLS] token) e.g. for Kinetics-400.
+
+        <Tip>
+
+            Note that it's possible to fine-tune ViT on higher resolution images than the ones it has been trained on, by
+            setting `interpolate_pos_encoding` to `True` in the forward of the model. This will interpolate the pre-trained
+            position embeddings to the higher resolution.
+
+        </Tip>
+    """
+)
+class VivitForVideoClassification(VivitPreTrainedModel):
+    def __init__(self, config: VivitConfig):
+        super().__init__(config)
+
+        self.num_labels = config.num_labels
+        self.vivit = VivitModel(config, add_pooling_layer=False)
+
+        # Classifier head
+        self.classifier = nn.Linear(config.hidden_size, config.num_labels) if config.num_labels > 0 else nn.Identity()
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    @can_return_tuple
+    @auto_docstring
+    def forward(
+        self,
+        pixel_values: torch.FloatTensor | None = None,
+        labels: torch.LongTensor | None = None,
+        interpolate_pos_encoding: bool = False,
+        **kwargs: Unpack[TransformersKwargs],
+    ) -> ImageClassifierOutput:
+        r"""
+        labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
+            Labels for computing the image classification/regression loss. Indices should be in `[0, ...,
+            config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If
+            `config.num_labels > 1` a classification loss is computed (Cross-Entropy).
+
+        Examples:
+
+        ```python
+        >>> import av
+        >>> import numpy as np
+        >>> import torch
+
+        >>> from transformers import VivitImageProcessor, VivitForVideoClassification
+        >>> from huggingface_hub import hf_hub_download
+
+        >>> np.random.seed(0)
+
+
+        >>> def read_video_pyav(container, indices):
+        ...     '''
+        ...     Decode the video with PyAV decoder.
+        ...     Args:
+        ...         container (`av.container.input.InputContainer`): PyAV container.
+        ...         indices (`list[int]`): List of frame indices to decode.
+        ...     Returns:
+        ...         result (np.ndarray): np array of decoded frames of shape (num_frames, height, width, 3).
+        ...     '''
+        ...     frames = []
+        ...     container.seek(0)
+        ...     start_index = indices[0]
+        ...     end_index = indices[-1]
+        ...     for i, frame in enumerate(container.decode(video=0)):
+        ...         if i > end_index:
+        ...             break
+        ...         if i >= start_index and i in indices:
+        ...             frames.append(frame)
+        ...     return np.stack([x.to_ndarray(format="rgb24") for x in frames])
+
+
+        >>> def sample_frame_indices(clip_len, frame_sample_rate, seg_len):
+        ...     '''
+        ...     Sample a given number of frame indices from the video.
+        ...     Args:
+        ...         clip_len (`int`): Total number of frames to sample.
+        ...         frame_sample_rate (`int`): Sample every n-th frame.
+        ...         seg_len (`int`): Maximum allowed index of sample's last frame.
+        ...     Returns:
+        ...         indices (`list[int]`): List of sampled frame indices
+        ...     '''
+        ...     converted_len = int(clip_len * frame_sample_rate)
+        ...     end_idx = np.random.randint(converted_len, seg_len)
+        ...     start_idx = end_idx - converted_len
+        ...     indices = np.linspace(start_idx, end_idx, num=clip_len)
+        ...     indices = np.clip(indices, start_idx, end_idx - 1).astype(np.int64)
+        ...     return indices
+
+
+        >>> # video clip consists of 300 frames (10 seconds at 30 FPS)
+        >>> file_path = hf_hub_download(
+        ...     repo_id="nielsr/video-demo", filename="eating_spaghetti.mp4", repo_type="dataset"
+        ... )
+        >>> container = av.open(file_path)
+
+        >>> # sample 32 frames
+        >>> indices = sample_frame_indices(clip_len=32, frame_sample_rate=4, seg_len=container.streams.video[0].frames)
+        >>> video = read_video_pyav(container=container, indices=indices)
+
+        >>> image_processor = VivitImageProcessor.from_pretrained("google/vivit-b-16x2-kinetics400")
+        >>> model = VivitForVideoClassification.from_pretrained("google/vivit-b-16x2-kinetics400")
+
+        >>> inputs = image_processor(list(video), return_tensors="pt")
+
+        >>> with torch.no_grad():
+        ...     outputs = model(**inputs)
+        ...     logits = outputs.logits
+
+        >>> # model predicts one of the 400 Kinetics-400 classes
+        >>> predicted_label = logits.argmax(-1).item()
+        >>> print(model.config.id2label[predicted_label])
+        LABEL_116
+        ```"""
+
+        outputs: BaseModelOutput = self.vivit(
+            pixel_values, interpolate_pos_encoding=interpolate_pos_encoding, **kwargs
+        )
+        sequence_output = outputs.last_hidden_state
+        logits = self.classifier(sequence_output[:, 0, :])
+
+        loss = None
+        if labels is not None:
+            loss = self.loss_function(labels, logits, self.config, **kwargs)
+
+        return ImageClassifierOutput(
+            loss=loss,
+            logits=logits,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )
+
+
+__all__ = ["VivitModel", "VivitPreTrainedModel", "VivitForVideoClassification"]

LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/transformers/models/vivit/modular_vivit.py

@@ -0,0 +1,399 @@
+# Copyright 2023 Google AI and The HuggingFace Inc. team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""PyTorch ViViT model - modular file inheriting transformer core from ViT."""
+
+from collections.abc import Iterable
+
+import torch
+from torch import nn
+
+from ... import initialization as init
+from ...masking_utils import create_bidirectional_mask
+from ...modeling_outputs import BaseModelOutput, BaseModelOutputWithPooling, ImageClassifierOutput
+from ...processing_utils import Unpack
+from ...utils import TransformersKwargs, auto_docstring, logging
+from ...utils.generic import can_return_tuple, merge_with_config_defaults
+from ...utils.output_capturing import capture_outputs
+from ..vit.modeling_vit import (
+    PreTrainedModel,
+    ViTAttention,
+    ViTEmbeddings,
+    ViTLayer,
+    ViTMLP,
+    ViTModel,
+    ViTPooler,
+    ViTPreTrainedModel,
+)
+from .configuration_vivit import VivitConfig
+
+
+logger = logging.get_logger(__name__)
+
+
+class VivitTubeletEmbeddings(nn.Module):
+    """
+    This class turns `pixel_values` of shape `(batch_size, num_frames, num_channels, height, width)` into the initial
+    `hidden_states` (tubelet embeddings) of shape `(batch_size, seq_length, hidden_size)` to be consumed by a
+    Transformer encoder.
+
+    The seq_length equals (num_frames // tubelet_size[0]) * (height // tubelet_size[1]) * (width // tubelet_size[2]).
+    """
+
+    def __init__(self, config: VivitConfig):
+        super().__init__()
+        tubelet_size = config.tubelet_size
+        image_size = config.image_size
+        image_size = image_size if isinstance(image_size, Iterable) else (image_size, image_size)
+
+        self.num_patches = (
+            (config.num_frames // tubelet_size[0])
+            * (image_size[0] // tubelet_size[1])
+            * (image_size[1] // tubelet_size[2])
+        )
+        self.image_size = image_size
+        self.projection = nn.Conv3d(
+            config.num_channels, config.hidden_size, kernel_size=tubelet_size, stride=tubelet_size
+        )
+
+    def forward(self, pixel_values: torch.Tensor) -> torch.Tensor:
+        # transpose (batch_size, num_channels, num_frames, height, width) for Conv3d
+        pixel_values = pixel_values.transpose(1, 2)
+        return self.projection(pixel_values).flatten(2).transpose(1, 2)
+
+
+class VivitEmbeddings(ViTEmbeddings):
+    """
+    Construct the CLS token, position and tubelet patch embeddings for video input.
+    """
+
+    def __init__(self, config: VivitConfig):
+        super().__init__()
+
+        self.cls_token = nn.Parameter(torch.zeros(1, 1, config.hidden_size))
+        self.patch_embeddings = VivitTubeletEmbeddings(config)
+        num_patches = self.patch_embeddings.num_patches
+        self.position_embeddings = nn.Parameter(torch.zeros(1, num_patches + 1, config.hidden_size))
+        # patch_size is the spatial (height, width) part of the tubelet for pos encoding interpolation
+        self.patch_size = config.tubelet_size[1:]
+        del self.mask_token
+
+    def interpolate_pos_encoding(self, embeddings: torch.Tensor, height: int, width: int) -> torch.Tensor:
+        super().interpolate_pos_encoding(embeddings, height, width)
+        # patch_size is a 2-tuple (height, width) for the spatial tubelet dimensions
+        new_height = height // self.patch_size[0]  # noqa: F841
+        new_width = width // self.patch_size[1]  # noqa: F841
+
+    def forward(self, pixel_values: torch.Tensor, interpolate_pos_encoding: bool = False) -> torch.Tensor:
+        batch_size, num_frames, num_channels, height, width = pixel_values.shape
+        embeddings = self.patch_embeddings(pixel_values)
+
+        # add the [CLS] token to the embedded patch tokens
+        cls_tokens = self.cls_token.expand(batch_size, -1, -1)
+        embeddings = torch.cat((cls_tokens, embeddings), dim=1)
+
+        if interpolate_pos_encoding:
+            embeddings = embeddings + self.interpolate_pos_encoding(embeddings, height, width)
+        else:
+            if height != self.image_size[0] or width != self.image_size[1]:
+                raise ValueError(
+                    f"Input image size ({height}*{width}) doesn't match model"
+                    f" ({self.image_size[0]}*{self.image_size[1]})."
+                )
+            embeddings = embeddings + self.position_embeddings
+
+        embeddings = self.dropout(embeddings)
+
+        return embeddings
+
+
+class VivitAttention(ViTAttention):
+    pass
+
+
+class VivitMLP(ViTMLP):
+    pass
+
+
+class VivitLayer(ViTLayer):
+    pass
+
+
+class VivitPooler(ViTPooler):
+    pass
+
+
+@auto_docstring
+class VivitPreTrainedModel(ViTPreTrainedModel):
+    config: VivitConfig
+    base_model_prefix = "vivit"
+    input_modalities = ("video",)
+    _no_split_modules = ["VivitEmbeddings", "VivitLayer"]
+
+    @torch.no_grad()
+    def _init_weights(self, module):
+        """Initialize the weights"""
+        PreTrainedModel._init_weights(self, module)
+        if isinstance(module, VivitEmbeddings):
+            init.zeros_(module.cls_token)
+            init.zeros_(module.position_embeddings)
+
+
+@auto_docstring
+class VivitModel(ViTModel):
+    def __init__(self, config: VivitConfig, add_pooling_layer: bool = True):
+        r"""
+        add_pooling_layer (bool, *optional*, defaults to `True`):
+            Whether to add a pooling layer
+        """
+        super().__init__(config)
+        self.embeddings = VivitEmbeddings(config)
+
+    @merge_with_config_defaults
+    @capture_outputs(tie_last_hidden_states=False)
+    @auto_docstring
+    def forward(
+        self,
+        pixel_values: torch.FloatTensor | None = None,
+        interpolate_pos_encoding: bool = False,
+        attention_mask: torch.Tensor | None = None,
+        **kwargs: Unpack[TransformersKwargs],
+    ) -> BaseModelOutputWithPooling:
+        r"""
+        Examples:
+
+        ```python
+        >>> import av
+        >>> import numpy as np
+
+        >>> from transformers import VivitImageProcessor, VivitModel
+        >>> from huggingface_hub import hf_hub_download
+
+        >>> np.random.seed(0)
+
+
+        >>> def read_video_pyav(container, indices):
+        ...     '''
+        ...     Decode the video with PyAV decoder.
+        ...     Args:
+        ...         container (`av.container.input.InputContainer`): PyAV container.
+        ...         indices (`list[int]`): List of frame indices to decode.
+        ...     Returns:
+        ...         result (np.ndarray): np array of decoded frames of shape (num_frames, height, width, 3).
+        ...     '''
+        ...     frames = []
+        ...     container.seek(0)
+        ...     start_index = indices[0]
+        ...     end_index = indices[-1]
+        ...     for i, frame in enumerate(container.decode(video=0)):
+        ...         if i > end_index:
+        ...             break
+        ...         if i >= start_index and i in indices:
+        ...             frames.append(frame)
+        ...     return np.stack([x.to_ndarray(format="rgb24") for x in frames])
+
+
+        >>> def sample_frame_indices(clip_len, frame_sample_rate, seg_len):
+        ...     '''
+        ...     Sample a given number of frame indices from the video.
+        ...     Args:
+        ...         clip_len (`int`): Total number of frames to sample.
+        ...         frame_sample_rate (`int`): Sample every n-th frame.
+        ...         seg_len (`int`): Maximum allowed index of sample's last frame.
+        ...     Returns:
+        ...         indices (`list[int]`): List of sampled frame indices
+        ...     '''
+        ...     converted_len = int(clip_len * frame_sample_rate)
+        ...     end_idx = np.random.randint(converted_len, seg_len)
+        ...     start_idx = end_idx - converted_len
+        ...     indices = np.linspace(start_idx, end_idx, num=clip_len)
+        ...     indices = np.clip(indices, start_idx, end_idx - 1).astype(np.int64)
+        ...     return indices
+
+
+        >>> # video clip consists of 300 frames (10 seconds at 30 FPS)
+        >>> file_path = hf_hub_download(
+        ...     repo_id="nielsr/video-demo", filename="eating_spaghetti.mp4", repo_type="dataset"
+        ... )
+        >>> container = av.open(file_path)
+
+        >>> # sample 32 frames
+        >>> indices = sample_frame_indices(clip_len=32, frame_sample_rate=1, seg_len=container.streams.video[0].frames)
+        >>> video = read_video_pyav(container=container, indices=indices)
+
+        >>> image_processor = VivitImageProcessor.from_pretrained("google/vivit-b-16x2-kinetics400")
+        >>> model = VivitModel.from_pretrained("google/vivit-b-16x2-kinetics400")
+
+        >>> # prepare video for the model
+        >>> inputs = image_processor(list(video), return_tensors="pt")
+
+        >>> # forward pass
+        >>> outputs = model(**inputs)
+        >>> last_hidden_states = outputs.last_hidden_state
+        >>> list(last_hidden_states.shape)
+        [1, 3137, 768]
+        ```"""
+
+        embedding_output = self.embeddings(pixel_values, interpolate_pos_encoding=interpolate_pos_encoding)
+        attention_mask = create_bidirectional_mask(
+            config=self.config,
+            inputs_embeds=embedding_output,
+            attention_mask=attention_mask,
+        )
+        hidden_states = embedding_output
+        for layer in self.layers:
+            hidden_states = layer(hidden_states, attention_mask, **kwargs)
+        sequence_output = self.layernorm(hidden_states)
+        pooled_output = self.pooler(sequence_output) if self.pooler is not None else None
+
+        return BaseModelOutputWithPooling(last_hidden_state=sequence_output, pooler_output=pooled_output)
+
+
+@auto_docstring(
+    custom_intro="""
+        ViViT Transformer model with a video classification head on top (a linear layer on top of the final hidden state of the
+    [CLS] token) e.g. for Kinetics-400.
+
+        <Tip>
+
+            Note that it's possible to fine-tune ViT on higher resolution images than the ones it has been trained on, by
+            setting `interpolate_pos_encoding` to `True` in the forward of the model. This will interpolate the pre-trained
+            position embeddings to the higher resolution.
+
+        </Tip>
+    """
+)
+class VivitForVideoClassification(VivitPreTrainedModel):
+    def __init__(self, config: VivitConfig):
+        super().__init__(config)
+
+        self.num_labels = config.num_labels
+        self.vivit = VivitModel(config, add_pooling_layer=False)
+
+        # Classifier head
+        self.classifier = nn.Linear(config.hidden_size, config.num_labels) if config.num_labels > 0 else nn.Identity()
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    @can_return_tuple
+    @auto_docstring
+    def forward(
+        self,
+        pixel_values: torch.FloatTensor | None = None,
+        labels: torch.LongTensor | None = None,
+        interpolate_pos_encoding: bool = False,
+        **kwargs: Unpack[TransformersKwargs],
+    ) -> ImageClassifierOutput:
+        r"""
+        labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
+            Labels for computing the image classification/regression loss. Indices should be in `[0, ...,
+            config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If
+            `config.num_labels > 1` a classification loss is computed (Cross-Entropy).
+
+        Examples:
+
+        ```python
+        >>> import av
+        >>> import numpy as np
+        >>> import torch
+
+        >>> from transformers import VivitImageProcessor, VivitForVideoClassification
+        >>> from huggingface_hub import hf_hub_download
+
+        >>> np.random.seed(0)
+
+
+        >>> def read_video_pyav(container, indices):
+        ...     '''
+        ...     Decode the video with PyAV decoder.
+        ...     Args:
+        ...         container (`av.container.input.InputContainer`): PyAV container.
+        ...         indices (`list[int]`): List of frame indices to decode.
+        ...     Returns:
+        ...         result (np.ndarray): np array of decoded frames of shape (num_frames, height, width, 3).
+        ...     '''
+        ...     frames = []
+        ...     container.seek(0)
+        ...     start_index = indices[0]
+        ...     end_index = indices[-1]
+        ...     for i, frame in enumerate(container.decode(video=0)):
+        ...         if i > end_index:
+        ...             break
+        ...         if i >= start_index and i in indices:
+        ...             frames.append(frame)
+        ...     return np.stack([x.to_ndarray(format="rgb24") for x in frames])
+
+
+        >>> def sample_frame_indices(clip_len, frame_sample_rate, seg_len):
+        ...     '''
+        ...     Sample a given number of frame indices from the video.
+        ...     Args:
+        ...         clip_len (`int`): Total number of frames to sample.
+        ...         frame_sample_rate (`int`): Sample every n-th frame.
+        ...         seg_len (`int`): Maximum allowed index of sample's last frame.
+        ...     Returns:
+        ...         indices (`list[int]`): List of sampled frame indices
+        ...     '''
+        ...     converted_len = int(clip_len * frame_sample_rate)
+        ...     end_idx = np.random.randint(converted_len, seg_len)
+        ...     start_idx = end_idx - converted_len
+        ...     indices = np.linspace(start_idx, end_idx, num=clip_len)
+        ...     indices = np.clip(indices, start_idx, end_idx - 1).astype(np.int64)
+        ...     return indices
+
+
+        >>> # video clip consists of 300 frames (10 seconds at 30 FPS)
+        >>> file_path = hf_hub_download(
+        ...     repo_id="nielsr/video-demo", filename="eating_spaghetti.mp4", repo_type="dataset"
+        ... )
+        >>> container = av.open(file_path)
+
+        >>> # sample 32 frames
+        >>> indices = sample_frame_indices(clip_len=32, frame_sample_rate=4, seg_len=container.streams.video[0].frames)
+        >>> video = read_video_pyav(container=container, indices=indices)
+
+        >>> image_processor = VivitImageProcessor.from_pretrained("google/vivit-b-16x2-kinetics400")
+        >>> model = VivitForVideoClassification.from_pretrained("google/vivit-b-16x2-kinetics400")
+
+        >>> inputs = image_processor(list(video), return_tensors="pt")
+
+        >>> with torch.no_grad():
+        ...     outputs = model(**inputs)
+        ...     logits = outputs.logits
+
+        >>> # model predicts one of the 400 Kinetics-400 classes
+        >>> predicted_label = logits.argmax(-1).item()
+        >>> print(model.config.id2label[predicted_label])
+        LABEL_116
+        ```"""
+
+        outputs: BaseModelOutput = self.vivit(
+            pixel_values, interpolate_pos_encoding=interpolate_pos_encoding, **kwargs
+        )
+        sequence_output = outputs.last_hidden_state
+        logits = self.classifier(sequence_output[:, 0, :])
+
+        loss = None
+        if labels is not None:
+            loss = self.loss_function(labels, logits, self.config, **kwargs)
+
+        return ImageClassifierOutput(
+            loss=loss,
+            logits=logits,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )
+
+
+__all__ = ["VivitModel", "VivitPreTrainedModel", "VivitForVideoClassification"]

LTA_openwebtext_dualt/mini_owt_logdirichlet/logs/build_owt_t5_clean_cache_ngram32.log

@@ -0,0 +1,762 @@
+[stream cache] files=80 workers=80 pack_len=1023 append_eos=1
+[stream cache] cache=cache/owt_t5_stream_pack1023_clean_ngram32_appendeos1.pt rejected=cache/owt_t5_stream_pack1023_clean_ngram32_rejected.txt
+[worker 27] docs=10000 accepted_docs=8570 rejected_docs=1430 rows=7628 buffered=287
+[worker 64] docs=10000 accepted_docs=8546 rejected_docs=1454 rows=7668 buffered=401
+[worker 69] docs=10000 accepted_docs=8528 rejected_docs=1472 rows=7681 buffered=958
+[worker 47] docs=10000 accepted_docs=8479 rejected_docs=1521 rows=7543 buffered=499
+[worker 30] docs=10000 accepted_docs=8605 rejected_docs=1395 rows=7742 buffered=433
+[worker 42] docs=10000 accepted_docs=8527 rejected_docs=1473 rows=7650 buffered=102
+[worker 9] docs=10000 accepted_docs=8565 rejected_docs=1435 rows=7650 buffered=396
+[worker 61] docs=10000 accepted_docs=8590 rejected_docs=1410 rows=7540 buffered=291
+[worker 38] docs=10000 accepted_docs=8597 rejected_docs=1403 rows=7714 buffered=370
+[worker 65] docs=10000 accepted_docs=8514 rejected_docs=1486 rows=7691 buffered=756
+[worker 77] docs=10000 accepted_docs=8567 rejected_docs=1433 rows=7649 buffered=537
+[worker 75] docs=10000 accepted_docs=8597 rejected_docs=1403 rows=7652 buffered=759
+[worker 8] docs=10000 accepted_docs=8567 rejected_docs=1433 rows=7714 buffered=914
+[worker 17] docs=10000 accepted_docs=8548 rejected_docs=1452 rows=7779 buffered=427
+[worker 57] docs=10000 accepted_docs=8501 rejected_docs=1499 rows=7536 buffered=12
+[worker 46] docs=10000 accepted_docs=8585 rejected_docs=1415 rows=7727 buffered=214
+[worker 78] docs=10000 accepted_docs=8539 rejected_docs=1461 rows=7640 buffered=1007
+[worker 60] docs=10000 accepted_docs=8489 rejected_docs=1511 rows=7605 buffered=723
+[worker 36] docs=10000 accepted_docs=8511 rejected_docs=1489 rows=7540 buffered=772
+[worker 68] docs=10000 accepted_docs=8492 rejected_docs=1508 rows=7594 buffered=297
+[worker 70] docs=10000 accepted_docs=8536 rejected_docs=1464 rows=7555 buffered=384
+[worker 7] docs=10000 accepted_docs=8521 rejected_docs=1479 rows=7554 buffered=174
+[worker 34] docs=10000 accepted_docs=8558 rejected_docs=1442 rows=7707 buffered=476
+[worker 49] docs=10000 accepted_docs=8471 rejected_docs=1529 rows=7592 buffered=669
+[worker 55] docs=10000 accepted_docs=8542 rejected_docs=1458 rows=7712 buffered=476
+[worker 48] docs=10000 accepted_docs=8552 rejected_docs=1448 rows=7657 buffered=870
+[worker 18] docs=10000 accepted_docs=8514 rejected_docs=1486 rows=7611 buffered=910
+[worker 5] docs=10000 accepted_docs=8443 rejected_docs=1557 rows=7515 buffered=959
+[worker 56] docs=10000 accepted_docs=8544 rejected_docs=1456 rows=7570 buffered=653
+[worker 66] docs=10000 accepted_docs=8560 rejected_docs=1440 rows=7503 buffered=306
+[worker 43] docs=10000 accepted_docs=8518 rejected_docs=1482 rows=7735 buffered=36
+[worker 74] docs=10000 accepted_docs=8538 rejected_docs=1462 rows=7599 buffered=549
+[worker 4] docs=10000 accepted_docs=8611 rejected_docs=1389 rows=7734 buffered=461
+[worker 76] docs=10000 accepted_docs=8516 rejected_docs=1484 rows=7506 buffered=917
+[worker 35] docs=10000 accepted_docs=8529 rejected_docs=1471 rows=7502 buffered=993
+[worker 44] docs=10000 accepted_docs=8569 rejected_docs=1431 rows=7778 buffered=358
+[worker 73] docs=10000 accepted_docs=8553 rejected_docs=1447 rows=7831 buffered=927
+[worker 63] docs=10000 accepted_docs=8579 rejected_docs=1421 rows=7643 buffered=948
+[worker 79] docs=10000 accepted_docs=8502 rejected_docs=1498 rows=7630 buffered=186
+[worker 71] docs=10000 accepted_docs=8501 rejected_docs=1499 rows=7589 buffered=130
+[worker 62] docs=10000 accepted_docs=8483 rejected_docs=1517 rows=7663 buffered=611
+[worker 37] docs=10000 accepted_docs=8542 rejected_docs=1458 rows=7679 buffered=637
+[worker 39] docs=10000 accepted_docs=8570 rejected_docs=1430 rows=7682 buffered=77
+[worker 10] docs=10000 accepted_docs=8584 rejected_docs=1416 rows=7701 buffered=360
+[worker 11] docs=10000 accepted_docs=8521 rejected_docs=1479 rows=7666 buffered=660
+[worker 20] docs=10000 accepted_docs=8496 rejected_docs=1504 rows=7705 buffered=690
+[worker 19] docs=10000 accepted_docs=8531 rejected_docs=1469 rows=7478 buffered=485
+[worker 25] docs=10000 accepted_docs=8524 rejected_docs=1476 rows=7601 buffered=834
+[worker 22] docs=10000 accepted_docs=8543 rejected_docs=1457 rows=7663 buffered=238
+[worker 6] docs=10000 accepted_docs=8568 rejected_docs=1432 rows=7716 buffered=794
+[worker 26] docs=10000 accepted_docs=8527 rejected_docs=1473 rows=7652 buffered=672
+[worker 1] docs=10000 accepted_docs=8446 rejected_docs=1554 rows=7645 buffered=535
+[worker 32] docs=10000 accepted_docs=8484 rejected_docs=1516 rows=7571 buffered=327
+[worker 54] docs=10000 accepted_docs=8544 rejected_docs=1456 rows=7678 buffered=283
+[worker 21] docs=10000 accepted_docs=8483 rejected_docs=1517 rows=7764 buffered=633
+[worker 0] docs=10000 accepted_docs=8580 rejected_docs=1420 rows=7594 buffered=914
+[worker 72] docs=10000 accepted_docs=8552 rejected_docs=1448 rows=7606 buffered=1015
+[worker 51] docs=10000 accepted_docs=8503 rejected_docs=1497 rows=7506 buffered=878
+[worker 15] docs=10000 accepted_docs=8519 rejected_docs=1481 rows=7783 buffered=274
+[worker 3] docs=10000 accepted_docs=8508 rejected_docs=1492 rows=7637 buffered=957
+[worker 2] docs=10000 accepted_docs=8548 rejected_docs=1452 rows=7651 buffered=699
+[worker 28] docs=10000 accepted_docs=8486 rejected_docs=1514 rows=7663 buffered=487
+[worker 12] docs=10000 accepted_docs=8494 rejected_docs=1506 rows=7595 buffered=274
+[worker 33] docs=10000 accepted_docs=8516 rejected_docs=1484 rows=7692 buffered=697
+[worker 29] docs=10000 accepted_docs=8499 rejected_docs=1501 rows=7641 buffered=41
+[worker 41] docs=10000 accepted_docs=8555 rejected_docs=1445 rows=7733 buffered=555
+[worker 16] docs=10000 accepted_docs=8555 rejected_docs=1445 rows=7910 buffered=654
+[worker 67] docs=10000 accepted_docs=8505 rejected_docs=1495 rows=7595 buffered=261
+[worker 27] docs=20000 accepted_docs=17100 rejected_docs=2900 rows=15346 buffered=168
+[worker 47] docs=20000 accepted_docs=17059 rejected_docs=2941 rows=15156 buffered=420
+[worker 64] docs=20000 accepted_docs=17059 rejected_docs=2941 rows=15299 buffered=253
+[worker 30] docs=20000 accepted_docs=17117 rejected_docs=2883 rows=15356 buffered=840
+[worker 7] docs=20000 accepted_docs=17110 rejected_docs=2890 rows=15322 buffered=333
+[worker 14] docs=20000 accepted_docs=17079 rejected_docs=2921 rows=15222 buffered=288
+[worker 9] docs=20000 accepted_docs=17157 rejected_docs=2843 rows=15381 buffered=945
+[worker 36] docs=20000 accepted_docs=17099 rejected_docs=2901 rows=15131 buffered=537
+[worker 49] docs=20000 accepted_docs=17005 rejected_docs=2995 rows=15191 buffered=893
+[worker 13] docs=20000 accepted_docs=17122 rejected_docs=2878 rows=15486 buffered=108
+[worker 53] docs=20000 accepted_docs=17124 rejected_docs=2876 rows=15366 buffered=211
+[worker 42] docs=20000 accepted_docs=17065 rejected_docs=2935 rows=15291 buffered=946
+[worker 62] docs=20000 accepted_docs=17012 rejected_docs=2988 rows=15357 buffered=501
+[worker 48] docs=20000 accepted_docs=17107 rejected_docs=2893 rows=15202 buffered=777
+[worker 5] docs=20000 accepted_docs=17003 rejected_docs=2997 rows=15156 buffered=377
+[worker 68] docs=20000 accepted_docs=16998 rejected_docs=3002 rows=15180 buffered=320
+[worker 70] docs=20000 accepted_docs=17101 rejected_docs=2899 rows=15099 buffered=920
+[worker 59] docs=20000 accepted_docs=17038 rejected_docs=2962 rows=15133 buffered=207
+[worker 63] docs=20000 accepted_docs=17124 rejected_docs=2876 rows=15257 buffered=185
+[worker 45] docs=20000 accepted_docs=17079 rejected_docs=2921 rows=15240 buffered=65
+[worker 18] docs=20000 accepted_docs=17006 rejected_docs=2994 rows=15341 buffered=765
+[worker 56] docs=20000 accepted_docs=17040 rejected_docs=2960 rows=15295 buffered=80
+[worker 20] docs=20000 accepted_docs=17084 rejected_docs=2916 rows=15223 buffered=790
+[worker 61] docs=20000 accepted_docs=17066 rejected_docs=2934 rows=15077 buffered=351
+[worker 11] docs=20000 accepted_docs=17046 rejected_docs=2954 rows=15289 buffered=479
+[worker 66] docs=20000 accepted_docs=17017 rejected_docs=2983 rows=15075 buffered=860
+[worker 17] docs=20000 accepted_docs=17055 rejected_docs=2945 rows=15557 buffered=408
+[worker 73] docs=20000 accepted_docs=17099 rejected_docs=2901 rows=15421 buffered=376
+[worker 35] docs=20000 accepted_docs=16970 rejected_docs=3030 rows=14937 buffered=576
+[worker 10] docs=20000 accepted_docs=17070 rejected_docs=2930 rows=15353 buffered=816
+[worker 65] docs=20000 accepted_docs=16985 rejected_docs=3015 rows=15213 buffered=210
+[worker 4] docs=20000 accepted_docs=17093 rejected_docs=2907 rows=15354 buffered=870
+[worker 77] docs=20000 accepted_docs=17077 rejected_docs=2923 rows=15361 buffered=850
+[worker 12] docs=20000 accepted_docs=17051 rejected_docs=2949 rows=15074 buffered=684
+[worker 55] docs=20000 accepted_docs=17080 rejected_docs=2920 rows=15381 buffered=147
+[worker 21] docs=20000 accepted_docs=17000 rejected_docs=3000 rows=15382 buffered=240
+[worker 76] docs=20000 accepted_docs=17022 rejected_docs=2978 rows=15172 buffered=836
+[worker 22] docs=20000 accepted_docs=17064 rejected_docs=2936 rows=15333 buffered=152
+[worker 24] docs=20000 accepted_docs=17030 rejected_docs=2970 rows=15296 buffered=824
+[worker 6] docs=20000 accepted_docs=17099 rejected_docs=2901 rows=15437 buffered=805
+[worker 44] docs=20000 accepted_docs=17109 rejected_docs=2891 rows=15499 buffered=963
+[worker 74] docs=20000 accepted_docs=17070 rejected_docs=2930 rows=15145 buffered=928
+[worker 29] docs=20000 accepted_docs=17045 rejected_docs=2955 rows=15236 buffered=552
+[worker 1] docs=20000 accepted_docs=16931 rejected_docs=3069 rows=15332 buffered=922
+[worker 78] docs=20000 accepted_docs=17028 rejected_docs=2972 rows=15363 buffered=412
+[worker 75] docs=20000 accepted_docs=17099 rejected_docs=2901 rows=15276 buffered=602
+[worker 31] docs=20000 accepted_docs=17100 rejected_docs=2900 rows=15295 buffered=332
+[worker 26] docs=20000 accepted_docs=17090 rejected_docs=2910 rows=15276 buffered=612
+[worker 15] docs=20000 accepted_docs=17074 rejected_docs=2926 rows=15372 buffered=579
+[worker 52] docs=20000 accepted_docs=16996 rejected_docs=3004 rows=15188 buffered=862
+[worker 54] docs=20000 accepted_docs=17161 rejected_docs=2839 rows=15521 buffered=554
+[worker 8] docs=20000 accepted_docs=17022 rejected_docs=2978 rows=15372 buffered=198
+[worker 28] docs=20000 accepted_docs=17019 rejected_docs=2981 rows=15184 buffered=405
+[worker 50] docs=20000 accepted_docs=17137 rejected_docs=2863 rows=15452 buffered=179
+[worker 16] docs=20000 accepted_docs=17091 rejected_docs=2909 rows=15484 buffered=841
+[worker 19] docs=20000 accepted_docs=16993 rejected_docs=3007 rows=15082 buffered=43
+[worker 2] docs=20000 accepted_docs=17090 rejected_docs=2910 rows=15260 buffered=649
+[worker 40] docs=20000 accepted_docs=17032 rejected_docs=2968 rows=15333 buffered=384
+[worker 3] docs=20000 accepted_docs=17028 rejected_docs=2972 rows=15346 buffered=284
+[worker 32] docs=20000 accepted_docs=16963 rejected_docs=3037 rows=15281 buffered=734
+[worker 0] docs=20000 accepted_docs=17058 rejected_docs=2942 rows=15207 buffered=160
+[worker 23] docs=20000 accepted_docs=16953 rejected_docs=3047 rows=15291 buffered=301
+[worker 67] docs=20000 accepted_docs=17004 rejected_docs=2996 rows=15093 buffered=507
+[worker 51] docs=20000 accepted_docs=17060 rejected_docs=2940 rows=15216 buffered=602
+[worker 58] docs=20000 accepted_docs=17121 rejected_docs=2879 rows=15308 buffered=373
+[worker 41] docs=20000 accepted_docs=17047 rejected_docs=2953 rows=15493 buffered=793
+[worker 71] docs=20000 accepted_docs=17011 rejected_docs=2989 rows=15140 buffered=869
+[worker 72] docs=20000 accepted_docs=17013 rejected_docs=2987 rows=15273 buffered=742
+[worker 27] docs=30000 accepted_docs=25633 rejected_docs=4367 rows=22998 buffered=367
+[worker 14] docs=30000 accepted_docs=25645 rejected_docs=4355 rows=22893 buffered=769
+[worker 69] docs=30000 accepted_docs=25596 rejected_docs=4404 rows=23033 buffered=17
+[worker 36] docs=30000 accepted_docs=25569 rejected_docs=4431 rows=22727 buffered=459
+[worker 47] docs=30000 accepted_docs=25588 rejected_docs=4412 rows=22869 buffered=684
+[worker 60] docs=30000 accepted_docs=25653 rejected_docs=4347 rows=22857 buffered=320
+[worker 9] docs=30000 accepted_docs=25727 rejected_docs=4273 rows=23193 buffered=254
+[worker 42] docs=30000 accepted_docs=25649 rejected_docs=4351 rows=22938 buffered=920
+[worker 30] docs=30000 accepted_docs=25614 rejected_docs=4386 rows=22988 buffered=891
+[worker 46] docs=30000 accepted_docs=25623 rejected_docs=4377 rows=22955 buffered=865
+[worker 7] docs=30000 accepted_docs=25611 rejected_docs=4389 rows=22912 buffered=764
+[worker 64] docs=30000 accepted_docs=25616 rejected_docs=4384 rows=23137 buffered=29
+[worker 63] docs=30000 accepted_docs=25686 rejected_docs=4314 rows=22802 buffered=537
+[worker 53] docs=30000 accepted_docs=25612 rejected_docs=4388 rows=22959 buffered=773
+[worker 34] docs=30000 accepted_docs=25621 rejected_docs=4379 rows=22972 buffered=959
+[worker 57] docs=30000 accepted_docs=25548 rejected_docs=4452 rows=22776 buffered=183
+[worker 62] docs=30000 accepted_docs=25578 rejected_docs=4422 rows=23050 buffered=913
+[worker 49] docs=30000 accepted_docs=25539 rejected_docs=4461 rows=22815 buffered=903
+[worker 43] docs=30000 accepted_docs=25570 rejected_docs=4430 rows=23032 buffered=0
+[worker 20] docs=30000 accepted_docs=25656 rejected_docs=4344 rows=22856 buffered=898
+[worker 25] docs=30000 accepted_docs=25566 rejected_docs=4434 rows=22888 buffered=616
+[worker 5] docs=30000 accepted_docs=25528 rejected_docs=4472 rows=22827 buffered=301
+[worker 59] docs=30000 accepted_docs=25619 rejected_docs=4381 rows=22792 buffered=780
+[worker 24] docs=30000 accepted_docs=25546 rejected_docs=4454 rows=22897 buffered=444
+[worker 73] docs=30000 accepted_docs=25639 rejected_docs=4361 rows=23036 buffered=732
+[worker 75] docs=30000 accepted_docs=25659 rejected_docs=4341 rows=22901 buffered=118
+[worker 11] docs=30000 accepted_docs=25527 rejected_docs=4473 rows=22798 buffered=448
+[worker 18] docs=30000 accepted_docs=25521 rejected_docs=4479 rows=22931 buffered=816
+[worker 76] docs=30000 accepted_docs=25566 rejected_docs=4434 rows=22861 buffered=753
+[worker 66] docs=30000 accepted_docs=25534 rejected_docs=4466 rows=22750 buffered=290
+[worker 65] docs=30000 accepted_docs=25559 rejected_docs=4441 rows=22883 buffered=182
+[worker 38] docs=30000 accepted_docs=25688 rejected_docs=4312 rows=23050 buffered=918
+[worker 70] docs=30000 accepted_docs=25622 rejected_docs=4378 rows=22791 buffered=936
+[worker 78] docs=30000 accepted_docs=25582 rejected_docs=4418 rows=22967 buffered=988
+[worker 4] docs=30000 accepted_docs=25638 rejected_docs=4362 rows=23066 buffered=965
+[worker 12] docs=30000 accepted_docs=25549 rejected_docs=4451 rows=22658 buffered=817
+[worker 74] docs=30000 accepted_docs=25617 rejected_docs=4383 rows=22697 buffered=1018
+[worker 79] docs=30000 accepted_docs=25607 rejected_docs=4393 rows=22910 buffered=398
+[worker 19] docs=30000 accepted_docs=25583 rejected_docs=4417 rows=22764 buffered=650
+[worker 21] docs=30000 accepted_docs=25498 rejected_docs=4502 rows=23009 buffered=668
+[worker 55] docs=30000 accepted_docs=25609 rejected_docs=4391 rows=23035 buffered=458
+[worker 35] docs=30000 accepted_docs=25506 rejected_docs=4494 rows=22599 buffered=934
+[worker 37] docs=30000 accepted_docs=25595 rejected_docs=4405 rows=22953 buffered=257
+[worker 15] docs=30000 accepted_docs=25638 rejected_docs=4362 rows=23094 buffered=415
+[worker 16] docs=30000 accepted_docs=25702 rejected_docs=4298 rows=23293 buffered=582
+[worker 6] docs=30000 accepted_docs=25638 rejected_docs=4362 rows=23019 buffered=890
+[worker 22] docs=30000 accepted_docs=25621 rejected_docs=4379 rows=23015 buffered=526
+[worker 8] docs=30000 accepted_docs=25617 rejected_docs=4383 rows=23088 buffered=187
+[worker 28] docs=30000 accepted_docs=25541 rejected_docs=4459 rows=22740 buffered=363
+[worker 29] docs=30000 accepted_docs=25567 rejected_docs=4433 rows=22926 buffered=642
+[worker 31] docs=30000 accepted_docs=25562 rejected_docs=4438 rows=22787 buffered=830
+[worker 26] docs=30000 accepted_docs=25672 rejected_docs=4328 rows=23054 buffered=863
+[worker 40] docs=30000 accepted_docs=25650 rejected_docs=4350 rows=23030 buffered=209
+[worker 1] docs=30000 accepted_docs=25419 rejected_docs=4581 rows=22944 buffered=885
+[worker 39] docs=30000 accepted_docs=25611 rejected_docs=4389 rows=22994 buffered=770
+[worker 10] docs=30000 accepted_docs=25575 rejected_docs=4425 rows=23047 buffered=706
+[worker 71] docs=30000 accepted_docs=25548 rejected_docs=4452 rows=22793 buffered=111
+[worker 50] docs=30000 accepted_docs=25656 rejected_docs=4344 rows=23012 buffered=1009
+[worker 3] docs=30000 accepted_docs=25553 rejected_docs=4447 rows=22968 buffered=652
+[worker 51] docs=30000 accepted_docs=25576 rejected_docs=4424 rows=22913 buffered=240
+[worker 67] docs=30000 accepted_docs=25532 rejected_docs=4468 rows=22763 buffered=738
+[worker 58] docs=30000 accepted_docs=25640 rejected_docs=4360 rows=22930 buffered=509
+[worker 72] docs=30000 accepted_docs=25558 rejected_docs=4442 rows=22995 buffered=864
+[worker 0] docs=30000 accepted_docs=25609 rejected_docs=4391 rows=22918 buffered=70
+[worker 27] docs=40000 accepted_docs=34218 rejected_docs=5782 rows=30691 buffered=886
+[worker 69] docs=40000 accepted_docs=34090 rejected_docs=5910 rows=30595 buffered=470
+[worker 60] docs=40000 accepted_docs=34150 rejected_docs=5850 rows=30450 buffered=368
+[worker 36] docs=40000 accepted_docs=34124 rejected_docs=5876 rows=30325 buffered=325
+[worker 14] docs=40000 accepted_docs=34150 rejected_docs=5850 rows=30492 buffered=706
+[worker 46] docs=40000 accepted_docs=34140 rejected_docs=5860 rows=30697 buffered=400
+[worker 9] docs=40000 accepted_docs=34223 rejected_docs=5777 rows=30784 buffered=464
+[worker 48] docs=40000 accepted_docs=34230 rejected_docs=5770 rows=30647 buffered=821
+[worker 56] docs=40000 accepted_docs=34090 rejected_docs=5910 rows=30636 buffered=1000
+[worker 49] docs=40000 accepted_docs=34130 rejected_docs=5870 rows=30554 buffered=931
+[worker 57] docs=40000 accepted_docs=34079 rejected_docs=5921 rows=30413 buffered=365
+[worker 62] docs=40000 accepted_docs=34110 rejected_docs=5890 rows=30787 buffered=16
+[worker 5] docs=40000 accepted_docs=34035 rejected_docs=5965 rows=30461 buffered=125
+[worker 43] docs=40000 accepted_docs=34110 rejected_docs=5890 rows=30749 buffered=76
+[worker 34] docs=40000 accepted_docs=34105 rejected_docs=5895 rows=30532 buffered=657
+[worker 30] docs=40000 accepted_docs=34136 rejected_docs=5864 rows=30510 buffered=757
+[worker 24] docs=40000 accepted_docs=34120 rejected_docs=5880 rows=30615 buffered=779
+[worker 45] docs=40000 accepted_docs=34147 rejected_docs=5853 rows=30465 buffered=153
+[worker 18] docs=40000 accepted_docs=34064 rejected_docs=5936 rows=30506 buffered=105
+[worker 63] docs=40000 accepted_docs=34221 rejected_docs=5779 rows=30447 buffered=869
+[worker 17] docs=40000 accepted_docs=34071 rejected_docs=5929 rows=30780 buffered=236
+[worker 42] docs=40000 accepted_docs=34202 rejected_docs=5798 rows=30608 buffered=530
+[worker 13] docs=40000 accepted_docs=34235 rejected_docs=5765 rows=30747 buffered=531
+[worker 53] docs=40000 accepted_docs=34180 rejected_docs=5820 rows=30678 buffered=47
+[worker 68] docs=40000 accepted_docs=34057 rejected_docs=5943 rows=30515 buffered=20
+[worker 25] docs=40000 accepted_docs=34016 rejected_docs=5984 rows=30483 buffered=24
+[worker 7] docs=40000 accepted_docs=34102 rejected_docs=5898 rows=30574 buffered=419
+[worker 66] docs=40000 accepted_docs=34049 rejected_docs=5951 rows=30283 buffered=340
+[worker 74] docs=40000 accepted_docs=34153 rejected_docs=5847 rows=30350 buffered=714
+[worker 73] docs=40000 accepted_docs=34181 rejected_docs=5819 rows=30559 buffered=395
+[worker 64] docs=40000 accepted_docs=34191 rejected_docs=5809 rows=30913 buffered=362
+[worker 4] docs=40000 accepted_docs=34247 rejected_docs=5753 rows=30629 buffered=735
+[worker 78] docs=40000 accepted_docs=34065 rejected_docs=5935 rows=30598 buffered=161
+[worker 20] docs=40000 accepted_docs=34177 rejected_docs=5823 rows=30478 buffered=945
+[worker 12] docs=40000 accepted_docs=34061 rejected_docs=5939 rows=30276 buffered=246
+[worker 19] docs=40000 accepted_docs=34176 rejected_docs=5824 rows=30343 buffered=716
+[worker 65] docs=40000 accepted_docs=34122 rejected_docs=5878 rows=30457 buffered=911
+[worker 61] docs=40000 accepted_docs=34059 rejected_docs=5941 rows=30249 buffered=708
+[worker 11] docs=40000 accepted_docs=34095 rejected_docs=5905 rows=30502 buffered=650
+[worker 79] docs=40000 accepted_docs=34121 rejected_docs=5879 rows=30534 buffered=162
+[worker 21] docs=40000 accepted_docs=34055 rejected_docs=5945 rows=30703 buffered=98
+[worker 76] docs=40000 accepted_docs=34158 rejected_docs=5842 rows=30608 buffered=915
+[worker 33] docs=40000 accepted_docs=34119 rejected_docs=5881 rows=30582 buffered=552
+[worker 16] docs=40000 accepted_docs=34228 rejected_docs=5772 rows=30957 buffered=837
+[worker 6] docs=40000 accepted_docs=34135 rejected_docs=5865 rows=30657 buffered=330
+[worker 10] docs=40000 accepted_docs=34045 rejected_docs=5955 rows=30600 buffered=615
+[worker 28] docs=40000 accepted_docs=34082 rejected_docs=5918 rows=30412 buffered=218
+[worker 71] docs=40000 accepted_docs=34107 rejected_docs=5893 rows=30381 buffered=604
+[worker 40] docs=40000 accepted_docs=34231 rejected_docs=5769 rows=30621 buffered=943
+[worker 22] docs=40000 accepted_docs=34194 rejected_docs=5806 rows=30776 buffered=1007
+[worker 37] docs=40000 accepted_docs=34117 rejected_docs=5883 rows=30681 buffered=1021
+[worker 26] docs=40000 accepted_docs=34208 rejected_docs=5792 rows=30718 buffered=62
+[worker 44] docs=40000 accepted_docs=34277 rejected_docs=5723 rows=30946 buffered=193
+[worker 31] docs=40000 accepted_docs=34134 rejected_docs=5866 rows=30545 buffered=1017
+[worker 32] docs=40000 accepted_docs=34111 rejected_docs=5889 rows=30600 buffered=556
+[worker 54] docs=40000 accepted_docs=34151 rejected_docs=5849 rows=30730 buffered=42
+[worker 50] docs=40000 accepted_docs=34208 rejected_docs=5792 rows=30727 buffered=760
+[worker 29] docs=40000 accepted_docs=34095 rejected_docs=5905 rows=30543 buffered=371
+[worker 39] docs=40000 accepted_docs=34187 rejected_docs=5813 rows=30720 buffered=787
+[worker 67] docs=40000 accepted_docs=34034 rejected_docs=5966 rows=30381 buffered=92
+[worker 52] docs=40000 accepted_docs=34152 rejected_docs=5848 rows=30588 buffered=180
+[worker 2] docs=40000 accepted_docs=34173 rejected_docs=5827 rows=30509 buffered=878
+[worker 3] docs=40000 accepted_docs=34102 rejected_docs=5898 rows=30616 buffered=567
+[worker 41] docs=40000 accepted_docs=34060 rejected_docs=5940 rows=30682 buffered=535
+[worker 23] docs=40000 accepted_docs=34004 rejected_docs=5996 rows=30679 buffered=438
+[worker 0] docs=40000 accepted_docs=34160 rejected_docs=5840 rows=30572 buffered=321
+[worker 58] docs=40000 accepted_docs=34237 rejected_docs=5763 rows=30638 buffered=797
+[worker 27] docs=50000 accepted_docs=42794 rejected_docs=7206 rows=38312 buffered=708
+[worker 69] docs=50000 accepted_docs=42666 rejected_docs=7334 rows=38263 buffered=530
+[worker 47] docs=50000 accepted_docs=42712 rejected_docs=7288 rows=38241 buffered=874
+[worker 56] docs=50000 accepted_docs=42644 rejected_docs=7356 rows=38214 buffered=86
+[worker 43] docs=50000 accepted_docs=42677 rejected_docs=7323 rows=38263 buffered=620
+[worker 48] docs=50000 accepted_docs=42795 rejected_docs=7205 rows=38239 buffered=259
+[worker 36] docs=50000 accepted_docs=42655 rejected_docs=7345 rows=37980 buffered=109
+[worker 49] docs=50000 accepted_docs=42648 rejected_docs=7352 rows=38233 buffered=348
+[worker 60] docs=50000 accepted_docs=42650 rejected_docs=7350 rows=38026 buffered=30
+[worker 34] docs=50000 accepted_docs=42664 rejected_docs=7336 rows=38109 buffered=69
+[worker 24] docs=50000 accepted_docs=42670 rejected_docs=7330 rows=38359 buffered=116
+[worker 5] docs=50000 accepted_docs=42527 rejected_docs=7473 rows=37992 buffered=550
+[worker 62] docs=50000 accepted_docs=42664 rejected_docs=7336 rows=38404 buffered=626
+[worker 7] docs=50000 accepted_docs=42654 rejected_docs=7346 rows=38268 buffered=299
+[worker 17] docs=50000 accepted_docs=42649 rejected_docs=7351 rows=38450 buffered=401
+[worker 78] docs=50000 accepted_docs=42650 rejected_docs=7350 rows=38301 buffered=941
+[worker 63] docs=50000 accepted_docs=42791 rejected_docs=7209 rows=38114 buffered=39
+[worker 65] docs=50000 accepted_docs=42653 rejected_docs=7347 rows=38176 buffered=998
+[worker 25] docs=50000 accepted_docs=42513 rejected_docs=7487 rows=38101 buffered=35
+[worker 57] docs=50000 accepted_docs=42642 rejected_docs=7358 rows=38189 buffered=396
+[worker 53] docs=50000 accepted_docs=42723 rejected_docs=7277 rows=38468 buffered=844
+[worker 18] docs=50000 accepted_docs=42616 rejected_docs=7384 rows=38143 buffered=266
+[worker 30] docs=50000 accepted_docs=42675 rejected_docs=7325 rows=38125 buffered=298
+[worker 12] docs=50000 accepted_docs=42591 rejected_docs=7409 rows=37983 buffered=338
+[worker 73] docs=50000 accepted_docs=42710 rejected_docs=7290 rows=38289 buffered=477
+[worker 59] docs=50000 accepted_docs=42740 rejected_docs=7260 rows=38100 buffered=91
+[worker 13] docs=50000 accepted_docs=42741 rejected_docs=7259 rows=38414 buffered=911
+[worker 61] docs=50000 accepted_docs=42583 rejected_docs=7417 rows=37818 buffered=902
+[worker 19] docs=50000 accepted_docs=42715 rejected_docs=7285 rows=37997 buffered=352
+[worker 11] docs=50000 accepted_docs=42649 rejected_docs=7351 rows=38238 buffered=479
+[worker 77] docs=50000 accepted_docs=42652 rejected_docs=7348 rows=38459 buffered=384
+[worker 79] docs=50000 accepted_docs=42648 rejected_docs=7352 rows=38206 buffered=910
+[worker 4] docs=50000 accepted_docs=42751 rejected_docs=7249 rows=38301 buffered=419
+[worker 55] docs=50000 accepted_docs=42559 rejected_docs=7441 rows=37997 buffered=385
+[worker 68] docs=50000 accepted_docs=42572 rejected_docs=7428 rows=38204 buffered=800
+[worker 70] docs=50000 accepted_docs=42728 rejected_docs=7272 rows=38121 buffered=96
+[worker 35] docs=50000 accepted_docs=42588 rejected_docs=7412 rows=37787 buffered=540
+[worker 21] docs=50000 accepted_docs=42576 rejected_docs=7424 rows=38378 buffered=924
+[worker 64] docs=50000 accepted_docs=42634 rejected_docs=7366 rows=38483 buffered=622
+[worker 38] docs=50000 accepted_docs=42802 rejected_docs=7198 rows=38306 buffered=946
+[worker 33] docs=50000 accepted_docs=42706 rejected_docs=7294 rows=38286 buffered=1003
+[worker 71] docs=50000 accepted_docs=42592 rejected_docs=7408 rows=37932 buffered=207
+[worker 6] docs=50000 accepted_docs=42669 rejected_docs=7331 rows=38267 buffered=649
+[worker 76] docs=50000 accepted_docs=42643 rejected_docs=7357 rows=38097 buffered=59
+[worker 16] docs=50000 accepted_docs=42779 rejected_docs=7221 rows=38657 buffered=349
+[worker 40] docs=50000 accepted_docs=42782 rejected_docs=7218 rows=38472 buffered=152
+[worker 31] docs=50000 accepted_docs=42591 rejected_docs=7409 rows=38178 buffered=96
+[worker 72] docs=50000 accepted_docs=42550 rejected_docs=7450 rows=38110 buffered=229
+[worker 15] docs=50000 accepted_docs=42618 rejected_docs=7382 rows=38306 buffered=10
+[worker 8] docs=50000 accepted_docs=42678 rejected_docs=7322 rows=38350 buffered=495
+[worker 32] docs=50000 accepted_docs=42651 rejected_docs=7349 rows=38312 buffered=867
+[worker 28] docs=50000 accepted_docs=42600 rejected_docs=7400 rows=38018 buffered=731
+[worker 10] docs=50000 accepted_docs=42608 rejected_docs=7392 rows=38376 buffered=428
+[worker 67] docs=50000 accepted_docs=42534 rejected_docs=7466 rows=38027 buffered=133
+[worker 51] docs=50000 accepted_docs=42634 rejected_docs=7366 rows=38377 buffered=819
+[worker 26] docs=50000 accepted_docs=42695 rejected_docs=7305 rows=38361 buffered=442
+[worker 29] docs=50000 accepted_docs=42654 rejected_docs=7346 rows=38207 buffered=308
+[worker 37] docs=50000 accepted_docs=42600 rejected_docs=7400 rows=38241 buffered=874
+[worker 44] docs=50000 accepted_docs=42818 rejected_docs=7182 rows=38550 buffered=105
+[worker 50] docs=50000 accepted_docs=42766 rejected_docs=7234 rows=38423 buffered=374
+[worker 1] docs=50000 accepted_docs=42499 rejected_docs=7501 rows=38332 buffered=224
+[worker 22] docs=50000 accepted_docs=42683 rejected_docs=7317 rows=38541 buffered=15
+[worker 39] docs=50000 accepted_docs=42735 rejected_docs=7265 rows=38404 buffered=593
+[worker 41] docs=50000 accepted_docs=42690 rejected_docs=7310 rows=38297 buffered=819
+[worker 52] docs=50000 accepted_docs=42648 rejected_docs=7352 rows=38173 buffered=551
+[worker 23] docs=50000 accepted_docs=42494 rejected_docs=7506 rows=38349 buffered=764
+[worker 2] docs=50000 accepted_docs=42700 rejected_docs=7300 rows=38148 buffered=117
+[worker 3] docs=50000 accepted_docs=42602 rejected_docs=7398 rows=38305 buffered=414
+[worker 0] docs=50000 accepted_docs=42715 rejected_docs=7285 rows=38342 buffered=450
+[worker 69] docs=60000 accepted_docs=51181 rejected_docs=8819 rows=45917 buffered=780
+[worker 47] docs=60000 accepted_docs=51287 rejected_docs=8713 rows=45860 buffered=714
+[worker 27] docs=60000 accepted_docs=51400 rejected_docs=8600 rows=45979 buffered=200
+[worker 46] docs=60000 accepted_docs=51202 rejected_docs=8798 rows=45965 buffered=352
+[worker 43] docs=60000 accepted_docs=51290 rejected_docs=8710 rows=46028 buffered=1013
+[worker 24] docs=60000 accepted_docs=51223 rejected_docs=8777 rows=45968 buffered=511
+[worker 60] docs=60000 accepted_docs=51168 rejected_docs=8832 rows=45617 buffered=852
+[worker 36] docs=60000 accepted_docs=51156 rejected_docs=8844 rows=45569 buffered=301
+[worker 49] docs=60000 accepted_docs=51226 rejected_docs=8774 rows=45920 buffered=344
+[worker 56] docs=60000 accepted_docs=51143 rejected_docs=8857 rows=45834 buffered=512
+[worker 34] docs=60000 accepted_docs=51170 rejected_docs=8830 rows=45789 buffered=744
+[worker 78] docs=60000 accepted_docs=51174 rejected_docs=8826 rows=45787 buffered=949
+[worker 66] docs=60000 accepted_docs=51179 rejected_docs=8821 rows=45456 buffered=309
+[worker 5] docs=60000 accepted_docs=50974 rejected_docs=9026 rows=45511 buffered=178
+[worker 9] docs=60000 accepted_docs=51299 rejected_docs=8701 rows=46084 buffered=68
+[worker 18] docs=60000 accepted_docs=51167 rejected_docs=8833 rows=45724 buffered=318
+[worker 65] docs=60000 accepted_docs=51239 rejected_docs=8761 rows=45981 buffered=754
+[worker 17] docs=60000 accepted_docs=51169 rejected_docs=8831 rows=46098 buffered=703
+[worker 62] docs=60000 accepted_docs=51233 rejected_docs=8767 rows=46077 buffered=451
+[worker 63] docs=60000 accepted_docs=51293 rejected_docs=8707 rows=45808 buffered=328
+[worker 12] docs=60000 accepted_docs=51221 rejected_docs=8779 rows=45768 buffered=270
+[worker 7] docs=60000 accepted_docs=51240 rejected_docs=8760 rows=45934 buffered=817
+[worker 73] docs=60000 accepted_docs=51269 rejected_docs=8731 rows=45947 buffered=419
+[worker 53] docs=60000 accepted_docs=51219 rejected_docs=8781 rows=46151 buffered=498
+[worker 74] docs=60000 accepted_docs=51157 rejected_docs=8843 rows=45677 buffered=138
+[worker 57] docs=60000 accepted_docs=51177 rejected_docs=8823 rows=45713 buffered=526
+[worker 61] docs=60000 accepted_docs=51123 rejected_docs=8877 rows=45505 buffered=894
+[worker 45] docs=60000 accepted_docs=51231 rejected_docs=8769 rows=45862 buffered=442
+[worker 42] docs=60000 accepted_docs=51257 rejected_docs=8743 rows=45837 buffered=381
+[worker 20] docs=60000 accepted_docs=51178 rejected_docs=8822 rows=45829 buffered=53
+[worker 59] docs=60000 accepted_docs=51299 rejected_docs=8701 rows=45780 buffered=941
+[worker 30] docs=60000 accepted_docs=51282 rejected_docs=8718 rows=45915 buffered=15
+[worker 11] docs=60000 accepted_docs=51136 rejected_docs=8864 rows=45689 buffered=536
+[worker 55] docs=60000 accepted_docs=51017 rejected_docs=8983 rows=45570 buffered=304
+[worker 79] docs=60000 accepted_docs=51158 rejected_docs=8842 rows=45890 buffered=94
+[worker 68] docs=60000 accepted_docs=51069 rejected_docs=8931 rows=45731 buffered=823
+[worker 21] docs=60000 accepted_docs=51116 rejected_docs=8884 rows=45986 buffered=501
+[worker 71] docs=60000 accepted_docs=51107 rejected_docs=8893 rows=45478 buffered=142
+[worker 70] docs=60000 accepted_docs=51250 rejected_docs=8750 rows=45754 buffered=874
+[worker 75] docs=60000 accepted_docs=51168 rejected_docs=8832 rows=45687 buffered=321
+[worker 33] docs=60000 accepted_docs=51257 rejected_docs=8743 rows=45810 buffered=147
+[worker 77] docs=60000 accepted_docs=51141 rejected_docs=8859 rows=46087 buffered=175
+[worker 31] docs=60000 accepted_docs=51160 rejected_docs=8840 rows=45669 buffered=444
+[worker 64] docs=60000 accepted_docs=51187 rejected_docs=8813 rows=46258 buffered=273
+[worker 19] docs=60000 accepted_docs=51244 rejected_docs=8756 rows=45632 buffered=597
+[worker 38] docs=60000 accepted_docs=51251 rejected_docs=8749 rows=45824 buffered=256
+[worker 6] docs=60000 accepted_docs=51186 rejected_docs=8814 rows=45958 buffered=247
+[worker 40] docs=60000 accepted_docs=51307 rejected_docs=8693 rows=46001 buffered=761
+[worker 35] docs=60000 accepted_docs=51128 rejected_docs=8872 rows=45345 buffered=694
+[worker 16] docs=60000 accepted_docs=51341 rejected_docs=8659 rows=46306 buffered=346
+[worker 37] docs=60000 accepted_docs=51180 rejected_docs=8820 rows=45852 buffered=65
+[worker 44] docs=60000 accepted_docs=51374 rejected_docs=8626 rows=46185 buffered=768
+[worker 72] docs=60000 accepted_docs=51094 rejected_docs=8906 rows=45786 buffered=684
+[worker 67] docs=60000 accepted_docs=51086 rejected_docs=8914 rows=45689 buffered=140
+[worker 15] docs=60000 accepted_docs=51126 rejected_docs=8874 rows=46014 buffered=724
+[worker 26] docs=60000 accepted_docs=51236 rejected_docs=8764 rows=46094 buffered=599
+[worker 8] docs=60000 accepted_docs=51173 rejected_docs=8827 rows=45969 buffered=56
+[worker 28] docs=60000 accepted_docs=51107 rejected_docs=8893 rows=45715 buffered=21
+[worker 29] docs=60000 accepted_docs=51158 rejected_docs=8842 rows=45804 buffered=275
+[worker 50] docs=60000 accepted_docs=51297 rejected_docs=8703 rows=46196 buffered=588
+[worker 54] docs=60000 accepted_docs=51177 rejected_docs=8823 rows=46091 buffered=968
+[worker 10] docs=60000 accepted_docs=51158 rejected_docs=8842 rows=46057 buffered=356
+[worker 51] docs=60000 accepted_docs=51088 rejected_docs=8912 rows=45969 buffered=716
+[worker 39] docs=60000 accepted_docs=51210 rejected_docs=8790 rows=45964 buffered=880
+[worker 23] docs=60000 accepted_docs=50993 rejected_docs=9007 rows=46002 buffered=250
+[worker 41] docs=60000 accepted_docs=51206 rejected_docs=8794 rows=45873 buffered=737
+[worker 1] docs=60000 accepted_docs=51064 rejected_docs=8936 rows=46146 buffered=112
+[worker 2] docs=60000 accepted_docs=51263 rejected_docs=8737 rows=45818 buffered=842
+[worker 52] docs=60000 accepted_docs=51087 rejected_docs=8913 rows=45710 buffered=864
+[worker 0] docs=60000 accepted_docs=51261 rejected_docs=8739 rows=46024 buffered=87
+[worker 3] docs=60000 accepted_docs=51106 rejected_docs=8894 rows=45866 buffered=8
+[worker 58] docs=60000 accepted_docs=51261 rejected_docs=8739 rows=45991 buffered=189
+[worker 47] docs=70000 accepted_docs=59814 rejected_docs=10186 rows=53460 buffered=249
+[worker 69] docs=70000 accepted_docs=59677 rejected_docs=10323 rows=53576 buffered=564
+[worker 27] docs=70000 accepted_docs=59940 rejected_docs=10060 rows=53693 buffered=533
+[worker 46] docs=70000 accepted_docs=59713 rejected_docs=10287 rows=53561 buffered=943
+[worker 24] docs=70000 accepted_docs=59779 rejected_docs=10221 rows=53541 buffered=364
+[worker 60] docs=70000 accepted_docs=59736 rejected_docs=10264 rows=53375 buffered=428
+[worker 43] docs=70000 accepted_docs=59791 rejected_docs=10209 rows=53686 buffered=977
+[worker 78] docs=70000 accepted_docs=59721 rejected_docs=10279 rows=53360 buffered=413
+[worker 14] docs=70000 accepted_docs=59762 rejected_docs=10238 rows=53724 buffered=942
+[worker 65] docs=70000 accepted_docs=59861 rejected_docs=10139 rows=53726 buffered=721
+[worker 49] docs=70000 accepted_docs=59807 rejected_docs=10193 rows=53548 buffered=606
+[worker 5] docs=70000 accepted_docs=59499 rejected_docs=10501 rows=53192 buffered=481
+[worker 36] docs=70000 accepted_docs=59610 rejected_docs=10390 rows=53137 buffered=1021
+[worker 34] docs=70000 accepted_docs=59736 rejected_docs=10264 rows=53460 buffered=222
+[worker 66] docs=70000 accepted_docs=59724 rejected_docs=10276 rows=53199 buffered=368
+[worker 13] docs=70000 accepted_docs=59838 rejected_docs=10162 rows=53674 buffered=486
+[worker 12] docs=70000 accepted_docs=59809 rejected_docs=10191 rows=53555 buffered=1010
+[worker 56] docs=70000 accepted_docs=59577 rejected_docs=10423 rows=53439 buffered=904
+[worker 9] docs=70000 accepted_docs=59877 rejected_docs=10123 rows=53736 buffered=730
+[worker 17] docs=70000 accepted_docs=59739 rejected_docs=10261 rows=53779 buffered=864
+[worker 61] docs=70000 accepted_docs=59671 rejected_docs=10329 rows=53140 buffered=423
+[worker 55] docs=70000 accepted_docs=59615 rejected_docs=10385 rows=53356 buffered=108
+[worker 42] docs=70000 accepted_docs=59738 rejected_docs=10262 rows=53499 buffered=655
+[worker 59] docs=70000 accepted_docs=59846 rejected_docs=10154 rows=53377 buffered=823
+[worker 63] docs=70000 accepted_docs=59781 rejected_docs=10219 rows=53409 buffered=644
+[worker 62] docs=70000 accepted_docs=59740 rejected_docs=10260 rows=53676 buffered=521
+[worker 68] docs=70000 accepted_docs=59573 rejected_docs=10427 rows=53278 buffered=626
+[worker 74] docs=70000 accepted_docs=59638 rejected_docs=10362 rows=53218 buffered=143
+[worker 57] docs=70000 accepted_docs=59760 rejected_docs=10240 rows=53496 buffered=515
+[worker 4] docs=70000 accepted_docs=59799 rejected_docs=10201 rows=53545 buffered=159
+[worker 71] docs=70000 accepted_docs=59644 rejected_docs=10356 rows=53050 buffered=334
+[worker 75] docs=70000 accepted_docs=59715 rejected_docs=10285 rows=53260 buffered=284
+[worker 11] docs=70000 accepted_docs=59647 rejected_docs=10353 rows=53290 buffered=428
+[worker 30] docs=70000 accepted_docs=59809 rejected_docs=10191 rows=53641 buffered=447
+[worker 21] docs=70000 accepted_docs=59624 rejected_docs=10376 rows=53639 buffered=430
+[worker 64] docs=70000 accepted_docs=59692 rejected_docs=10308 rows=54004 buffered=228
+[worker 45] docs=70000 accepted_docs=59733 rejected_docs=10267 rows=53623 buffered=442
+[worker 20] docs=70000 accepted_docs=59739 rejected_docs=10261 rows=53649 buffered=424
+[worker 33] docs=70000 accepted_docs=59822 rejected_docs=10178 rows=53294 buffered=48
+[worker 70] docs=70000 accepted_docs=59748 rejected_docs=10252 rows=53306 buffered=310
+[worker 37] docs=70000 accepted_docs=59769 rejected_docs=10231 rows=53547 buffered=119
+[worker 77] docs=70000 accepted_docs=59739 rejected_docs=10261 rows=53830 buffered=782
+[worker 72] docs=70000 accepted_docs=59619 rejected_docs=10381 rows=53420 buffered=66
+[worker 6] docs=70000 accepted_docs=59803 rejected_docs=10197 rows=53735 buffered=78
+[worker 31] docs=70000 accepted_docs=59643 rejected_docs=10357 rows=53437 buffered=1007
+[worker 8] docs=70000 accepted_docs=59719 rejected_docs=10281 rows=53643 buffered=417
+[worker 44] docs=70000 accepted_docs=59838 rejected_docs=10162 rows=53749 buffered=162
+[worker 35] docs=70000 accepted_docs=59660 rejected_docs=10340 rows=52830 buffered=126
+[worker 26] docs=70000 accepted_docs=59743 rejected_docs=10257 rows=53763 buffered=991
+[worker 19] docs=70000 accepted_docs=59759 rejected_docs=10241 rows=53211 buffered=560
+[worker 67] docs=70000 accepted_docs=59615 rejected_docs=10385 rows=53348 buffered=1021
+[worker 38] docs=70000 accepted_docs=59791 rejected_docs=10209 rows=53527 buffered=108
+[worker 39] docs=70000 accepted_docs=59773 rejected_docs=10227 rows=53701 buffered=573
+[worker 16] docs=70000 accepted_docs=59934 rejected_docs=10066 rows=54097 buffered=498
+[worker 50] docs=70000 accepted_docs=59847 rejected_docs=10153 rows=53936 buffered=17
+[worker 32] docs=70000 accepted_docs=59733 rejected_docs=10267 rows=53771 buffered=778
+[worker 23] docs=70000 accepted_docs=59560 rejected_docs=10440 rows=53698 buffered=383
+[worker 1] docs=70000 accepted_docs=59675 rejected_docs=10325 rows=53885 buffered=910
+[worker 22] docs=70000 accepted_docs=59695 rejected_docs=10305 rows=53879 buffered=965
+[worker 76] docs=70000 accepted_docs=59711 rejected_docs=10289 rows=53615 buffered=553
+[worker 28] docs=70000 accepted_docs=59638 rejected_docs=10362 rows=53257 buffered=11
+[worker 10] docs=70000 accepted_docs=59742 rejected_docs=10258 rows=53802 buffered=347
+[worker 41] docs=70000 accepted_docs=59722 rejected_docs=10278 rows=53517 buffered=864
+[worker 52] docs=70000 accepted_docs=59630 rejected_docs=10370 rows=53372 buffered=542
+[worker 0] docs=70000 accepted_docs=59735 rejected_docs=10265 rows=53573 buffered=710
+[worker 2] docs=70000 accepted_docs=59801 rejected_docs=10199 rows=53571 buffered=148
+[worker 58] docs=70000 accepted_docs=59754 rejected_docs=10246 rows=53512 buffered=755
+[worker 3] docs=70000 accepted_docs=59644 rejected_docs=10356 rows=53510 buffered=220
+[worker 69] docs=80000 accepted_docs=68238 rejected_docs=11762 rows=61281 buffered=99
+[worker 27] docs=80000 accepted_docs=68424 rejected_docs=11576 rows=61263 buffered=308
+[worker 46] docs=80000 accepted_docs=68239 rejected_docs=11761 rows=61177 buffered=653
+[worker 60] docs=80000 accepted_docs=68251 rejected_docs=11749 rows=61045 buffered=545
+[worker 24] docs=80000 accepted_docs=68345 rejected_docs=11655 rows=61133 buffered=78
+[worker 14] docs=80000 accepted_docs=68361 rejected_docs=11639 rows=61439 buffered=720
+[worker 43] docs=80000 accepted_docs=68286 rejected_docs=11714 rows=61437 buffered=295
+[worker 13] docs=80000 accepted_docs=68330 rejected_docs=11670 rows=61292 buffered=926
+[worker 5] docs=80000 accepted_docs=68060 rejected_docs=11940 rows=60723 buffered=123
+[worker 78] docs=80000 accepted_docs=68240 rejected_docs=11760 rows=61057 buffered=983
+[worker 34] docs=80000 accepted_docs=68290 rejected_docs=11710 rows=61185 buffered=447
+[worker 66] docs=80000 accepted_docs=68215 rejected_docs=11785 rows=60708 buffered=462
+[worker 49] docs=80000 accepted_docs=68349 rejected_docs=11651 rows=61250 buffered=577
+[worker 48] docs=80000 accepted_docs=68323 rejected_docs=11677 rows=61146 buffered=433
+[worker 61] docs=80000 accepted_docs=68185 rejected_docs=11815 rows=60772 buffered=25
+[worker 53] docs=80000 accepted_docs=68271 rejected_docs=11729 rows=61292 buffered=210
+[worker 55] docs=80000 accepted_docs=68212 rejected_docs=11788 rows=61055 buffered=565
+[worker 68] docs=80000 accepted_docs=68151 rejected_docs=11849 rows=60886 buffered=173
+[worker 42] docs=80000 accepted_docs=68241 rejected_docs=11759 rows=61161 buffered=704
+[worker 73] docs=80000 accepted_docs=68351 rejected_docs=11649 rows=61237 buffered=223
+[worker 79] docs=80000 accepted_docs=68250 rejected_docs=11750 rows=61125 buffered=715
+[worker 56] docs=80000 accepted_docs=68060 rejected_docs=11940 rows=60918 buffered=352
+[worker 59] docs=80000 accepted_docs=68379 rejected_docs=11621 rows=61025 buffered=310
+[worker 63] docs=80000 accepted_docs=68346 rejected_docs=11654 rows=61082 buffered=273
+[worker 75] docs=80000 accepted_docs=68250 rejected_docs=11750 rows=61002 buffered=797
+[worker 25] docs=80000 accepted_docs=68172 rejected_docs=11828 rows=61024 buffered=176
+[worker 9] docs=80000 accepted_docs=68362 rejected_docs=11638 rows=61275 buffered=468
+[worker 7] docs=80000 accepted_docs=68280 rejected_docs=11720 rows=61310 buffered=721
+[worker 17] docs=80000 accepted_docs=68344 rejected_docs=11656 rows=61525 buffered=614
+[worker 74] docs=80000 accepted_docs=68171 rejected_docs=11829 rows=60864 buffered=28
+[worker 62] docs=80000 accepted_docs=68295 rejected_docs=11705 rows=61263 buffered=414
+[worker 45] docs=80000 accepted_docs=68286 rejected_docs=11714 rows=61334 buffered=318
+[worker 70] docs=80000 accepted_docs=68293 rejected_docs=11707 rows=60874 buffered=420
+[worker 37] docs=80000 accepted_docs=68336 rejected_docs=11664 rows=61292 buffered=827
+[worker 11] docs=80000 accepted_docs=68160 rejected_docs=11840 rows=60945 buffered=401
+[worker 64] docs=80000 accepted_docs=68176 rejected_docs=11824 rows=61662 buffered=204
+[worker 20] docs=80000 accepted_docs=68222 rejected_docs=11778 rows=61340 buffered=65
+[worker 31] docs=80000 accepted_docs=68140 rejected_docs=11860 rows=60919 buffered=165
+[worker 71] docs=80000 accepted_docs=68172 rejected_docs=11828 rows=60733 buffered=745
+[worker 30] docs=80000 accepted_docs=68360 rejected_docs=11640 rows=61229 buffered=828
+[worker 72] docs=80000 accepted_docs=68149 rejected_docs=11851 rows=61035 buffered=659
+[worker 21] docs=80000 accepted_docs=68149 rejected_docs=11851 rows=61282 buffered=717
+[worker 8] docs=80000 accepted_docs=68257 rejected_docs=11743 rows=61331 buffered=8
+[worker 77] docs=80000 accepted_docs=68271 rejected_docs=11729 rows=61485 buffered=485
+[worker 33] docs=80000 accepted_docs=68284 rejected_docs=11716 rows=60751 buffered=316
+[worker 40] docs=80000 accepted_docs=68384 rejected_docs=11616 rows=61351 buffered=694
+[worker 26] docs=80000 accepted_docs=68293 rejected_docs=11707 rows=61477 buffered=728
+[worker 67] docs=80000 accepted_docs=68216 rejected_docs=11784 rows=61013 buffered=776
+[worker 29] docs=80000 accepted_docs=68242 rejected_docs=11758 rows=61080 buffered=79
+[worker 23] docs=80000 accepted_docs=68086 rejected_docs=11914 rows=61246 buffered=693
+[worker 39] docs=80000 accepted_docs=68287 rejected_docs=11713 rows=61294 buffered=107
+[worker 6] docs=80000 accepted_docs=68333 rejected_docs=11667 rows=61438 buffered=385
+[worker 19] docs=80000 accepted_docs=68295 rejected_docs=11705 rows=60815 buffered=840
+[worker 28] docs=80000 accepted_docs=68194 rejected_docs=11806 rows=60994 buffered=260
+[worker 35] docs=80000 accepted_docs=68221 rejected_docs=11779 rows=60529 buffered=525
+[worker 22] docs=80000 accepted_docs=68252 rejected_docs=11748 rows=61531 buffered=879
+[worker 32] docs=80000 accepted_docs=68242 rejected_docs=11758 rows=61473 buffered=376
+[worker 16] docs=80000 accepted_docs=68446 rejected_docs=11554 rows=61701 buffered=780
+[worker 76] docs=80000 accepted_docs=68245 rejected_docs=11755 rows=61225 buffered=526
+[worker 38] docs=80000 accepted_docs=68304 rejected_docs=11696 rows=61258 buffered=496
+[worker 41] docs=80000 accepted_docs=68236 rejected_docs=11764 rows=61178 buffered=293
+[worker 54] docs=80000 accepted_docs=68257 rejected_docs=11743 rows=61392 buffered=953
+[worker 10] docs=80000 accepted_docs=68267 rejected_docs=11733 rows=61435 buffered=388
+[worker 2] docs=80000 accepted_docs=68328 rejected_docs=11672 rows=61241 buffered=178
+[worker 58] docs=80000 accepted_docs=68362 rejected_docs=11638 rows=61186 buffered=907
+[worker 3] docs=80000 accepted_docs=68188 rejected_docs=11812 rows=61242 buffered=608
+[worker 47] docs=90000 accepted_docs=76750 rejected_docs=13250 rows=68536 buffered=599
+[worker 69] docs=90000 accepted_docs=76847 rejected_docs=13153 rows=69092 buffered=664
+[worker 60] docs=90000 accepted_docs=76745 rejected_docs=13255 rows=68557 buffered=427
+[worker 46] docs=90000 accepted_docs=76797 rejected_docs=13203 rows=68908 buffered=99
+[worker 24] docs=90000 accepted_docs=76908 rejected_docs=13092 rows=68772 buffered=652
+[worker 14] docs=90000 accepted_docs=76874 rejected_docs=13126 rows=69040 buffered=1019
+[worker 5] docs=90000 accepted_docs=76564 rejected_docs=13436 rows=68296 buffered=227
+[worker 78] docs=90000 accepted_docs=76830 rejected_docs=13170 rows=68761 buffered=538
+[worker 65] docs=90000 accepted_docs=76914 rejected_docs=13086 rows=68795 buffered=894
+[worker 43] docs=90000 accepted_docs=76833 rejected_docs=13167 rows=69125 buffered=766
+[worker 53] docs=90000 accepted_docs=76795 rejected_docs=13205 rows=68974 buffered=433
+[worker 18] docs=90000 accepted_docs=76816 rejected_docs=13184 rows=68667 buffered=529
+[worker 12] docs=90000 accepted_docs=76893 rejected_docs=13107 rows=68782 buffered=866
+[worker 34] docs=90000 accepted_docs=76805 rejected_docs=13195 rows=68902 buffered=488
+[worker 13] docs=90000 accepted_docs=76901 rejected_docs=13099 rows=69014 buffered=370
+[worker 59] docs=90000 accepted_docs=76883 rejected_docs=13117 rows=68615 buffered=273
+[worker 66] docs=90000 accepted_docs=76759 rejected_docs=13241 rows=68414 buffered=984
+[worker 79] docs=90000 accepted_docs=76772 rejected_docs=13228 rows=68781 buffered=162
+[worker 48] docs=90000 accepted_docs=76821 rejected_docs=13179 rows=68804 buffered=73
+[worker 74] docs=90000 accepted_docs=76702 rejected_docs=13298 rows=68464 buffered=883
+[worker 49] docs=90000 accepted_docs=76916 rejected_docs=13084 rows=68990 buffered=953
+[worker 68] docs=90000 accepted_docs=76687 rejected_docs=13313 rows=68715 buffered=108
+[worker 45] docs=90000 accepted_docs=76800 rejected_docs=13200 rows=69066 buffered=686
+[worker 37] docs=90000 accepted_docs=76849 rejected_docs=13151 rows=68806 buffered=268
+[worker 55] docs=90000 accepted_docs=76774 rejected_docs=13226 rows=68659 buffered=48
+[worker 20] docs=90000 accepted_docs=76756 rejected_docs=13244 rows=68962 buffered=308
+[worker 7] docs=90000 accepted_docs=76849 rejected_docs=13151 rows=68949 buffered=798
+[worker 70] docs=90000 accepted_docs=76811 rejected_docs=13189 rows=68439 buffered=340
+[worker 4] docs=90000 accepted_docs=76952 rejected_docs=13048 rows=68935 buffered=53
+[worker 64] docs=90000 accepted_docs=76682 rejected_docs=13318 rows=69262 buffered=380
+[worker 63] docs=90000 accepted_docs=76849 rejected_docs=13151 rows=68730 buffered=540
+[worker 36] docs=90000 accepted_docs=76694 rejected_docs=13306 rows=68533 buffered=518
+[worker 31] docs=90000 accepted_docs=76650 rejected_docs=13350 rows=68513 buffered=314
+[worker 9] docs=90000 accepted_docs=76905 rejected_docs=13095 rows=68925 buffered=772
+[worker 8] docs=90000 accepted_docs=76759 rejected_docs=13241 rows=69042 buffered=760
+[worker 17] docs=90000 accepted_docs=76910 rejected_docs=13090 rows=69121 buffered=810
+[worker 62] docs=90000 accepted_docs=76837 rejected_docs=13163 rows=68850 buffered=524
+[worker 77] docs=90000 accepted_docs=76738 rejected_docs=13262 rows=69010 buffered=404
+[worker 33] docs=90000 accepted_docs=76830 rejected_docs=13170 rows=68366 buffered=283
+[worker 25] docs=90000 accepted_docs=76684 rejected_docs=13316 rows=68817 buffered=144
+[worker 72] docs=90000 accepted_docs=76688 rejected_docs=13312 rows=68724 buffered=982
+[worker 11] docs=90000 accepted_docs=76686 rejected_docs=13314 rows=68517 buffered=720
+[worker 57] docs=90000 accepted_docs=76840 rejected_docs=13160 rows=68866 buffered=28
+[worker 21] docs=90000 accepted_docs=76658 rejected_docs=13342 rows=68951 buffered=71
+[worker 30] docs=90000 accepted_docs=76857 rejected_docs=13143 rows=68756 buffered=608
+[worker 67] docs=90000 accepted_docs=76780 rejected_docs=13220 rows=68723 buffered=840
+[worker 29] docs=90000 accepted_docs=76786 rejected_docs=13214 rows=68728 buffered=2
+[worker 26] docs=90000 accepted_docs=76789 rejected_docs=13211 rows=69143 buffered=1015
+[worker 40] docs=90000 accepted_docs=76935 rejected_docs=13065 rows=69006 buffered=193
+[worker 32] docs=90000 accepted_docs=76798 rejected_docs=13202 rows=69115 buffered=304
+[worker 23] docs=90000 accepted_docs=76671 rejected_docs=13329 rows=69157 buffered=227
+[worker 28] docs=90000 accepted_docs=76782 rejected_docs=13218 rows=68693 buffered=239
+[worker 6] docs=90000 accepted_docs=76868 rejected_docs=13132 rows=69030 buffered=62
+[worker 44] docs=90000 accepted_docs=76881 rejected_docs=13119 rows=68954 buffered=944
+[worker 41] docs=90000 accepted_docs=76771 rejected_docs=13229 rows=68851 buffered=843
+[worker 22] docs=90000 accepted_docs=76800 rejected_docs=13200 rows=69354 buffered=714
+[worker 19] docs=90000 accepted_docs=76808 rejected_docs=13192 rows=68460 buffered=949
+[worker 15] docs=90000 accepted_docs=76701 rejected_docs=13299 rows=68972 buffered=21
+[worker 50] docs=90000 accepted_docs=76931 rejected_docs=13069 rows=69301 buffered=354
+[worker 35] docs=90000 accepted_docs=76763 rejected_docs=13237 rows=68306 buffered=458
+[worker 54] docs=90000 accepted_docs=76849 rejected_docs=13151 rows=69188 buffered=372
+[worker 10] docs=90000 accepted_docs=76792 rejected_docs=13208 rows=69099 buffered=10
+[worker 1] docs=90000 accepted_docs=76711 rejected_docs=13289 rows=69238 buffered=127
+[worker 0] docs=90000 accepted_docs=76796 rejected_docs=13204 rows=68829 buffered=898
+[worker 51] docs=90000 accepted_docs=76651 rejected_docs=13349 rows=68780 buffered=402
+[worker 52] docs=90000 accepted_docs=76571 rejected_docs=13429 rows=68591 buffered=348
+[worker 58] docs=90000 accepted_docs=76828 rejected_docs=13172 rows=68875 buffered=101
+[worker 2] docs=90000 accepted_docs=76849 rejected_docs=13151 rows=69062 buffered=905
+[worker 69] docs=100000 accepted_docs=85393 rejected_docs=14607 rows=76802 buffered=908
+[stream cache] worker=69 done docs=100172 rows=76933 accepted_docs=85538 rejected_docs=14634 kept_rows=76933 leftover=363
+[worker 60] docs=100000 accepted_docs=85329 rejected_docs=14671 rows=76219 buffered=559
+[worker 47] docs=100000 accepted_docs=85126 rejected_docs=14874 rows=76233 buffered=526
+[stream cache] worker=60 done docs=100172 rows=76347 accepted_docs=85478 rejected_docs=14694 kept_rows=76347 leftover=852
+[stream cache] worker=47 done docs=100172 rows=76365 accepted_docs=85269 rejected_docs=14903 kept_rows=76365 leftover=951
+[worker 27] docs=100000 accepted_docs=85487 rejected_docs=14513 rows=76507 buffered=526
+[worker 14] docs=100000 accepted_docs=85409 rejected_docs=14591 rows=76783 buffered=334
+[worker 46] docs=100000 accepted_docs=85296 rejected_docs=14704 rows=76401 buffered=560
+[worker 24] docs=100000 accepted_docs=85464 rejected_docs=14536 rows=76476 buffered=302
+[stream cache] worker=27 done docs=100172 rows=76663 accepted_docs=85634 rejected_docs=14538 kept_rows=76663 leftover=282
+[worker 78] docs=100000 accepted_docs=85401 rejected_docs=14599 rows=76478 buffered=630
+[stream cache] worker=46 done docs=100172 rows=76517 accepted_docs=85444 rejected_docs=14728 kept_rows=76517 leftover=714
+[stream cache] worker=14 done docs=100172 rows=76917 accepted_docs=85552 rejected_docs=14620 kept_rows=76917 leftover=142
+[stream cache] worker=24 done docs=100172 rows=76607 accepted_docs=85615 rejected_docs=14557 kept_rows=76607 leftover=1019
+[stream cache] worker=78 done docs=100172 rows=76610 accepted_docs=85542 rejected_docs=14630 kept_rows=76610 leftover=6
+[stream cache] worker=53 done docs=100172 rows=76879 accepted_docs=85443 rejected_docs=14729 kept_rows=76879 leftover=944
+[worker 12] docs=100000 accepted_docs=85432 rejected_docs=14568 rows=76459 buffered=1010
+[worker 59] docs=100000 accepted_docs=85475 rejected_docs=14525 rows=76181 buffered=92
+[worker 5] docs=100000 accepted_docs=85052 rejected_docs=14948 rows=75940 buffered=258
+[stream cache] worker=12 done docs=100172 rows=76585 accepted_docs=85580 rejected_docs=14592 kept_rows=76585 leftover=395
+[worker 74] docs=100000 accepted_docs=85278 rejected_docs=14722 rows=76134 buffered=941
+[stream cache] worker=34 done docs=100172 rows=76731 accepted_docs=85495 rejected_docs=14677 kept_rows=76731 leftover=128
+[worker 61] docs=100000 accepted_docs=85249 rejected_docs=14751 rows=75994 buffered=195
+[stream cache] worker=59 done docs=100172 rows=76299 accepted_docs=85624 rejected_docs=14548 kept_rows=76299 leftover=844
+[stream cache] worker=65 done docs=100172 rows=76704 accepted_docs=85578 rejected_docs=14594 kept_rows=76704 leftover=302
+[stream cache] worker=42 done docs=100172 rows=76452 accepted_docs=85490 rejected_docs=14682 kept_rows=76452 leftover=531
+[stream cache] worker=74 done docs=100172 rows=76272 accepted_docs=85428 rejected_docs=14744 kept_rows=76272 leftover=359
+[worker 79] docs=100000 accepted_docs=85308 rejected_docs=14692 rows=76583 buffered=767
+[stream cache] worker=5 done docs=100173 rows=76098 accepted_docs=85199 rejected_docs=14974 kept_rows=76098 leftover=466
+[worker 18] docs=100000 accepted_docs=85297 rejected_docs=14703 rows=76137 buffered=623
+[worker 45] docs=100000 accepted_docs=85352 rejected_docs=14648 rows=76683 buffered=571
+[worker 64] docs=100000 accepted_docs=85263 rejected_docs=14737 rows=76912 buffered=636
+[worker 20] docs=100000 accepted_docs=85348 rejected_docs=14652 rows=76707 buffered=620
+[stream cache] worker=61 done docs=100172 rows=76109 accepted_docs=85386 rejected_docs=14786 kept_rows=76109 leftover=150
+[worker 13] docs=100000 accepted_docs=85414 rejected_docs=14586 rows=76667 buffered=853
+[worker 43] docs=100000 accepted_docs=85314 rejected_docs=14686 rows=76775 buffered=773
+[stream cache] worker=66 done docs=100172 rows=76152 accepted_docs=85503 rejected_docs=14669 kept_rows=76152 leftover=89
+[stream cache] worker=79 done docs=100172 rows=76701 accepted_docs=85453 rejected_docs=14719 kept_rows=76701 leftover=291
+[stream cache] worker=45 done docs=100172 rows=76819 accepted_docs=85505 rejected_docs=14667 kept_rows=76819 leftover=830
+[stream cache] worker=64 done docs=100172 rows=77031 accepted_docs=85411 rejected_docs=14761 kept_rows=77031 leftover=468
+[worker 70] docs=100000 accepted_docs=85392 rejected_docs=14608 rows=76155 buffered=195
+[stream cache] worker=18 done docs=100172 rows=76268 accepted_docs=85444 rejected_docs=14728 kept_rows=76268 leftover=719
+[worker 48] docs=100000 accepted_docs=85414 rejected_docs=14586 rows=76467 buffered=370
+[stream cache] worker=73 done docs=100172 rows=76631 accepted_docs=85656 rejected_docs=14516 kept_rows=76631 leftover=728
+[stream cache] worker=20 done docs=100172 rows=76839 accepted_docs=85497 rejected_docs=14675 kept_rows=76839 leftover=388
+[worker 68] docs=100000 accepted_docs=85180 rejected_docs=14820 rows=76243 buffered=686
+[stream cache] worker=43 done docs=100172 rows=76906 accepted_docs=85468 rejected_docs=14704 kept_rows=76906 leftover=445
+[worker 8] docs=100000 accepted_docs=85293 rejected_docs=14707 rows=76715 buffered=505
+[stream cache] worker=13 done docs=100172 rows=76810 accepted_docs=85560 rejected_docs=14612 kept_rows=76810 leftover=757
+[worker 37] docs=100000 accepted_docs=85379 rejected_docs=14621 rows=76506 buffered=449
+[worker 31] docs=100000 accepted_docs=85112 rejected_docs=14888 rows=76044 buffered=70
+[worker 56] docs=100000 accepted_docs=85203 rejected_docs=14797 rows=76303 buffered=28
+[stream cache] worker=48 done docs=100172 rows=76600 accepted_docs=85564 rejected_docs=14608 kept_rows=76600 leftover=548
+[stream cache] worker=70 done docs=100172 rows=76286 accepted_docs=85535 rejected_docs=14637 kept_rows=76286 leftover=825
+[worker 7] docs=100000 accepted_docs=85356 rejected_docs=14644 rows=76615 buffered=455
+[stream cache] worker=68 done docs=100172 rows=76384 accepted_docs=85325 rejected_docs=14847 kept_rows=76384 leftover=517
+[stream cache] worker=8 done docs=100173 rows=76840 accepted_docs=85443 rejected_docs=14730 kept_rows=76840 leftover=663
+[worker 4] docs=100000 accepted_docs=85453 rejected_docs=14547 rows=76583 buffered=233
+[worker 9] docs=100000 accepted_docs=85479 rejected_docs=14521 rows=76645 buffered=515
+[stream cache] worker=37 done docs=100172 rows=76650 accepted_docs=85529 rejected_docs=14643 kept_rows=76650 leftover=734
+[stream cache] worker=56 done docs=100172 rows=76439 accepted_docs=85352 rejected_docs=14820 kept_rows=76439 leftover=68
+[stream cache] worker=31 done docs=100172 rows=76183 accepted_docs=85257 rejected_docs=14915 kept_rows=76183 leftover=147
+[worker 49] docs=100000 accepted_docs=85431 rejected_docs=14569 rows=76634 buffered=980
+[worker 72] docs=100000 accepted_docs=85343 rejected_docs=14657 rows=76457 buffered=835
+[stream cache] worker=7 done docs=100173 rows=76746 accepted_docs=85504 rejected_docs=14669 kept_rows=76746 leftover=522
+[worker 36] docs=100000 accepted_docs=85232 rejected_docs=14768 rows=76162 buffered=937
+[stream cache] worker=4 done docs=100173 rows=76705 accepted_docs=85599 rejected_docs=14574 kept_rows=76705 leftover=977
+[worker 77] docs=100000 accepted_docs=85329 rejected_docs=14671 rows=76689 buffered=146
+[stream cache] worker=9 done docs=100172 rows=76772 accepted_docs=85624 rejected_docs=14548 kept_rows=76772 leftover=742
+[stream cache] worker=63 done docs=100172 rows=76354 accepted_docs=85543 rejected_docs=14629 kept_rows=76354 leftover=482
+[stream cache] worker=49 done docs=100172 rows=76778 accepted_docs=85584 rejected_docs=14588 kept_rows=76778 leftover=738
+[stream cache] worker=72 done docs=100172 rows=76586 accepted_docs=85486 rejected_docs=14686 kept_rows=76586 leftover=692
+[worker 25] docs=100000 accepted_docs=85241 rejected_docs=14759 rows=76512 buffered=94
+[stream cache] worker=36 done docs=100172 rows=76300 accepted_docs=85376 rejected_docs=14796 kept_rows=76300 leftover=396
+[worker 55] docs=100000 accepted_docs=85336 rejected_docs=14664 rows=76388 buffered=547
+[stream cache] worker=77 done docs=100172 rows=76811 accepted_docs=85479 rejected_docs=14693 kept_rows=76811 leftover=388
+[worker 33] docs=100000 accepted_docs=85358 rejected_docs=14642 rows=76000 buffered=943
+[worker 62] docs=100000 accepted_docs=85395 rejected_docs=14605 rows=76608 buffered=798
+[stream cache] worker=25 done docs=100172 rows=76665 accepted_docs=85385 rejected_docs=14787 kept_rows=76665 leftover=543
+[worker 17] docs=100000 accepted_docs=85497 rejected_docs=14503 rows=76743 buffered=995
+[stream cache] worker=21 done docs=100172 rows=76719 accepted_docs=85356 rejected_docs=14816 kept_rows=76719 leftover=618
+[stream cache] worker=55 done docs=100172 rows=76517 accepted_docs=85475 rejected_docs=14697 kept_rows=76517 leftover=16
+[stream cache] worker=33 done docs=100172 rows=76126 accepted_docs=85503 rejected_docs=14669 kept_rows=76126 leftover=149
+[worker 75] docs=100000 accepted_docs=85234 rejected_docs=14766 rows=76267 buffered=297
+[stream cache] worker=62 done docs=100172 rows=76747 accepted_docs=85545 rejected_docs=14627 kept_rows=76747 leftover=555
+[stream cache] worker=17 done docs=100172 rows=76877 accepted_docs=85647 rejected_docs=14525 kept_rows=76877 leftover=222
+[worker 39] docs=100000 accepted_docs=85419 rejected_docs=14581 rows=76640 buffered=909
+[worker 32] docs=100000 accepted_docs=85405 rejected_docs=14595 rows=76934 buffered=38
+[stream cache] worker=75 done docs=100172 rows=76399 accepted_docs=85380 rejected_docs=14792 kept_rows=76399 leftover=776
+[worker 57] docs=100000 accepted_docs=85337 rejected_docs=14663 rows=76503 buffered=677
+[stream cache] worker=11 done docs=100172 rows=76252 accepted_docs=85357 rejected_docs=14815 kept_rows=76252 leftover=134
+[stream cache] worker=39 done docs=100172 rows=76772 accepted_docs=85563 rejected_docs=14609 kept_rows=76772 leftover=592
+[stream cache] worker=32 done docs=100172 rows=77067 accepted_docs=85543 rejected_docs=14629 kept_rows=77067 leftover=923
+[worker 23] docs=100000 accepted_docs=85220 rejected_docs=14780 rows=76824 buffered=453
+[stream cache] worker=57 done docs=100172 rows=76647 accepted_docs=85486 rejected_docs=14686 kept_rows=76647 leftover=276
+[worker 30] docs=100000 accepted_docs=85332 rejected_docs=14668 rows=76396 buffered=986
+[worker 71] docs=100000 accepted_docs=85176 rejected_docs=14824 rows=75993 buffered=743
+[worker 67] docs=100000 accepted_docs=85282 rejected_docs=14718 rows=76309 buffered=8
+[worker 40] docs=100000 accepted_docs=85458 rejected_docs=14542 rows=76808 buffered=303
+[stream cache] worker=23 done docs=100172 rows=76976 accepted_docs=85363 rejected_docs=14809 kept_rows=76976 leftover=431
+[stream cache] worker=30 done docs=100172 rows=76519 accepted_docs=85477 rejected_docs=14695 kept_rows=76519 leftover=283
+[worker 22] docs=100000 accepted_docs=85374 rejected_docs=14626 rows=77014 buffered=391
+[worker 26] docs=100000 accepted_docs=85299 rejected_docs=14701 rows=76808 buffered=351
+[stream cache] worker=71 done docs=100172 rows=76127 accepted_docs=85322 rejected_docs=14850 kept_rows=76127 leftover=899
+[worker 6] docs=100000 accepted_docs=85384 rejected_docs=14616 rows=76555 buffered=1021
+[worker 28] docs=100000 accepted_docs=85348 rejected_docs=14652 rows=76332 buffered=775
+[stream cache] worker=67 done docs=100172 rows=76436 accepted_docs=85428 rejected_docs=14744 kept_rows=76436 leftover=697
+[stream cache] worker=22 done docs=100172 rows=77152 accepted_docs=85518 rejected_docs=14654 kept_rows=77152 leftover=1006
+[stream cache] worker=40 done docs=100172 rows=76943 accepted_docs=85611 rejected_docs=14561 kept_rows=76943 leftover=399
+[stream cache] worker=26 done docs=100172 rows=76933 accepted_docs=85450 rejected_docs=14722 kept_rows=76933 leftover=562
+[worker 35] docs=100000 accepted_docs=85298 rejected_docs=14702 rows=75782 buffered=943
+[worker 29] docs=100000 accepted_docs=85294 rejected_docs=14706 rows=76339 buffered=469
+[stream cache] worker=6 done docs=100173 rows=76689 accepted_docs=85537 rejected_docs=14636 kept_rows=76689 leftover=801
+[stream cache] worker=28 done docs=100172 rows=76469 accepted_docs=85497 rejected_docs=14675 kept_rows=76469 leftover=186
+[stream cache] worker=35 done docs=100172 rows=75906 accepted_docs=85446 rejected_docs=14726 kept_rows=75906 leftover=266
+[worker 44] docs=100000 accepted_docs=85435 rejected_docs=14565 rows=76493 buffered=458
+[worker 38] docs=100000 accepted_docs=85390 rejected_docs=14610 rows=76616 buffered=472
+[stream cache] worker=29 done docs=100172 rows=76497 accepted_docs=85437 rejected_docs=14735 kept_rows=76497 leftover=555
+[worker 41] docs=100000 accepted_docs=85277 rejected_docs=14723 rows=76551 buffered=688
+[stream cache] worker=38 done docs=100172 rows=76754 accepted_docs=85538 rejected_docs=14634 kept_rows=76754 leftover=25
+[stream cache] worker=44 done docs=100172 rows=76622 accepted_docs=85584 rejected_docs=14588 kept_rows=76622 leftover=794
+[worker 19] docs=100000 accepted_docs=85352 rejected_docs=14648 rows=76059 buffered=640
+[stream cache] worker=41 done docs=100172 rows=76698 accepted_docs=85427 rejected_docs=14745 kept_rows=76698 leftover=312
+[worker 54] docs=100000 accepted_docs=85406 rejected_docs=14594 rows=76820 buffered=575
+[worker 50] docs=100000 accepted_docs=85520 rejected_docs=14480 rows=76890 buffered=898
+[stream cache] worker=19 done docs=100172 rows=76180 accepted_docs=85493 rejected_docs=14679 kept_rows=76180 leftover=586
+[worker 15] docs=100000 accepted_docs=85199 rejected_docs=14801 rows=76558 buffered=851
+[stream cache] worker=54 done docs=100172 rows=76957 accepted_docs=85562 rejected_docs=14610 kept_rows=76957 leftover=990
+[stream cache] worker=50 done docs=100172 rows=77030 accepted_docs=85669 rejected_docs=14503 kept_rows=77030 leftover=380
+[stream cache] worker=16 done docs=100172 rows=77184 accepted_docs=85581 rejected_docs=14591 kept_rows=77184 leftover=282
+[stream cache] worker=15 done docs=100172 rows=76699 accepted_docs=85345 rejected_docs=14827 kept_rows=76699 leftover=644
+[worker 76] docs=100000 accepted_docs=85324 rejected_docs=14676 rows=76579 buffered=484
+[worker 10] docs=100000 accepted_docs=85365 rejected_docs=14635 rows=76740 buffered=181
+[worker 1] docs=100000 accepted_docs=85276 rejected_docs=14724 rows=76900 buffered=421
+[stream cache] worker=76 done docs=100172 rows=76694 accepted_docs=85463 rejected_docs=14709 kept_rows=76694 leftover=280
+[stream cache] worker=10 done docs=100172 rows=76860 accepted_docs=85508 rejected_docs=14664 kept_rows=76860 leftover=242
+[stream cache] worker=1 done docs=100173 rows=77028 accepted_docs=85425 rejected_docs=14748 kept_rows=77028 leftover=721
+[worker 0] docs=100000 accepted_docs=85354 rejected_docs=14646 rows=76394 buffered=488
+[stream cache] worker=0 done docs=100173 rows=76529 accepted_docs=85501 rejected_docs=14672 kept_rows=76529 leftover=857
+[worker 51] docs=100000 accepted_docs=85130 rejected_docs=14870 rows=76465 buffered=562
+[stream cache] worker=51 done docs=100172 rows=76612 accepted_docs=85277 rejected_docs=14895 kept_rows=76612 leftover=397
+[worker 52] docs=100000 accepted_docs=85147 rejected_docs=14853 rows=76232 buffered=956
+[stream cache] worker=52 done docs=100172 rows=76376 accepted_docs=85300 rejected_docs=14872 kept_rows=76376 leftover=200
+[worker 3] docs=100000 accepted_docs=85269 rejected_docs=14731 rows=76525 buffered=968
+[stream cache] worker=58 done docs=100172 rows=76597 accepted_docs=85519 rejected_docs=14653 kept_rows=76597 leftover=999
+[stream cache] worker=3 done docs=100173 rows=76649 accepted_docs=85411 rejected_docs=14762 kept_rows=76649 leftover=470
+[worker 2] docs=100000 accepted_docs=85397 rejected_docs=14603 rows=76860 buffered=639
+[stream cache] worker=2 done docs=100173 rows=76997 accepted_docs=85536 rejected_docs=14637 kept_rows=76997 leftover=255
+[stream cache] saved=cache/owt_t5_stream_pack1023_clean_ngram32_appendeos1.pt shape=(6129580, 1024) docs=8013769 raw_rows=6129580 accepted_docs=6838457 rejected_docs=1175312 kept_rows=6129580 reject_txt=cache/owt_t5_stream_pack1023_clean_ngram32_rejected.txt

LTA_openwebtext_dualt/mini_owt_logdirichlet/logs/debug_articlefull_elfopt_no_usersite.log

@@ -0,0 +1,49 @@
+Muon: 54 2D params; Nesterov-AdamW: 76 other params
+{
+  "data_mode": "cache",
+  "cache_path": "cache/owt_t5_llmclean_qwen36_35b_articlefull_pack1023_10k_appendeos1.pt",
+  "data_path": "/e2e-data/evad-tech-vla/wanghan58/data/small_benchmarks/langflow_2604_11748/openwebtext",
+  "tokenizer_path": "/e2e-data/evad-tech-vla/wanghan58/models/hf/t5-small/tokenizer.json",
+  "text_column": "text",
+  "pack_len": 1023,
+  "append_eos": 1,
+  "num_workers": 0,
+  "shuffle_buffer": 8192,
+  "reject_txt": "cache/online_rejected.txt",
+  "out_dir": "runs/debug_articlefull_elfopt_no_usersite",
+  "subset_size": 10000,
+  "resume": "",
+  "steps": 1,
+  "batch_size": 2,
+  "grad_accum": 1,
+  "lr": 7.8125e-06,
+  "blr": 0.001,
+  "min_lr": 0.0,
+  "lr_schedule": "constant",
+  "warmup_steps": 2500,
+  "warmup_epochs": 0.5,
+  "optimizer": "muon",
+  "weight_decay": 0.0,
+  "adam_beta1": 0.9,
+  "adam_beta2": 0.95,
+  "adam_eps": 1e-08,
+  "grad_clip": 1.0,
+  "log_every": 1,
+  "save_every": 999999,
+  "dim": 768,
+  "layers": 12,
+  "heads": 12,
+  "mlp_dim": 3072,
+  "time_tokens": 4,
+  "c_min": 1.0,
+  "c_max": 1024.0,
+  "c_schedule": "sqrt",
+  "seed": 1234,
+  "loader_batches_per_rank": 5000,
+  "optimizer_steps_per_epoch": 5000,
+  "steps_per_epoch": 5000,
+  "effective_batch_size": 2
+}
+[data] mode=cache rows=10000 length=1024 vocab=32100 seen=24862 dropped=2100 bos=1:</s> eos=1:</s>
+[optim] optimizer=muon lr=7.812500e-06 blr=1.000000e-03 effective_batch=2 warmup_steps=2500 lr_schedule=constant wd=0.0 loader_batches_per_rank=5000 optimizer_steps_per_epoch=5000
+step=1 lr=3.125000e-09 loss=10.5609 {'pos0_bos_p': 2.3223959942697547e-05, 'pos0_bos_top1': 0, 'last_eos_p': 2.623751242936123e-05, 'last_eos_top1': 0}

LTA_openwebtext_dualt/mini_owt_logdirichlet/logs/owt_t5_elftokenized_full_len1024_C1_to_1024_pow1_d768_l12_h12_gbs512_8gpu_5ep_elfopt_t5embed_unfixed_probadd_selfcond_ce_20260531_174225.log

@@ -0,0 +1,949 @@
+[data] path=/e2e-data/evad-tech-vla/wanghan58/data/embedded-language-flows/openwebtext-t5
+[data] t5_tokenized_arrow_shards=75 rows=9737184 steps_per_epoch_gbs512=19018 total_steps_5ep=95090
+W0531 17:42:27.055000 10365 torch/distributed/run.py:792] 
+W0531 17:42:27.055000 10365 torch/distributed/run.py:792] *****************************************
+W0531 17:42:27.055000 10365 torch/distributed/run.py:792] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. 
+W0531 17:42:27.055000 10365 torch/distributed/run.py:792] *****************************************
+[W531 17:42:29.733675142 Utils.hpp:136] Warning: Environment variable NCCL_ASYNC_ERROR_HANDLING is deprecated; use TORCH_NCCL_ASYNC_ERROR_HANDLING instead (function operator())
+[W531 17:42:29.743356103 Utils.hpp:136] Warning: Environment variable NCCL_ASYNC_ERROR_HANDLING is deprecated; use TORCH_NCCL_ASYNC_ERROR_HANDLING instead (function operator())
+[W531 17:42:29.764802164 Utils.hpp:136] Warning: Environment variable NCCL_ASYNC_ERROR_HANDLING is deprecated; use TORCH_NCCL_ASYNC_ERROR_HANDLING instead (function operator())
+[W531 17:42:29.767936896 Utils.hpp:136] Warning: Environment variable NCCL_ASYNC_ERROR_HANDLING is deprecated; use TORCH_NCCL_ASYNC_ERROR_HANDLING instead (function operator())
+[W531 17:42:29.770258401 Utils.hpp:136] Warning: Environment variable NCCL_ASYNC_ERROR_HANDLING is deprecated; use TORCH_NCCL_ASYNC_ERROR_HANDLING instead (function operator())
+[W531 17:42:29.770939740 Utils.hpp:136] Warning: Environment variable NCCL_ASYNC_ERROR_HANDLING is deprecated; use TORCH_NCCL_ASYNC_ERROR_HANDLING instead (function operator())
+[W531 17:42:29.771197877 Utils.hpp:136] Warning: Environment variable NCCL_ASYNC_ERROR_HANDLING is deprecated; use TORCH_NCCL_ASYNC_ERROR_HANDLING instead (function operator())
+[W531 17:42:29.771434826 Utils.hpp:136] Warning: Environment variable NCCL_ASYNC_ERROR_HANDLING is deprecated; use TORCH_NCCL_ASYNC_ERROR_HANDLING instead (function operator())
+t-20260601014141-vst28-worker-0:10432:10432 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth1
+t-20260601014141-vst28-worker-0:10432:10432 [0] NCCL INFO Bootstrap: Using eth1:10.82.112.4<0>
+t-20260601014141-vst28-worker-0:10432:10432 [0] NCCL INFO cudaDriverVersion 12080
+t-20260601014141-vst28-worker-0:10432:10432 [0] NCCL INFO NCCL version 2.25.1+cuda12.8
+t-20260601014141-vst28-worker-0:10432:10432 [0] NCCL INFO Comm config Blocking set to 1
+t-20260601014141-vst28-worker-0:10439:10439 [6] NCCL INFO cudaDriverVersion 12080
+t-20260601014141-vst28-worker-0:10439:10439 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth1
+t-20260601014141-vst28-worker-0:10438:10438 [5] NCCL INFO cudaDriverVersion 12080
+t-20260601014141-vst28-worker-0:10438:10438 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth1
+t-20260601014141-vst28-worker-0:10439:10439 [6] NCCL INFO Bootstrap: Using eth1:10.82.112.4<0>
+t-20260601014141-vst28-worker-0:10438:10438 [5] NCCL INFO Bootstrap: Using eth1:10.82.112.4<0>
+t-20260601014141-vst28-worker-0:10438:10438 [5] NCCL INFO NCCL version 2.25.1+cuda12.8
+t-20260601014141-vst28-worker-0:10439:10439 [6] NCCL INFO NCCL version 2.25.1+cuda12.8
+t-20260601014141-vst28-worker-0:10433:10433 [1] NCCL INFO cudaDriverVersion 12080
+t-20260601014141-vst28-worker-0:10433:10433 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth1
+t-20260601014141-vst28-worker-0:10433:10433 [1] NCCL INFO Bootstrap: Using eth1:10.82.112.4<0>
+t-20260601014141-vst28-worker-0:10433:10433 [1] NCCL INFO NCCL version 2.25.1+cuda12.8
+t-20260601014141-vst28-worker-0:10438:10438 [5] NCCL INFO Comm config Blocking set to 1
+t-20260601014141-vst28-worker-0:10439:10439 [6] NCCL INFO Comm config Blocking set to 1
+t-20260601014141-vst28-worker-0:10436:10436 [4] NCCL INFO cudaDriverVersion 12080
+t-20260601014141-vst28-worker-0:10436:10436 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth1
+t-20260601014141-vst28-worker-0:10433:10433 [1] NCCL INFO Comm config Blocking set to 1
+t-20260601014141-vst28-worker-0:10440:10440 [7] NCCL INFO cudaDriverVersion 12080
+t-20260601014141-vst28-worker-0:10440:10440 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth1
+t-20260601014141-vst28-worker-0:10436:10436 [4] NCCL INFO Bootstrap: Using eth1:10.82.112.4<0>
+t-20260601014141-vst28-worker-0:10436:10436 [4] NCCL INFO NCCL version 2.25.1+cuda12.8
+t-20260601014141-vst28-worker-0:10440:10440 [7] NCCL INFO Bootstrap: Using eth1:10.82.112.4<0>
+t-20260601014141-vst28-worker-0:10440:10440 [7] NCCL INFO NCCL version 2.25.1+cuda12.8
+t-20260601014141-vst28-worker-0:10434:10434 [2] NCCL INFO cudaDriverVersion 12080
+t-20260601014141-vst28-worker-0:10434:10434 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth1
+t-20260601014141-vst28-worker-0:10434:10434 [2] NCCL INFO Bootstrap: Using eth1:10.82.112.4<0>
+t-20260601014141-vst28-worker-0:10434:10434 [2] NCCL INFO NCCL version 2.25.1+cuda12.8
+t-20260601014141-vst28-worker-0:10435:10435 [3] NCCL INFO cudaDriverVersion 12080
+t-20260601014141-vst28-worker-0:10435:10435 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth1
+t-20260601014141-vst28-worker-0:10435:10435 [3] NCCL INFO Bootstrap: Using eth1:10.82.112.4<0>
+t-20260601014141-vst28-worker-0:10435:10435 [3] NCCL INFO NCCL version 2.25.1+cuda12.8
+t-20260601014141-vst28-worker-0:10436:10436 [4] NCCL INFO Comm config Blocking set to 1
+t-20260601014141-vst28-worker-0:10440:10440 [7] NCCL INFO Comm config Blocking set to 1
+t-20260601014141-vst28-worker-0:10434:10434 [2] NCCL INFO Comm config Blocking set to 1
+t-20260601014141-vst28-worker-0:10435:10435 [3] NCCL INFO Comm config Blocking set to 1
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO NET/Plugin: Loaded net plugin NCCL RDMA Plugin v9 (v9)
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO NET/Plugin: Loaded collnet plugin SHARP (v9)
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO Plugin Path : /opt/hpcx/nccl_rdma_sharp_plugin/lib/libnccl-net.so
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO P2P plugin v9 IBext_v9
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth1
+t-20260601014141-vst28-worker-0:10433:10531 [1] NCCL INFO NET/Plugin: Loaded net plugin NCCL RDMA Plugin v9 (v9)
+t-20260601014141-vst28-worker-0:10433:10531 [1] NCCL INFO NET/Plugin: Loaded collnet plugin SHARP (v9)
+t-20260601014141-vst28-worker-0:10433:10531 [1] NCCL INFO Plugin Path : /opt/hpcx/nccl_rdma_sharp_plugin/lib/libnccl-net.so
+t-20260601014141-vst28-worker-0:10433:10531 [1] NCCL INFO P2P plugin v9 IBext_v9
+t-20260601014141-vst28-worker-0:10433:10531 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth1
+t-20260601014141-vst28-worker-0:10439:10530 [6] NCCL INFO NET/Plugin: Loaded net plugin NCCL RDMA Plugin v9 (v9)
+t-20260601014141-vst28-worker-0:10439:10530 [6] NCCL INFO NET/Plugin: Loaded collnet plugin SHARP (v9)
+t-20260601014141-vst28-worker-0:10439:10530 [6] NCCL INFO Plugin Path : /opt/hpcx/nccl_rdma_sharp_plugin/lib/libnccl-net.so
+t-20260601014141-vst28-worker-0:10439:10530 [6] NCCL INFO P2P plugin v9 IBext_v9
+t-20260601014141-vst28-worker-0:10439:10530 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth1
+t-20260601014141-vst28-worker-0:10440:10533 [7] NCCL INFO NET/Plugin: Loaded net plugin NCCL RDMA Plugin v9 (v9)
+t-20260601014141-vst28-worker-0:10440:10533 [7] NCCL INFO NET/Plugin: Loaded collnet plugin SHARP (v9)
+t-20260601014141-vst28-worker-0:10440:10533 [7] NCCL INFO Plugin Path : /opt/hpcx/nccl_rdma_sharp_plugin/lib/libnccl-net.so
+t-20260601014141-vst28-worker-0:10440:10533 [7] NCCL INFO P2P plugin v9 IBext_v9
+t-20260601014141-vst28-worker-0:10436:10532 [4] NCCL INFO NET/Plugin: Loaded net plugin NCCL RDMA Plugin v9 (v9)
+t-20260601014141-vst28-worker-0:10436:10532 [4] NCCL INFO NET/Plugin: Loaded collnet plugin SHARP (v9)
+t-20260601014141-vst28-worker-0:10440:10533 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth1
+t-20260601014141-vst28-worker-0:10436:10532 [4] NCCL INFO Plugin Path : /opt/hpcx/nccl_rdma_sharp_plugin/lib/libnccl-net.so
+t-20260601014141-vst28-worker-0:10436:10532 [4] NCCL INFO P2P plugin v9 IBext_v9
+t-20260601014141-vst28-worker-0:10436:10532 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth1
+t-20260601014141-vst28-worker-0:10435:10535 [3] NCCL INFO NET/Plugin: Loaded net plugin NCCL RDMA Plugin v9 (v9)
+t-20260601014141-vst28-worker-0:10435:10535 [3] NCCL INFO NET/Plugin: Loaded collnet plugin SHARP (v9)
+t-20260601014141-vst28-worker-0:10435:10535 [3] NCCL INFO Plugin Path : /opt/hpcx/nccl_rdma_sharp_plugin/lib/libnccl-net.so
+t-20260601014141-vst28-worker-0:10435:10535 [3] NCCL INFO P2P plugin v9 IBext_v9
+t-20260601014141-vst28-worker-0:10435:10535 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth1
+t-20260601014141-vst28-worker-0:10434:10534 [2] NCCL INFO NET/Plugin: Loaded net plugin NCCL RDMA Plugin v9 (v9)
+t-20260601014141-vst28-worker-0:10434:10534 [2] NCCL INFO NET/Plugin: Loaded collnet plugin SHARP (v9)
+t-20260601014141-vst28-worker-0:10434:10534 [2] NCCL INFO Plugin Path : /opt/hpcx/nccl_rdma_sharp_plugin/lib/libnccl-net.so
+t-20260601014141-vst28-worker-0:10434:10534 [2] NCCL INFO P2P plugin v9 IBext_v9
+t-20260601014141-vst28-worker-0:10434:10534 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth1
+t-20260601014141-vst28-worker-0:10438:10529 [5] NCCL INFO NET/Plugin: Loaded net plugin NCCL RDMA Plugin v9 (v9)
+t-20260601014141-vst28-worker-0:10438:10529 [5] NCCL INFO NET/Plugin: Loaded collnet plugin SHARP (v9)
+t-20260601014141-vst28-worker-0:10438:10529 [5] NCCL INFO Plugin Path : /opt/hpcx/nccl_rdma_sharp_plugin/lib/libnccl-net.so
+t-20260601014141-vst28-worker-0:10438:10529 [5] NCCL INFO P2P plugin v9 IBext_v9
+t-20260601014141-vst28-worker-0:10438:10529 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth1
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO NCCL_IB_PCI_RELAXED_ORDERING set by environment to 1.
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO NET/IB : Using [0]mlx5_1:1/RoCE [1]mlx5_4:1/RoCE [2]mlx5_5:1/RoCE [3]mlx5_6:1/RoCE [4]mlx5_7:1/RoCE [5]mlx5_8:1/RoCE [6]mlx5_9:1/RoCE [7]mlx5_10:1/RoCE [RO]; OOB eth1:10.82.112.4<0>
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO PROFILER/Plugin: Could not find: libnccl-profiler.so.
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO Using network IBext_v9
+t-20260601014141-vst28-worker-0:10440:10533 [7] NCCL INFO NCCL_IB_PCI_RELAXED_ORDERING set by environment to 1.
+t-20260601014141-vst28-worker-0:10440:10533 [7] NCCL INFO NET/IB : Using [0]mlx5_1:1/RoCE [1]mlx5_4:1/RoCE [2]mlx5_5:1/RoCE [3]mlx5_6:1/RoCE [4]mlx5_7:1/RoCE [5]mlx5_8:1/RoCE [6]mlx5_9:1/RoCE [7]mlx5_10:1/RoCE [RO]; OOB eth1:10.82.112.4<0>
+t-20260601014141-vst28-worker-0:10440:10533 [7] NCCL INFO PROFILER/Plugin: Could not find: libnccl-profiler.so.
+t-20260601014141-vst28-worker-0:10440:10533 [7] NCCL INFO Using network IBext_v9
+t-20260601014141-vst28-worker-0:10433:10531 [1] NCCL INFO NCCL_IB_PCI_RELAXED_ORDERING set by environment to 1.
+t-20260601014141-vst28-worker-0:10433:10531 [1] NCCL INFO NET/IB : Using [0]mlx5_1:1/RoCE [1]mlx5_4:1/RoCE [2]mlx5_5:1/RoCE [3]mlx5_6:1/RoCE [4]mlx5_7:1/RoCE [5]mlx5_8:1/RoCE [6]mlx5_9:1/RoCE [7]mlx5_10:1/RoCE [RO]; OOB eth1:10.82.112.4<0>
+t-20260601014141-vst28-worker-0:10436:10532 [4] NCCL INFO NCCL_IB_PCI_RELAXED_ORDERING set by environment to 1.
+t-20260601014141-vst28-worker-0:10436:10532 [4] NCCL INFO NET/IB : Using [0]mlx5_1:1/RoCE [1]mlx5_4:1/RoCE [2]mlx5_5:1/RoCE [3]mlx5_6:1/RoCE [4]mlx5_7:1/RoCE [5]mlx5_8:1/RoCE [6]mlx5_9:1/RoCE [7]mlx5_10:1/RoCE [RO]; OOB eth1:10.82.112.4<0>
+t-20260601014141-vst28-worker-0:10438:10529 [5] NCCL INFO NCCL_IB_PCI_RELAXED_ORDERING set by environment to 1.
+t-20260601014141-vst28-worker-0:10438:10529 [5] NCCL INFO NET/IB : Using [0]mlx5_1:1/RoCE [1]mlx5_4:1/RoCE [2]mlx5_5:1/RoCE [3]mlx5_6:1/RoCE [4]mlx5_7:1/RoCE [5]mlx5_8:1/RoCE [6]mlx5_9:1/RoCE [7]mlx5_10:1/RoCE [RO]; OOB eth1:10.82.112.4<0>
+t-20260601014141-vst28-worker-0:10434:10534 [2] NCCL INFO NCCL_IB_PCI_RELAXED_ORDERING set by environment to 1.
+t-20260601014141-vst28-worker-0:10434:10534 [2] NCCL INFO NET/IB : Using [0]mlx5_1:1/RoCE [1]mlx5_4:1/RoCE [2]mlx5_5:1/RoCE [3]mlx5_6:1/RoCE [4]mlx5_7:1/RoCE [5]mlx5_8:1/RoCE [6]mlx5_9:1/RoCE [7]mlx5_10:1/RoCE [RO]; OOB eth1:10.82.112.4<0>
+t-20260601014141-vst28-worker-0:10439:10530 [6] NCCL INFO NCCL_IB_PCI_RELAXED_ORDERING set by environment to 1.
+t-20260601014141-vst28-worker-0:10439:10530 [6] NCCL INFO NET/IB : Using [0]mlx5_1:1/RoCE [1]mlx5_4:1/RoCE [2]mlx5_5:1/RoCE [3]mlx5_6:1/RoCE [4]mlx5_7:1/RoCE [5]mlx5_8:1/RoCE [6]mlx5_9:1/RoCE [7]mlx5_10:1/RoCE [RO]; OOB eth1:10.82.112.4<0>
+t-20260601014141-vst28-worker-0:10433:10531 [1] NCCL INFO PROFILER/Plugin: Could not find: libnccl-profiler.so.
+t-20260601014141-vst28-worker-0:10436:10532 [4] NCCL INFO PROFILER/Plugin: Could not find: libnccl-profiler.so.
+t-20260601014141-vst28-worker-0:10436:10532 [4] NCCL INFO Using network IBext_v9
+t-20260601014141-vst28-worker-0:10433:10531 [1] NCCL INFO Using network IBext_v9
+t-20260601014141-vst28-worker-0:10435:10535 [3] NCCL INFO NCCL_IB_PCI_RELAXED_ORDERING set by environment to 1.
+t-20260601014141-vst28-worker-0:10435:10535 [3] NCCL INFO NET/IB : Using [0]mlx5_1:1/RoCE [1]mlx5_4:1/RoCE [2]mlx5_5:1/RoCE [3]mlx5_6:1/RoCE [4]mlx5_7:1/RoCE [5]mlx5_8:1/RoCE [6]mlx5_9:1/RoCE [7]mlx5_10:1/RoCE [RO]; OOB eth1:10.82.112.4<0>
+t-20260601014141-vst28-worker-0:10438:10529 [5] NCCL INFO PROFILER/Plugin: Could not find: libnccl-profiler.so.
+t-20260601014141-vst28-worker-0:10438:10529 [5] NCCL INFO Using network IBext_v9
+t-20260601014141-vst28-worker-0:10439:10530 [6] NCCL INFO PROFILER/Plugin: Could not find: libnccl-profiler.so.
+t-20260601014141-vst28-worker-0:10439:10530 [6] NCCL INFO Using network IBext_v9
+t-20260601014141-vst28-worker-0:10434:10534 [2] NCCL INFO PROFILER/Plugin: Could not find: libnccl-profiler.so.
+t-20260601014141-vst28-worker-0:10434:10534 [2] NCCL INFO Using network IBext_v9
+t-20260601014141-vst28-worker-0:10435:10535 [3] NCCL INFO PROFILER/Plugin: Could not find: libnccl-profiler.so.
+t-20260601014141-vst28-worker-0:10435:10535 [3] NCCL INFO Using network IBext_v9
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO ncclCommInitRankConfig comm 0x854acd0 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 65040 commId 0x4fb4c8e8f3932906 - Init START
+t-20260601014141-vst28-worker-0:10440:10533 [7] NCCL INFO ncclCommInitRankConfig comm 0xa65cbb0 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId 75020 commId 0x4fb4c8e8f3932906 - Init START
+t-20260601014141-vst28-worker-0:10438:10529 [5] NCCL INFO ncclCommInitRankConfig comm 0x8dc95b0 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 71020 commId 0x4fb4c8e8f3932906 - Init START
+t-20260601014141-vst28-worker-0:10434:10534 [2] NCCL INFO ncclCommInitRankConfig comm 0x9a70eb0 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 69020 commId 0x4fb4c8e8f3932906 - Init START
+t-20260601014141-vst28-worker-0:10435:10535 [3] NCCL INFO ncclCommInitRankConfig comm 0x9613d70 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 6b020 commId 0x4fb4c8e8f3932906 - Init START
+t-20260601014141-vst28-worker-0:10439:10530 [6] NCCL INFO ncclCommInitRankConfig comm 0xa3f72f0 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId 73020 commId 0x4fb4c8e8f3932906 - Init START
+t-20260601014141-vst28-worker-0:10439:10530 [6] NCCL INFO RAS client listening socket at ::1<28028>
+t-20260601014141-vst28-worker-0:10440:10533 [7] NCCL INFO RAS client listening socket at ::1<28028>
+t-20260601014141-vst28-worker-0:10433:10531 [1] NCCL INFO ncclCommInitRankConfig comm 0x99599b0 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 67020 commId 0x4fb4c8e8f3932906 - Init START
+t-20260601014141-vst28-worker-0:10434:10534 [2] NCCL INFO RAS client listening socket at ::1<28028>
+t-20260601014141-vst28-worker-0:10433:10531 [1] NCCL INFO RAS client listening socket at ::1<28028>
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO RAS client listening socket at ::1<28028>
+t-20260601014141-vst28-worker-0:10436:10532 [4] NCCL INFO ncclCommInitRankConfig comm 0x9001380 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 6f020 commId 0x4fb4c8e8f3932906 - Init START
+t-20260601014141-vst28-worker-0:10436:10532 [4] NCCL INFO RAS client listening socket at ::1<28028>
+t-20260601014141-vst28-worker-0:10438:10529 [5] NCCL INFO RAS client listening socket at ::1<28028>
+t-20260601014141-vst28-worker-0:10435:10535 [3] NCCL INFO RAS client listening socket at ::1<28028>
+t-20260601014141-vst28-worker-0:10433:10531 [1] NCCL INFO Bootstrap timings total 0.032309 (create 0.000021, send 0.000067, recv 0.000082, ring 0.031849, delay 0.000001)
+t-20260601014141-vst28-worker-0:10439:10530 [6] NCCL INFO Bootstrap timings total 0.033407 (create 0.000021, send 0.000079, recv 0.000123, ring 0.032897, delay 0.000001)
+t-20260601014141-vst28-worker-0:10436:10532 [4] NCCL INFO Bootstrap timings total 0.031744 (create 0.000021, send 0.000078, recv 0.000079, ring 0.031268, delay 0.000000)
+t-20260601014141-vst28-worker-0:10434:10534 [2] NCCL INFO Bootstrap timings total 0.034463 (create 0.000024, send 0.000067, recv 0.000677, ring 0.031845, delay 0.000001)
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO Bootstrap timings total 0.212386 (create 0.000027, send 0.000090, recv 0.180117, ring 0.000156, delay 0.000001)
+t-20260601014141-vst28-worker-0:10435:10535 [3] NCCL INFO Bootstrap timings total 0.033898 (create 0.000024, send 0.000067, recv 0.002213, ring 0.031241, delay 0.000001)
+t-20260601014141-vst28-worker-0:10438:10529 [5] NCCL INFO Bootstrap timings total 0.035490 (create 0.000021, send 0.000065, recv 0.002162, ring 0.031248, delay 0.000001)
+t-20260601014141-vst28-worker-0:10440:10533 [7] NCCL INFO Bootstrap timings total 0.059097 (create 0.000023, send 0.000064, recv 0.000098, ring 0.032863, delay 0.000001)
+t-20260601014141-vst28-worker-0:10433:10531 [1] NCCL INFO MNNVL busId 0x67020 fabric UUID 0.0 cliqueId 0x0 state 3 healthMask 0x0
+t-20260601014141-vst28-worker-0:10439:10530 [6] NCCL INFO MNNVL busId 0x73020 fabric UUID 0.0 cliqueId 0x0 state 3 healthMask 0x0
+t-20260601014141-vst28-worker-0:10436:10532 [4] NCCL INFO MNNVL busId 0x6f020 fabric UUID 0.0 cliqueId 0x0 state 3 healthMask 0x0
+t-20260601014141-vst28-worker-0:10434:10534 [2] NCCL INFO MNNVL busId 0x69020 fabric UUID 0.0 cliqueId 0x0 state 3 healthMask 0x0
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO MNNVL busId 0x65040 fabric UUID 0.0 cliqueId 0x0 state 3 healthMask 0x0
+t-20260601014141-vst28-worker-0:10438:10529 [5] NCCL INFO MNNVL busId 0x71020 fabric UUID 0.0 cliqueId 0x0 state 3 healthMask 0x0
+t-20260601014141-vst28-worker-0:10435:10535 [3] NCCL INFO MNNVL busId 0x6b020 fabric UUID 0.0 cliqueId 0x0 state 3 healthMask 0x0
+t-20260601014141-vst28-worker-0:10440:10533 [7] NCCL INFO MNNVL busId 0x75020 fabric UUID 0.0 cliqueId 0x0 state 3 healthMask 0x0
+t-20260601014141-vst28-worker-0:10435:10535 [3] NCCL INFO NCCL_TOPO_FILE set by environment to /var/run/nvidia-topologyd/virtualTopology.xml
+t-20260601014141-vst28-worker-0:10438:10529 [5] NCCL INFO NCCL_TOPO_FILE set by environment to /var/run/nvidia-topologyd/virtualTopology.xml
+t-20260601014141-vst28-worker-0:10433:10531 [1] NCCL INFO NCCL_TOPO_FILE set by environment to /var/run/nvidia-topologyd/virtualTopology.xml
+t-20260601014141-vst28-worker-0:10434:10534 [2] NCCL INFO NCCL_TOPO_FILE set by environment to /var/run/nvidia-topologyd/virtualTopology.xml
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO NCCL_TOPO_FILE set by environment to /var/run/nvidia-topologyd/virtualTopology.xml
+t-20260601014141-vst28-worker-0:10436:10532 [4] NCCL INFO NCCL_TOPO_FILE set by environment to /var/run/nvidia-topologyd/virtualTopology.xml
+t-20260601014141-vst28-worker-0:10439:10530 [6] NCCL INFO NCCL_TOPO_FILE set by environment to /var/run/nvidia-topologyd/virtualTopology.xml
+t-20260601014141-vst28-worker-0:10440:10533 [7] NCCL INFO NCCL_TOPO_FILE set by environment to /var/run/nvidia-topologyd/virtualTopology.xml
+t-20260601014141-vst28-worker-0:10438:10529 [5] NCCL INFO Setting affinity for GPU 5 to 0fffff,ffffffff,ffffffff,fc000000,00000000,00000000
+t-20260601014141-vst28-worker-0:10439:10530 [6] NCCL INFO Setting affinity for GPU 6 to 0fffff,ffffffff,ffffffff,fc000000,00000000,00000000
+t-20260601014141-vst28-worker-0:10438:10529 [5] NCCL INFO NVLS multicast support is available on dev 5
+t-20260601014141-vst28-worker-0:10434:10534 [2] NCCL INFO Setting affinity for GPU 2 to 03ffffff,ffffffff,ffffffff
+t-20260601014141-vst28-worker-0:10433:10531 [1] NCCL INFO Setting affinity for GPU 1 to 03ffffff,ffffffff,ffffffff
+t-20260601014141-vst28-worker-0:10434:10534 [2] NCCL INFO NVLS multicast support is available on dev 2
+t-20260601014141-vst28-worker-0:10436:10532 [4] NCCL INFO Setting affinity for GPU 4 to 0fffff,ffffffff,ffffffff,fc000000,00000000,00000000
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO Setting affinity for GPU 0 to 03ffffff,ffffffff,ffffffff
+t-20260601014141-vst28-worker-0:10436:10532 [4] NCCL INFO NVLS multicast support is available on dev 4
+t-20260601014141-vst28-worker-0:10439:10530 [6] NCCL INFO NVLS multicast support is available on dev 6
+t-20260601014141-vst28-worker-0:10435:10535 [3] NCCL INFO Setting affinity for GPU 3 to 03ffffff,ffffffff,ffffffff
+t-20260601014141-vst28-worker-0:10435:10535 [3] NCCL INFO NVLS multicast support is available on dev 3
+t-20260601014141-vst28-worker-0:10440:10533 [7] NCCL INFO Setting affinity for GPU 7 to 0fffff,ffffffff,ffffffff,fc000000,00000000,00000000
+t-20260601014141-vst28-worker-0:10440:10533 [7] NCCL INFO NVLS multicast support is available on dev 7
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO NVLS multicast support is available on dev 0
+t-20260601014141-vst28-worker-0:10433:10531 [1] NCCL INFO NVLS multicast support is available on dev 1
+t-20260601014141-vst28-worker-0:10439:10530 [6] NCCL INFO comm 0xa3f72f0 rank 6 nRanks 8 nNodes 1 localRanks 8 localRank 6 MNNVL 0
+t-20260601014141-vst28-worker-0:10440:10533 [7] NCCL INFO comm 0xa65cbb0 rank 7 nRanks 8 nNodes 1 localRanks 8 localRank 7 MNNVL 0
+t-20260601014141-vst28-worker-0:10438:10529 [5] NCCL INFO comm 0x8dc95b0 rank 5 nRanks 8 nNodes 1 localRanks 8 localRank 5 MNNVL 0
+t-20260601014141-vst28-worker-0:10436:10532 [4] NCCL INFO comm 0x9001380 rank 4 nRanks 8 nNodes 1 localRanks 8 localRank 4 MNNVL 0
+t-20260601014141-vst28-worker-0:10433:10531 [1] NCCL INFO comm 0x99599b0 rank 1 nRanks 8 nNodes 1 localRanks 8 localRank 1 MNNVL 0
+t-20260601014141-vst28-worker-0:10435:10535 [3] NCCL INFO comm 0x9613d70 rank 3 nRanks 8 nNodes 1 localRanks 8 localRank 3 MNNVL 0
+t-20260601014141-vst28-worker-0:10434:10534 [2] NCCL INFO comm 0x9a70eb0 rank 2 nRanks 8 nNodes 1 localRanks 8 localRank 2 MNNVL 0
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO comm 0x854acd0 rank 0 nRanks 8 nNodes 1 localRanks 8 localRank 0 MNNVL 0
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO Channel 00/24 : 0 1 2 3 4 5 6 7
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO Channel 01/24 : 0 1 2 3 4 5 6 7
+t-20260601014141-vst28-worker-0:10436:10532 [4] NCCL INFO Trees [0] 5/-1/-1->4->3 [1] 5/-1/-1->4->3 [2] 5/-1/-1->4->3 [3] 5/-1/-1->4->3 [4] 5/-1/-1->4->3 [5] 5/-1/-1->4->3 [6] 5/-1/-1->4->3 [7] 5/-1/-1->4->3 [8] 5/-1/-1->4->3 [9] 5/-1/-1->4->3 [10] 5/-1/-1->4->3 [11] 5/-1/-1->4->3 [12] 5/-1/-1->4->3 [13] 5/-1/-1->4->3 [14] 5/-1/-1->4->3 [15] 5/-1/-1->4->3 [16] 5/-1/-1->4->3 [17] 5/-1/-1->4->3 [18] 5/-1/-1->4->3 [19] 5/-1/-1->4->3 [20] 5/-1/-1->4->3 [21] 5/-1/-1->4->3 [22] 5/-1/-1->4->3 [23] 5/-1/-1->4->3
+t-20260601014141-vst28-worker-0:10435:10535 [3] NCCL INFO Trees [0] 4/-1/-1->3->2 [1] 4/-1/-1->3->2 [2] 4/-1/-1->3->2 [3] 4/-1/-1->3->2 [4] 4/-1/-1->3->2 [5] 4/-1/-1->3->2 [6] 4/-1/-1->3->2 [7] 4/-1/-1->3->2 [8] 4/-1/-1->3->2 [9] 4/-1/-1->3->2 [10] 4/-1/-1->3->2 [11] 4/-1/-1->3->2 [12] 4/-1/-1->3->2 [13] 4/-1/-1->3->2 [14] 4/-1/-1->3->2 [15] 4/-1/-1->3->2 [16] 4/-1/-1->3->2 [17] 4/-1/-1->3->2 [18] 4/-1/-1->3->2 [19] 4/-1/-1->3->2 [20] 4/-1/-1->3->2 [21] 4/-1/-1->3->2 [22] 4/-1/-1->3->2 [23] 4/-1/-1->3->2
+t-20260601014141-vst28-worker-0:10440:10533 [7] NCCL INFO Trees [0] -1/-1/-1->7->6 [1] -1/-1/-1->7->6 [2] -1/-1/-1->7->6 [3] -1/-1/-1->7->6 [4] -1/-1/-1->7->6 [5] -1/-1/-1->7->6 [6] -1/-1/-1->7->6 [7] -1/-1/-1->7->6 [8] -1/-1/-1->7->6 [9] -1/-1/-1->7->6 [10] -1/-1/-1->7->6 [11] -1/-1/-1->7->6 [12] -1/-1/-1->7->6 [13] -1/-1/-1->7->6 [14] -1/-1/-1->7->6 [15] -1/-1/-1->7->6 [16] -1/-1/-1->7->6 [17] -1/-1/-1->7->6 [18] -1/-1/-1->7->6 [19] -1/-1/-1->7->6 [20] -1/-1/-1->7->6 [21] -1/-1/-1->7->6 [22] -1/-1/-1->7->6 [23] -1/-1/-1->7->6
+t-20260601014141-vst28-worker-0:10439:10530 [6] NCCL INFO Trees [0] 7/-1/-1->6->5 [1] 7/-1/-1->6->5 [2] 7/-1/-1->6->5 [3] 7/-1/-1->6->5 [4] 7/-1/-1->6->5 [5] 7/-1/-1->6->5 [6] 7/-1/-1->6->5 [7] 7/-1/-1->6->5 [8] 7/-1/-1->6->5 [9] 7/-1/-1->6->5 [10] 7/-1/-1->6->5 [11] 7/-1/-1->6->5 [12] 7/-1/-1->6->5 [13] 7/-1/-1->6->5 [14] 7/-1/-1->6->5 [15] 7/-1/-1->6->5 [16] 7/-1/-1->6->5 [17] 7/-1/-1->6->5 [18] 7/-1/-1->6->5 [19] 7/-1/-1->6->5 [20] 7/-1/-1->6->5 [21] 7/-1/-1->6->5 [22] 7/-1/-1->6->5 [23] 7/-1/-1->6->5
+t-20260601014141-vst28-worker-0:10438:10529 [5] NCCL INFO Trees [0] 6/-1/-1->5->4 [1] 6/-1/-1->5->4 [2] 6/-1/-1->5->4 [3] 6/-1/-1->5->4 [4] 6/-1/-1->5->4 [5] 6/-1/-1->5->4 [6] 6/-1/-1->5->4 [7] 6/-1/-1->5->4 [8] 6/-1/-1->5->4 [9] 6/-1/-1->5->4 [10] 6/-1/-1->5->4 [11] 6/-1/-1->5->4 [12] 6/-1/-1->5->4 [13] 6/-1/-1->5->4 [14] 6/-1/-1->5->4 [15] 6/-1/-1->5->4 [16] 6/-1/-1->5->4 [17] 6/-1/-1->5->4 [18] 6/-1/-1->5->4 [19] 6/-1/-1->5->4 [20] 6/-1/-1->5->4 [21] 6/-1/-1->5->4 [22] 6/-1/-1->5->4 [23] 6/-1/-1->5->4
+t-20260601014141-vst28-worker-0:10440:10533 [7] NCCL INFO P2P Chunksize set to 524288
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO Channel 02/24 : 0 1 2 3 4 5 6 7
+t-20260601014141-vst28-worker-0:10436:10532 [4] NCCL INFO P2P Chunksize set to 524288
+t-20260601014141-vst28-worker-0:10433:10531 [1] NCCL INFO Trees [0] 2/-1/-1->1->0 [1] 2/-1/-1->1->0 [2] 2/-1/-1->1->0 [3] 2/-1/-1->1->0 [4] 2/-1/-1->1->0 [5] 2/-1/-1->1->0 [6] 2/-1/-1->1->0 [7] 2/-1/-1->1->0 [8] 2/-1/-1->1->0 [9] 2/-1/-1->1->0 [10] 2/-1/-1->1->0 [11] 2/-1/-1->1->0 [12] 2/-1/-1->1->0 [13] 2/-1/-1->1->0 [14] 2/-1/-1->1->0 [15] 2/-1/-1->1->0 [16] 2/-1/-1->1->0 [17] 2/-1/-1->1->0 [18] 2/-1/-1->1->0 [19] 2/-1/-1->1->0 [20] 2/-1/-1->1->0 [21] 2/-1/-1->1->0 [22] 2/-1/-1->1->0 [23] 2/-1/-1->1->0
+t-20260601014141-vst28-worker-0:10435:10535 [3] NCCL INFO P2P Chunksize set to 524288
+t-20260601014141-vst28-worker-0:10439:10530 [6] NCCL INFO P2P Chunksize set to 524288
+t-20260601014141-vst28-worker-0:10434:10534 [2] NCCL INFO Trees [0] 3/-1/-1->2->1 [1] 3/-1/-1->2->1 [2] 3/-1/-1->2->1 [3] 3/-1/-1->2->1 [4] 3/-1/-1->2->1 [5] 3/-1/-1->2->1 [6] 3/-1/-1->2->1 [7] 3/-1/-1->2->1 [8] 3/-1/-1->2->1 [9] 3/-1/-1->2->1 [10] 3/-1/-1->2->1 [11] 3/-1/-1->2->1 [12] 3/-1/-1->2->1 [13] 3/-1/-1->2->1 [14] 3/-1/-1->2->1 [15] 3/-1/-1->2->1 [16] 3/-1/-1->2->1 [17] 3/-1/-1->2->1 [18] 3/-1/-1->2->1 [19] 3/-1/-1->2->1 [20] 3/-1/-1->2->1 [21] 3/-1/-1->2->1 [22] 3/-1/-1->2->1 [23] 3/-1/-1->2->1
+t-20260601014141-vst28-worker-0:10438:10529 [5] NCCL INFO P2P Chunksize set to 524288
+t-20260601014141-vst28-worker-0:10434:10534 [2] NCCL INFO P2P Chunksize set to 524288
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO Channel 03/24 : 0 1 2 3 4 5 6 7
+t-20260601014141-vst28-worker-0:10433:10531 [1] NCCL INFO P2P Chunksize set to 524288
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO Channel 04/24 : 0 1 2 3 4 5 6 7
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO Channel 05/24 : 0 1 2 3 4 5 6 7
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO Channel 06/24 : 0 1 2 3 4 5 6 7
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO Channel 07/24 : 0 1 2 3 4 5 6 7
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO Channel 08/24 : 0 1 2 3 4 5 6 7
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO Channel 09/24 : 0 1 2 3 4 5 6 7
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO Channel 10/24 : 0 1 2 3 4 5 6 7
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO Channel 11/24 : 0 1 2 3 4 5 6 7
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO Channel 12/24 : 0 1 2 3 4 5 6 7
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO Channel 13/24 : 0 1 2 3 4 5 6 7
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO Channel 14/24 : 0 1 2 3 4 5 6 7
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO Channel 15/24 : 0 1 2 3 4 5 6 7
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO Channel 16/24 : 0 1 2 3 4 5 6 7
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO Channel 17/24 : 0 1 2 3 4 5 6 7
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO Channel 18/24 : 0 1 2 3 4 5 6 7
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO Channel 19/24 : 0 1 2 3 4 5 6 7
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO Channel 20/24 : 0 1 2 3 4 5 6 7
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO Channel 21/24 : 0 1 2 3 4 5 6 7
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO Channel 22/24 : 0 1 2 3 4 5 6 7
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO Channel 23/24 : 0 1 2 3 4 5 6 7
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO Trees [0] 1/-1/-1->0->-1 [1] 1/-1/-1->0->-1 [2] 1/-1/-1->0->-1 [3] 1/-1/-1->0->-1 [4] 1/-1/-1->0->-1 [5] 1/-1/-1->0->-1 [6] 1/-1/-1->0->-1 [7] 1/-1/-1->0->-1 [8] 1/-1/-1->0->-1 [9] 1/-1/-1->0->-1 [10] 1/-1/-1->0->-1 [11] 1/-1/-1->0->-1 [12] 1/-1/-1->0->-1 [13] 1/-1/-1->0->-1 [14] 1/-1/-1->0->-1 [15] 1/-1/-1->0->-1 [16] 1/-1/-1->0->-1 [17] 1/-1/-1->0->-1 [18] 1/-1/-1->0->-1 [19] 1/-1/-1->0->-1 [20] 1/-1/-1->0->-1 [21] 1/-1/-1->0->-1 [22] 1/-1/-1->0->-1 [23] 1/-1/-1->0->-1
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO P2P Chunksize set to 524288
+t-20260601014141-vst28-worker-0:10440:10632 [7] NCCL INFO [Proxy Service] Device 7 CPU core 167
+t-20260601014141-vst28-worker-0:10440:10633 [7] NCCL INFO [Proxy Service UDS] Device 7 CPU core 168
+t-20260601014141-vst28-worker-0:10439:10634 [6] NCCL INFO [Proxy Service] Device 6 CPU core 167
+t-20260601014141-vst28-worker-0:10439:10636 [6] NCCL INFO [Proxy Service UDS] Device 6 CPU core 168
+t-20260601014141-vst28-worker-0:10436:10635 [4] NCCL INFO [Proxy Service] Device 4 CPU core 112
+t-20260601014141-vst28-worker-0:10436:10637 [4] NCCL INFO [Proxy Service UDS] Device 4 CPU core 114
+t-20260601014141-vst28-worker-0:10438:10638 [5] NCCL INFO [Proxy Service] Device 5 CPU core 92
+t-20260601014141-vst28-worker-0:10438:10639 [5] NCCL INFO [Proxy Service UDS] Device 5 CPU core 94
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO Check P2P Type intraNodeP2pSupport 1 directMode 0
+t-20260601014141-vst28-worker-0:10433:10640 [1] NCCL INFO [Proxy Service] Device 1 CPU core 42
+t-20260601014141-vst28-worker-0:10433:10641 [1] NCCL INFO [Proxy Service UDS] Device 1 CPU core 44
+t-20260601014141-vst28-worker-0:10432:10642 [0] NCCL INFO [Proxy Service] Device 0 CPU core 2
+t-20260601014141-vst28-worker-0:10432:10643 [0] NCCL INFO [Proxy Service UDS] Device 0 CPU core 4
+t-20260601014141-vst28-worker-0:10435:10644 [3] NCCL INFO [Proxy Service] Device 3 CPU core 81
+t-20260601014141-vst28-worker-0:10435:10645 [3] NCCL INFO [Proxy Service UDS] Device 3 CPU core 82
+t-20260601014141-vst28-worker-0:10434:10646 [2] NCCL INFO [Proxy Service] Device 2 CPU core 2
+t-20260601014141-vst28-worker-0:10434:10647 [2] NCCL INFO [Proxy Service UDS] Device 2 CPU core 4
+t-20260601014141-vst28-worker-0:10439:10530 [6] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512
+t-20260601014141-vst28-worker-0:10439:10530 [6] NCCL INFO 24 coll channels, 24 collnet channels, 16 nvls channels, 32 p2p channels, 32 p2p channels per peer
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO 24 coll channels, 24 collnet channels, 16 nvls channels, 32 p2p channels, 32 p2p channels per peer
+t-20260601014141-vst28-worker-0:10435:10535 [3] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512
+t-20260601014141-vst28-worker-0:10435:10535 [3] NCCL INFO 24 coll channels, 24 collnet channels, 16 nvls channels, 32 p2p channels, 32 p2p channels per peer
+t-20260601014141-vst28-worker-0:10434:10534 [2] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512
+t-20260601014141-vst28-worker-0:10434:10534 [2] NCCL INFO 24 coll channels, 24 collnet channels, 16 nvls channels, 32 p2p channels, 32 p2p channels per peer
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO CC Off, workFifoBytes 1048576
+t-20260601014141-vst28-worker-0:10438:10529 [5] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512
+t-20260601014141-vst28-worker-0:10438:10529 [5] NCCL INFO 24 coll channels, 24 collnet channels, 16 nvls channels, 32 p2p channels, 32 p2p channels per peer
+t-20260601014141-vst28-worker-0:10440:10533 [7] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512
+t-20260601014141-vst28-worker-0:10440:10533 [7] NCCL INFO 24 coll channels, 24 collnet channels, 16 nvls channels, 32 p2p channels, 32 p2p channels per peer
+t-20260601014141-vst28-worker-0:10433:10531 [1] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512
+t-20260601014141-vst28-worker-0:10433:10531 [1] NCCL INFO 24 coll channels, 24 collnet channels, 16 nvls channels, 32 p2p channels, 32 p2p channels per peer
+t-20260601014141-vst28-worker-0:10436:10532 [4] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512
+t-20260601014141-vst28-worker-0:10436:10532 [4] NCCL INFO 24 coll channels, 24 collnet channels, 16 nvls channels, 32 p2p channels, 32 p2p channels per peer
+t-20260601014141-vst28-worker-0:10436:10532 [4] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v4 symbol.
+t-20260601014141-vst28-worker-0:10438:10529 [5] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v4 symbol.
+t-20260601014141-vst28-worker-0:10438:10529 [5] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v3 symbol.
+t-20260601014141-vst28-worker-0:10436:10532 [4] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v3 symbol.
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v4 symbol.
+t-20260601014141-vst28-worker-0:10438:10529 [5] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v2 symbol, using internal tuner instead.
+t-20260601014141-vst28-worker-0:10436:10532 [4] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v2 symbol, using internal tuner instead.
+t-20260601014141-vst28-worker-0:10435:10535 [3] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v4 symbol.
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v3 symbol.
+t-20260601014141-vst28-worker-0:10438:10529 [5] NCCL INFO ncclCommInitRankConfig comm 0x8dc95b0 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 71020 commId 0x4fb4c8e8f3932906 - Init COMPLETE
+t-20260601014141-vst28-worker-0:10440:10533 [7] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v4 symbol.
+t-20260601014141-vst28-worker-0:10439:10530 [6] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v4 symbol.
+t-20260601014141-vst28-worker-0:10436:10532 [4] NCCL INFO ncclCommInitRankConfig comm 0x9001380 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 6f020 commId 0x4fb4c8e8f3932906 - Init COMPLETE
+t-20260601014141-vst28-worker-0:10435:10535 [3] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v3 symbol.
+t-20260601014141-vst28-worker-0:10439:10530 [6] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v3 symbol.
+t-20260601014141-vst28-worker-0:10433:10531 [1] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v4 symbol.
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v2 symbol, using internal tuner instead.
+t-20260601014141-vst28-worker-0:10434:10534 [2] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v4 symbol.
+t-20260601014141-vst28-worker-0:10440:10533 [7] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v3 symbol.
+t-20260601014141-vst28-worker-0:10438:10529 [5] NCCL INFO Init timings - ncclCommInitRankConfig: rank 5 nranks 8 total 2.13 (kernels 0.20, alloc 1.07, bootstrap 0.04, allgathers 0.01, topo 0.54, graphs 0.01, connections 0.24, rest 0.02)
+t-20260601014141-vst28-worker-0:10435:10535 [3] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v2 symbol, using internal tuner instead.
+t-20260601014141-vst28-worker-0:10439:10530 [6] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v2 symbol, using internal tuner instead.
+t-20260601014141-vst28-worker-0:10436:10532 [4] NCCL INFO Init timings - ncclCommInitRankConfig: rank 4 nranks 8 total 2.13 (kernels 0.19, alloc 1.08, bootstrap 0.03, allgathers 0.01, topo 0.54, graphs 0.01, connections 0.24, rest 0.02)
+t-20260601014141-vst28-worker-0:10435:10535 [3] NCCL INFO ncclCommInitRankConfig comm 0x9613d70 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 6b020 commId 0x4fb4c8e8f3932906 - Init COMPLETE
+t-20260601014141-vst28-worker-0:10433:10531 [1] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v3 symbol.
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO ncclCommInitRankConfig comm 0x854acd0 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 65040 commId 0x4fb4c8e8f3932906 - Init COMPLETE
+t-20260601014141-vst28-worker-0:10434:10534 [2] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v3 symbol.
+t-20260601014141-vst28-worker-0:10433:10531 [1] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v2 symbol, using internal tuner instead.
+t-20260601014141-vst28-worker-0:10440:10533 [7] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v2 symbol, using internal tuner instead.
+t-20260601014141-vst28-worker-0:10439:10530 [6] NCCL INFO ncclCommInitRankConfig comm 0xa3f72f0 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId 73020 commId 0x4fb4c8e8f3932906 - Init COMPLETE
+t-20260601014141-vst28-worker-0:10434:10534 [2] NCCL INFO TUNER/Plugin: Failed to find ncclTunerPlugin_v2 symbol, using internal tuner instead.
+t-20260601014141-vst28-worker-0:10433:10531 [1] NCCL INFO ncclCommInitRankConfig comm 0x99599b0 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 67020 commId 0x4fb4c8e8f3932906 - Init COMPLETE
+t-20260601014141-vst28-worker-0:10435:10535 [3] NCCL INFO Init timings - ncclCommInitRankConfig: rank 3 nranks 8 total 2.13 (kernels 0.19, alloc 1.08, bootstrap 0.03, allgathers 0.01, topo 0.54, graphs 0.01, connections 0.24, rest 0.03)
+t-20260601014141-vst28-worker-0:10432:10528 [0] NCCL INFO Init timings - ncclCommInitRankConfig: rank 0 nranks 8 total 2.14 (kernels 0.20, alloc 0.91, bootstrap 0.21, allgathers 0.00, topo 0.54, graphs 0.01, connections 0.24, rest 0.03)
+t-20260601014141-vst28-worker-0:10439:10530 [6] NCCL INFO Init timings - ncclCommInitRankConfig: rank 6 nranks 8 total 2.13 (kernels 0.19, alloc 1.08, bootstrap 0.03, allgathers 0.01, topo 0.54, graphs 0.01, connections 0.24, rest 0.02)
+t-20260601014141-vst28-worker-0:10440:10533 [7] NCCL INFO ncclCommInitRankConfig comm 0xa65cbb0 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId 75020 commId 0x4fb4c8e8f3932906 - Init COMPLETE
+t-20260601014141-vst28-worker-0:10434:10534 [2] NCCL INFO ncclCommInitRankConfig comm 0x9a70eb0 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 69020 commId 0x4fb4c8e8f3932906 - Init COMPLETE
+t-20260601014141-vst28-worker-0:10433:10531 [1] NCCL INFO Init timings - ncclCommInitRankConfig: rank 1 nranks 8 total 2.13 (kernels 0.19, alloc 1.08, bootstrap 0.03, allgathers 0.00, topo 0.54, graphs 0.01, connections 0.24, rest 0.02)
+t-20260601014141-vst28-worker-0:10440:10533 [7] NCCL INFO Init timings - ncclCommInitRankConfig: rank 7 nranks 8 total 2.13 (kernels 0.19, alloc 1.05, bootstrap 0.06, allgathers 0.01, topo 0.54, graphs 0.01, connections 0.24, rest 0.02)
+t-20260601014141-vst28-worker-0:10434:10534 [2] NCCL INFO Init timings - ncclCommInitRankConfig: rank 2 nranks 8 total 2.13 (kernels 0.19, alloc 1.08, bootstrap 0.03, allgathers 0.00, topo 0.54, graphs 0.01, connections 0.24, rest 0.03)
+t-20260601014141-vst28-worker-0:10436:10650 [4] NCCL INFO Channel 00/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10440:10651 [7] NCCL INFO Channel 00/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10432:10654 [0] NCCL INFO Channel 00/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10440:10651 [7] NCCL INFO Channel 01/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10432:10654 [0] NCCL INFO Channel 01/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10440:10651 [7] NCCL INFO Channel 02/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10432:10654 [0] NCCL INFO Channel 02/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10440:10651 [7] NCCL INFO Channel 03/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10432:10654 [0] NCCL INFO Channel 03/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10440:10651 [7] NCCL INFO Channel 04/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10432:10654 [0] NCCL INFO Channel 04/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10440:10651 [7] NCCL INFO Channel 05/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10436:10650 [4] NCCL INFO Channel 01/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10432:10654 [0] NCCL INFO Channel 05/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10433:10648 [1] NCCL INFO Channel 00/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10440:10651 [7] NCCL INFO Channel 06/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10436:10650 [4] NCCL INFO Channel 02/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10433:10648 [1] NCCL INFO Channel 01/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10433:10648 [1] NCCL INFO Channel 02/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10433:10648 [1] NCCL INFO Channel 03/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10433:10648 [1] NCCL INFO Channel 04/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10433:10648 [1] NCCL INFO Channel 05/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10438:10653 [5] NCCL INFO Channel 00/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10433:10648 [1] NCCL INFO Channel 06/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10438:10653 [5] NCCL INFO Channel 01/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10433:10648 [1] NCCL INFO Channel 07/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10438:10653 [5] NCCL INFO Channel 02/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10433:10648 [1] NCCL INFO Channel 08/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10439:10652 [6] NCCL INFO Channel 00/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10438:10653 [5] NCCL INFO Channel 03/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10433:10648 [1] NCCL INFO Channel 09/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10439:10652 [6] NCCL INFO Channel 01/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10438:10653 [5] NCCL INFO Channel 04/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10433:10648 [1] NCCL INFO Channel 10/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10434:10649 [2] NCCL INFO Channel 00/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10439:10652 [6] NCCL INFO Channel 02/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10432:10654 [0] NCCL INFO Channel 06/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10438:10653 [5] NCCL INFO Channel 05/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10433:10648 [1] NCCL INFO Channel 11/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10440:10651 [7] NCCL INFO Channel 07/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10434:10649 [2] NCCL INFO Channel 01/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10436:10650 [4] NCCL INFO Channel 03/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10439:10652 [6] NCCL INFO Channel 03/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10432:10654 [0] NCCL INFO Channel 07/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10438:10653 [5] NCCL INFO Channel 06/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10433:10648 [1] NCCL INFO Channel 12/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10440:10651 [7] NCCL INFO Channel 08/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10434:10649 [2] NCCL INFO Channel 02/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10436:10650 [4] NCCL INFO Channel 04/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10439:10652 [6] NCCL INFO Channel 04/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10432:10654 [0] NCCL INFO Channel 08/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10438:10653 [5] NCCL INFO Channel 07/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10433:10648 [1] NCCL INFO Channel 13/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10440:10651 [7] NCCL INFO Channel 09/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10434:10649 [2] NCCL INFO Channel 03/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10436:10650 [4] NCCL INFO Channel 05/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10439:10652 [6] NCCL INFO Channel 05/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10432:10654 [0] NCCL INFO Channel 09/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10438:10653 [5] NCCL INFO Channel 08/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10433:10648 [1] NCCL INFO Channel 14/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10440:10651 [7] NCCL INFO Channel 10/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10434:10649 [2] NCCL INFO Channel 04/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10436:10650 [4] NCCL INFO Channel 06/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10439:10652 [6] NCCL INFO Channel 06/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10432:10654 [0] NCCL INFO Channel 10/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10438:10653 [5] NCCL INFO Channel 09/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10433:10648 [1] NCCL INFO Channel 15/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10440:10651 [7] NCCL INFO Channel 11/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10434:10649 [2] NCCL INFO Channel 05/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10436:10650 [4] NCCL INFO Channel 07/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10439:10652 [6] NCCL INFO Channel 07/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10432:10654 [0] NCCL INFO Channel 11/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10438:10653 [5] NCCL INFO Channel 10/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10433:10648 [1] NCCL INFO Channel 16/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10440:10651 [7] NCCL INFO Channel 12/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10434:10649 [2] NCCL INFO Channel 06/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10436:10650 [4] NCCL INFO Channel 08/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10439:10652 [6] NCCL INFO Channel 08/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10432:10654 [0] NCCL INFO Channel 12/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10438:10653 [5] NCCL INFO Channel 11/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10432:10654 [0] NCCL INFO Channel 13/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10432:10654 [0] NCCL INFO Channel 14/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10432:10654 [0] NCCL INFO Channel 15/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10432:10654 [0] NCCL INFO Channel 16/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10432:10654 [0] NCCL INFO Channel 17/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10432:10654 [0] NCCL INFO Channel 18/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10432:10654 [0] NCCL INFO Channel 19/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10432:10654 [0] NCCL INFO Channel 20/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10432:10654 [0] NCCL INFO Channel 21/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10432:10654 [0] NCCL INFO Channel 22/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10432:10654 [0] NCCL INFO Channel 23/0 : 0[0] -> 1[1] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10439:10652 [6] NCCL INFO Channel 09/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10440:10651 [7] NCCL INFO Channel 13/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10434:10649 [2] NCCL INFO Channel 07/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10438:10653 [5] NCCL INFO Channel 12/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10440:10651 [7] NCCL INFO Channel 14/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10434:10649 [2] NCCL INFO Channel 08/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10438:10653 [5] NCCL INFO Channel 13/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10440:10651 [7] NCCL INFO Channel 15/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10434:10649 [2] NCCL INFO Channel 09/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10433:10648 [1] NCCL INFO Channel 17/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10438:10653 [5] NCCL INFO Channel 14/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10440:10651 [7] NCCL INFO Channel 16/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10436:10650 [4] NCCL INFO Channel 09/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10434:10649 [2] NCCL INFO Channel 10/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10438:10653 [5] NCCL INFO Channel 15/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10433:10648 [1] NCCL INFO Channel 18/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10440:10651 [7] NCCL INFO Channel 17/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10436:10650 [4] NCCL INFO Channel 10/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10434:10649 [2] NCCL INFO Channel 11/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10438:10653 [5] NCCL INFO Channel 16/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10433:10648 [1] NCCL INFO Channel 19/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10440:10651 [7] NCCL INFO Channel 18/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10436:10650 [4] NCCL INFO Channel 11/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10434:10649 [2] NCCL INFO Channel 12/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10438:10653 [5] NCCL INFO Channel 17/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10433:10648 [1] NCCL INFO Channel 20/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10440:10651 [7] NCCL INFO Channel 19/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10436:10650 [4] NCCL INFO Channel 12/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10434:10649 [2] NCCL INFO Channel 13/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10439:10652 [6] NCCL INFO Channel 10/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10438:10653 [5] NCCL INFO Channel 18/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10433:10648 [1] NCCL INFO Channel 21/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10440:10651 [7] NCCL INFO Channel 20/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10436:10650 [4] NCCL INFO Channel 13/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10434:10649 [2] NCCL INFO Channel 14/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10439:10652 [6] NCCL INFO Channel 11/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10438:10653 [5] NCCL INFO Channel 19/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10433:10648 [1] NCCL INFO Channel 22/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10440:10651 [7] NCCL INFO Channel 21/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10436:10650 [4] NCCL INFO Channel 14/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10434:10649 [2] NCCL INFO Channel 15/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10439:10652 [6] NCCL INFO Channel 12/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10438:10653 [5] NCCL INFO Channel 20/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10433:10648 [1] NCCL INFO Channel 23/0 : 1[1] -> 2[2] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10440:10651 [7] NCCL INFO Channel 22/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10436:10650 [4] NCCL INFO Channel 15/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10434:10649 [2] NCCL INFO Channel 16/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10439:10652 [6] NCCL INFO Channel 13/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10438:10653 [5] NCCL INFO Channel 21/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10440:10651 [7] NCCL INFO Channel 23/0 : 7[7] -> 0[0] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10436:10650 [4] NCCL INFO Channel 16/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10434:10649 [2] NCCL INFO Channel 17/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10439:10652 [6] NCCL INFO Channel 14/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10438:10653 [5] NCCL INFO Channel 22/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10436:10650 [4] NCCL INFO Channel 17/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10434:10649 [2] NCCL INFO Channel 18/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10439:10652 [6] NCCL INFO Channel 15/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10438:10653 [5] NCCL INFO Channel 23/0 : 5[5] -> 6[6] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10439:10652 [6] NCCL INFO Channel 16/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10436:10650 [4] NCCL INFO Channel 18/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10439:10652 [6] NCCL INFO Channel 17/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10434:10649 [2] NCCL INFO Channel 19/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10439:10652 [6] NCCL INFO Channel 18/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10434:10649 [2] NCCL INFO Channel 20/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10435:10655 [3] NCCL INFO Channel 00/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10439:10652 [6] NCCL INFO Channel 19/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10434:10649 [2] NCCL INFO Channel 21/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10435:10655 [3] NCCL INFO Channel 01/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10439:10652 [6] NCCL INFO Channel 20/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10434:10649 [2] NCCL INFO Channel 22/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10435:10655 [3] NCCL INFO Channel 02/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10439:10652 [6] NCCL INFO Channel 21/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10434:10649 [2] NCCL INFO Channel 23/0 : 2[2] -> 3[3] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10435:10655 [3] NCCL INFO Channel 03/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10439:10652 [6] NCCL INFO Channel 22/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10435:10655 [3] NCCL INFO Channel 04/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10439:10652 [6] NCCL INFO Channel 23/0 : 6[6] -> 7[7] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10436:10650 [4] NCCL INFO Channel 19/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10435:10655 [3] NCCL INFO Channel 05/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10436:10650 [4] NCCL INFO Channel 20/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10435:10655 [3] NCCL INFO Channel 06/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10436:10650 [4] NCCL INFO Channel 21/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10435:10655 [3] NCCL INFO Channel 07/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10435:10655 [3] NCCL INFO Channel 08/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10436:10650 [4] NCCL INFO Channel 22/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10435:10655 [3] NCCL INFO Channel 09/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10436:10650 [4] NCCL INFO Channel 23/0 : 4[4] -> 5[5] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10435:10655 [3] NCCL INFO Channel 10/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10435:10655 [3] NCCL INFO Channel 11/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10435:10655 [3] NCCL INFO Channel 12/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10435:10655 [3] NCCL INFO Channel 13/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10435:10655 [3] NCCL INFO Channel 14/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10435:10655 [3] NCCL INFO Channel 15/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10435:10655 [3] NCCL INFO Channel 16/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10435:10655 [3] NCCL INFO Channel 17/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10435:10655 [3] NCCL INFO Channel 18/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10435:10655 [3] NCCL INFO Channel 19/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10435:10655 [3] NCCL INFO Channel 20/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10435:10655 [3] NCCL INFO Channel 21/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10435:10655 [3] NCCL INFO Channel 22/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10435:10655 [3] NCCL INFO Channel 23/0 : 3[3] -> 4[4] via P2P/CUMEM
+t-20260601014141-vst28-worker-0:10434:10649 [2] NCCL INFO Connected all rings, use ring PXN 0 GDR 1
+t-20260601014141-vst28-worker-0:10435:10655 [3] NCCL INFO Connected all rings, use ring PXN 0 GDR 1
+t-20260601014141-vst28-worker-0:10439:10652 [6] NCCL INFO Connected all rings, use ring PXN 0 GDR 1
+t-20260601014141-vst28-worker-0:10433:10648 [1] NCCL INFO Connected all rings, use ring PXN 0 GDR 1
+t-20260601014141-vst28-worker-0:10438:10653 [5] NCCL INFO Connected all rings, use ring PXN 0 GDR 1
+t-20260601014141-vst28-worker-0:10440:10651 [7] NCCL INFO Connected all rings, use ring PXN 0 GDR 1
+t-20260601014141-vst28-worker-0:10436:10650 [4] NCCL INFO Connected all rings, use ring PXN 0 GDR 1
+t-20260601014141-vst28-worker-0:10432:10654 [0] NCCL INFO Connected all rings, use ring PXN 0 GDR 1
+[rank0]:[W531 17:42:33.924682541 Utils.hpp:111] Warning: Environment variable NCCL_ASYNC_ERROR_HANDLING is deprecated; use TORCH_NCCL_ASYNC_ERROR_HANDLING instead (function operator())
+[rank1]:[W531 17:42:33.924814203 Utils.hpp:111] Warning: Environment variable NCCL_ASYNC_ERROR_HANDLING is deprecated; use TORCH_NCCL_ASYNC_ERROR_HANDLING instead (function operator())
+[rank6]:[W531 17:42:33.924869513 Utils.hpp:111] Warning: Environment variable NCCL_ASYNC_ERROR_HANDLING is deprecated; use TORCH_NCCL_ASYNC_ERROR_HANDLING instead (function operator())
+[rank2]:[W531 17:42:33.924888140 Utils.hpp:111] Warning: Environment variable NCCL_ASYNC_ERROR_HANDLING is deprecated; use TORCH_NCCL_ASYNC_ERROR_HANDLING instead (function operator())
+[rank3]:[W531 17:42:33.925175467 Utils.hpp:111] Warning: Environment variable NCCL_ASYNC_ERROR_HANDLING is deprecated; use TORCH_NCCL_ASYNC_ERROR_HANDLING instead (function operator())
+[rank4]:[W531 17:42:33.925186926 Utils.hpp:111] Warning: Environment variable NCCL_ASYNC_ERROR_HANDLING is deprecated; use TORCH_NCCL_ASYNC_ERROR_HANDLING instead (function operator())
+[rank7]:[W531 17:42:33.925236434 Utils.hpp:111] Warning: Environment variable NCCL_ASYNC_ERROR_HANDLING is deprecated; use TORCH_NCCL_ASYNC_ERROR_HANDLING instead (function operator())
+[rank5]:[W531 17:42:33.925258143 Utils.hpp:111] Warning: Environment variable NCCL_ASYNC_ERROR_HANDLING is deprecated; use TORCH_NCCL_ASYNC_ERROR_HANDLING instead (function operator())
+Muon: 58 2D params; Nesterov-AdamW: 78 other paramsMuon: 58 2D params; Nesterov-AdamW: 78 other paramsMuon: 58 2D params; Nesterov-AdamW: 78 other params
+
+
+Muon: 58 2D params; Nesterov-AdamW: 78 other params
+{
+  "data_mode": "elf_hfds",
+  "cache_path": "cache/owt_t5_payload1022_appendeos1.pt",
+  "data_path": "/e2e-data/evad-tech-vla/wanghan58/data/embedded-language-flows/openwebtext-t5",
+  "tokenizer_path": "/e2e-data/evad-tech-vla/wanghan58/models/hf/t5-small/tokenizer.json",
+  "text_column": "text",
+  "pack_len": 1023,
+  "append_eos": 1,
+  "num_workers": 0,
+  "shuffle_buffer": 8192,
+  "reject_txt": "cache/online_rejected.txt",
+  "out_dir": "runs/owt_t5_elftokenized_full_len1024_C1_to_1024_pow1_d768_l12_h12_gbs512_8gpu_5ep_elfopt_t5embed_unfixed_probadd_selfcond_ce_20260531_174225",
+  "subset_size": 0,
+  "resume": "",
+  "steps": 95090,
+  "batch_size": 16,
+  "grad_accum": 4,
+  "lr": 0.002,
+  "blr": 0.001,
+  "min_lr": 0.0,
+  "lr_schedule": "constant",
+  "warmup_steps": 9,
+  "warmup_epochs": -1.0,
+  "optimizer": "muon",
+  "weight_decay": 0.0,
+  "adam_beta1": 0.9,
+  "adam_beta2": 0.95,
+  "adam_eps": 1e-08,
+  "grad_clip": 1.0,
+  "log_every": 50,
+  "save_every": 1000,
+  "dim": 768,
+  "layers": 12,
+  "heads": 12,
+  "mlp_dim": 3072,
+  "time_tokens": 4,
+  "token_embed_init": "t5_shared",
+  "token_embed_model_path": "/e2e-data/evad-tech-vla/wanghan58/models/hf/t5-small",
+  "token_embed_dim": 512,
+  "freeze_token_embed": 0,
+  "c_min": 1.0,
+  "c_max": 1024.0,
+  "c_schedule": "exp",
+  "concat_self_cond": 0,
+  "self_cond_ce": 1,
+  "self_cond_use_cfg_scale": 0,
+  "self_cond_input_cfg": 0,
+  "self_cond_cfg_distill": 0,
+  "self_cond_prob_add": 1,
+  "self_cond_prob_add_scale": 1.0,
+  "self_cond_prob": 0.5,
+  "self_cond_cfg_min": 0.5,
+  "self_cond_cfg_max": 5.0,
+  "self_cond_cfg_formula": "elf",
+  "self_cond_kl_weight": 0.05,
+  "self_cond_kl_warmup": 1000,
+  "self_cond_teacher_temp": 1.0,
+  "seed": 1234,
+  "loader_batches_per_rank": 0,
+  "optimizer_steps_per_epoch": 0,
+  "steps_per_epoch": 0,
+  "effective_batch_size": 512
+}
+[data] mode=elf_hfds rows=elf_hfds length=1024 vocab=32100 seen=0 dropped=0 bos=1:</s> eos=1:</s>
+[optim] optimizer=muon lr=2.000000e-03 blr=1.000000e-03 effective_batch=512 warmup_steps=9 lr_schedule=constant wd=0.0 loader_batches_per_rank=0 optimizer_steps_per_epoch=0
+[embed] init=t5_shared dim=512 freeze=False model_path=/e2e-data/evad-tech-vla/wanghan58/models/hf/t5-small
+[self_cond] ce=True use_cfg_scale=False input_cfg=False distill=False prob_add=True prob_add_scale=1.0 concat=False prob=0.5 scale=[0.5,5.0] formula=elf kl_weight=0.05 kl_warmup=1000 teacher_temp=1.0
+Muon: 58 2D params; Nesterov-AdamW: 78 other params
+Muon: 58 2D params; Nesterov-AdamW: 78 other params
+Muon: 58 2D params; Nesterov-AdamW: 78 other params
+Muon: 58 2D params; Nesterov-AdamW: 78 other params
+[rank6]: Traceback (most recent call last):
+[rank6]:   File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/mini_owt_logdirichlet/train.py", line 347, in main
+[rank6]:     ids = next(it)
+[rank6]:           ^^^^^^^^
+[rank6]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/dataloader.py", line 708, in __next__
+[rank6]:     data = self._next_data()
+[rank6]:            ^^^^^^^^^^^^^^^^^
+[rank6]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/dataloader.py", line 764, in _next_data
+[rank6]:     data = self._dataset_fetcher.fetch(index)  # may raise StopIteration
+[rank6]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank6]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/_utils/fetch.py", line 40, in fetch
+[rank6]:     raise StopIteration
+[rank6]: StopIteration
+
+[rank6]: During handling of the above exception, another exception occurred:
+
+[rank6]: Traceback (most recent call last):
+[rank6]:   File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/mini_owt_logdirichlet/train.py", line 473, in <module>
+[rank6]:     main()
+[rank6]:   File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/mini_owt_logdirichlet/train.py", line 352, in main
+[rank6]:     ids = next(it)
+[rank6]:           ^^^^^^^^
+[rank6]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/dataloader.py", line 708, in __next__
+[rank6]:     data = self._next_data()
+[rank6]:            ^^^^^^^^^^^^^^^^^
+[rank6]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/dataloader.py", line 764, in _next_data
+[rank6]:     data = self._dataset_fetcher.fetch(index)  # may raise StopIteration
+[rank6]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank6]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/_utils/fetch.py", line 40, in fetch
+[rank6]:     raise StopIteration
+[rank6]: StopIteration
+[rank2]: Traceback (most recent call last):
+[rank2]:   File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/mini_owt_logdirichlet/train.py", line 347, in main
+[rank2]:     ids = next(it)
+[rank2]:           ^^^^^^^^
+[rank2]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/dataloader.py", line 708, in __next__
+[rank2]:     data = self._next_data()
+[rank2]:            ^^^^^^^^^^^^^^^^^
+[rank2]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/dataloader.py", line 764, in _next_data
+[rank2]:     data = self._dataset_fetcher.fetch(index)  # may raise StopIteration
+[rank2]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank2]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/_utils/fetch.py", line 40, in fetch
+[rank2]:     raise StopIteration
+[rank2]: StopIteration
+
+[rank2]: During handling of the above exception, another exception occurred:
+
+[rank2]: Traceback (most recent call last):
+[rank2]:   File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/mini_owt_logdirichlet/train.py", line 473, in <module>
+[rank2]:     main()
+[rank2]:   File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/mini_owt_logdirichlet/train.py", line 352, in main
+[rank2]:     ids = next(it)
+[rank2]:           ^^^^^^^^
+[rank2]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/dataloader.py", line 708, in __next__
+[rank2]:     data = self._next_data()
+[rank2]:            ^^^^^^^^^^^^^^^^^
+[rank2]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/dataloader.py", line 764, in _next_data
+[rank2]:     data = self._dataset_fetcher.fetch(index)  # may raise StopIteration
+[rank2]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank2]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/_utils/fetch.py", line 40, in fetch
+[rank2]:     raise StopIteration
+[rank2]: StopIteration
+[rank1]: Traceback (most recent call last):
+[rank1]:   File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/mini_owt_logdirichlet/train.py", line 347, in main
+[rank1]:     ids = next(it)
+[rank1]:           ^^^^^^^^
+[rank1]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/dataloader.py", line 708, in __next__
+[rank1]:     data = self._next_data()
+[rank1]:            ^^^^^^^^^^^^^^^^^
+[rank1]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/dataloader.py", line 764, in _next_data
+[rank1]:     data = self._dataset_fetcher.fetch(index)  # may raise StopIteration
+[rank1]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank1]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/_utils/fetch.py", line 40, in fetch
+[rank1]:     raise StopIteration
+[rank1]: StopIteration
+
+[rank1]: During handling of the above exception, another exception occurred:
+
+[rank1]: Traceback (most recent call last):
+[rank1]:   File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/mini_owt_logdirichlet/train.py", line 473, in <module>
+[rank1]:     main()
+[rank1]:   File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/mini_owt_logdirichlet/train.py", line 352, in main
+[rank1]:     ids = next(it)
+[rank1]:           ^^^^^^^^
+[rank1]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/dataloader.py", line 708, in __next__
+[rank1]:     data = self._next_data()
+[rank1]:            ^^^^^^^^^^^^^^^^^
+[rank1]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/dataloader.py", line 764, in _next_data
+[rank1]:     data = self._dataset_fetcher.fetch(index)  # may raise StopIteration
+[rank1]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank1]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/_utils/fetch.py", line 40, in fetch
+[rank1]:     raise StopIteration
+[rank1]: StopIteration
+[rank3]: Traceback (most recent call last):
+[rank3]:   File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/mini_owt_logdirichlet/train.py", line 347, in main
+[rank3]:     ids = next(it)
+[rank3]:           ^^^^^^^^
+[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/dataloader.py", line 708, in __next__
+[rank3]:     data = self._next_data()
+[rank3]:            ^^^^^^^^^^^^^^^^^
+[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/dataloader.py", line 764, in _next_data
+[rank3]:     data = self._dataset_fetcher.fetch(index)  # may raise StopIteration
+[rank3]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/_utils/fetch.py", line 40, in fetch
+[rank3]:     raise StopIteration
+[rank3]: StopIteration
+
+[rank3]: During handling of the above exception, another exception occurred:
+
+[rank3]: Traceback (most recent call last):
+[rank3]:   File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/mini_owt_logdirichlet/train.py", line 473, in <module>
+[rank3]:     main()
+[rank3]:   File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/mini_owt_logdirichlet/train.py", line 352, in main
+[rank3]:     ids = next(it)
+[rank3]:           ^^^^^^^^
+[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/dataloader.py", line 708, in __next__
+[rank3]:     data = self._next_data()
+[rank3]:            ^^^^^^^^^^^^^^^^^
+[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/dataloader.py", line 764, in _next_data
+[rank3]:     data = self._dataset_fetcher.fetch(index)  # may raise StopIteration
+[rank3]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/_utils/fetch.py", line 40, in fetch
+[rank3]:     raise StopIteration
+[rank3]: StopIteration
+[rank7]: Traceback (most recent call last):
+[rank7]:   File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/mini_owt_logdirichlet/train.py", line 347, in main
+[rank7]:     ids = next(it)
+[rank7]:           ^^^^^^^^
+[rank7]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/dataloader.py", line 708, in __next__
+[rank7]:     data = self._next_data()
+[rank7]:            ^^^^^^^^^^^^^^^^^
+[rank7]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/dataloader.py", line 764, in _next_data
+[rank7]:     data = self._dataset_fetcher.fetch(index)  # may raise StopIteration
+[rank7]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank7]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/_utils/fetch.py", line 40, in fetch
+[rank7]:     raise StopIteration
+[rank7]: StopIteration
+
+[rank7]: During handling of the above exception, another exception occurred:
+
+[rank7]: Traceback (most recent call last):
+[rank7]:   File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/mini_owt_logdirichlet/train.py", line 473, in <module>
+[rank7]:     main()
+[rank7]:   File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/mini_owt_logdirichlet/train.py", line 352, in main
+[rank7]:     ids = next(it)
+[rank7]:           ^^^^^^^^
+[rank7]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/dataloader.py", line 708, in __next__
+[rank7]:     data = self._next_data()
+[rank7]:            ^^^^^^^^^^^^^^^^^
+[rank7]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/dataloader.py", line 764, in _next_data
+[rank7]:     data = self._dataset_fetcher.fetch(index)  # may raise StopIteration
+[rank7]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank7]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/_utils/fetch.py", line 40, in fetch
+[rank7]:     raise StopIteration
+[rank7]: StopIteration
+[rank5]: Traceback (most recent call last):
+[rank5]:   File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/mini_owt_logdirichlet/train.py", line 347, in main
+[rank5]:     ids = next(it)
+[rank5]:           ^^^^^^^^
+[rank5]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/dataloader.py", line 708, in __next__
+[rank5]:     data = self._next_data()
+[rank5]:            ^^^^^^^^^^^^^^^^^
+[rank5]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/dataloader.py", line 764, in _next_data
+[rank5]:     data = self._dataset_fetcher.fetch(index)  # may raise StopIteration
+[rank5]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank5]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/_utils/fetch.py", line 40, in fetch
+[rank5]:     raise StopIteration
+[rank5]: StopIteration
+
+[rank5]: During handling of the above exception, another exception occurred:
+
+[rank5]: Traceback (most recent call last):
+[rank5]:   File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/mini_owt_logdirichlet/train.py", line 473, in <module>
+[rank5]:     main()
+[rank5]:   File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/mini_owt_logdirichlet/train.py", line 352, in main
+[rank5]:     ids = next(it)
+[rank5]:           ^^^^^^^^
+[rank5]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/dataloader.py", line 708, in __next__
+[rank5]:     data = self._next_data()
+[rank5]:            ^^^^^^^^^^^^^^^^^
+[rank5]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/dataloader.py", line 764, in _next_data
+[rank5]:     data = self._dataset_fetcher.fetch(index)  # may raise StopIteration
+[rank5]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank5]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/_utils/fetch.py", line 40, in fetch
+[rank5]:     raise StopIteration
+[rank5]: StopIteration
+[rank4]: Traceback (most recent call last):
+[rank4]:   File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/mini_owt_logdirichlet/train.py", line 347, in main
+[rank4]:     ids = next(it)
+[rank4]:           ^^^^^^^^
+[rank4]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/dataloader.py", line 708, in __next__
+[rank4]:     data = self._next_data()
+[rank4]:            ^^^^^^^^^^^^^^^^^
+[rank4]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/dataloader.py", line 764, in _next_data
+[rank4]:     data = self._dataset_fetcher.fetch(index)  # may raise StopIteration
+[rank4]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank4]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/_utils/fetch.py", line 40, in fetch
+[rank4]:     raise StopIteration
+[rank4]: StopIteration
+
+[rank4]: During handling of the above exception, another exception occurred:
+
+[rank4]: Traceback (most recent call last):
+[rank4]:   File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/mini_owt_logdirichlet/train.py", line 473, in <module>
+[rank4]:     main()
+[rank4]:   File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/mini_owt_logdirichlet/train.py", line 352, in main
+[rank4]:     ids = next(it)
+[rank4]:           ^^^^^^^^
+[rank4]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/dataloader.py", line 708, in __next__
+[rank4]:     data = self._next_data()
+[rank4]:            ^^^^^^^^^^^^^^^^^
+[rank4]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/dataloader.py", line 764, in _next_data
+[rank4]:     data = self._dataset_fetcher.fetch(index)  # may raise StopIteration
+[rank4]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+[rank4]:   File "/usr/local/lib/python3.12/dist-packages/torch/utils/data/_utils/fetch.py", line 40, in fetch
+[rank4]:     raise StopIteration
+[rank4]: StopIteration
+t-20260601014141-vst28-worker-0:10433:10685 [1] NCCL INFO misc/socket.cc:64 -> 3
+t-20260601014141-vst28-worker-0:10433:10685 [1] NCCL INFO misc/socket.cc:80 -> 3
+t-20260601014141-vst28-worker-0:10433:10685 [1] NCCL INFO misc/socket.cc:828 -> 3
+t-20260601014141-vst28-worker-0:10433:10685 [1] NCCL INFO misc/socket.cc:64 -> 3
+t-20260601014141-vst28-worker-0:10433:10685 [1] NCCL INFO misc/socket.cc:80 -> 3
+t-20260601014141-vst28-worker-0:10433:10685 [1] NCCL INFO misc/socket.cc:828 -> 3
+t-20260601014141-vst28-worker-0:10433:10685 [1] NCCL INFO misc/socket.cc:64 -> 3
+t-20260601014141-vst28-worker-0:10433:10685 [1] NCCL INFO misc/socket.cc:80 -> 3
+t-20260601014141-vst28-worker-0:10433:10685 [1] NCCL INFO misc/socket.cc:828 -> 3
+t-20260601014141-vst28-worker-0:10433:10640 [1] NCCL INFO misc/socket.cc:880 -> 3
+t-20260601014141-vst28-worker-0:10434:10687 [2] NCCL INFO misc/socket.cc:64 -> 3
+t-20260601014141-vst28-worker-0:10434:10687 [2] NCCL INFO misc/socket.cc:80 -> 3
+t-20260601014141-vst28-worker-0:10434:10687 [2] NCCL INFO misc/socket.cc:828 -> 3
+t-20260601014141-vst28-worker-0:10434:10687 [2] NCCL INFO misc/socket.cc:64 -> 3
+t-20260601014141-vst28-worker-0:10434:10687 [2] NCCL INFO misc/socket.cc:80 -> 3
+t-20260601014141-vst28-worker-0:10434:10687 [2] NCCL INFO misc/socket.cc:828 -> 3
+t-20260601014141-vst28-worker-0:10434:10687 [2] NCCL INFO misc/socket.cc:64 -> 3
+t-20260601014141-vst28-worker-0:10434:10687 [2] NCCL INFO misc/socket.cc:80 -> 3
+t-20260601014141-vst28-worker-0:10434:10687 [2] NCCL INFO misc/socket.cc:828 -> 3
+t-20260601014141-vst28-worker-0:10433:10640 [1] NCCL INFO misc/socket.cc:880 -> 3
+t-20260601014141-vst28-worker-0:10434:10687 [2] NCCL INFO misc/socket.cc:64 -> 3
+t-20260601014141-vst28-worker-0:10434:10687 [2] NCCL INFO misc/socket.cc:80 -> 3
+t-20260601014141-vst28-worker-0:10434:10687 [2] NCCL INFO misc/socket.cc:828 -> 3
+t-20260601014141-vst28-worker-0:10434:10646 [2] NCCL INFO misc/socket.cc:880 -> 3
+t-20260601014141-vst28-worker-0:10439:10692 [6] NCCL INFO misc/socket.cc:64 -> 3
+t-20260601014141-vst28-worker-0:10439:10692 [6] NCCL INFO misc/socket.cc:80 -> 3
+t-20260601014141-vst28-worker-0:10439:10692 [6] NCCL INFO misc/socket.cc:828 -> 3
+t-20260601014141-vst28-worker-0:10439:10692 [6] NCCL INFO misc/socket.cc:64 -> 3
+t-20260601014141-vst28-worker-0:10439:10692 [6] NCCL INFO misc/socket.cc:80 -> 3
+t-20260601014141-vst28-worker-0:10439:10692 [6] NCCL INFO misc/socket.cc:828 -> 3
+t-20260601014141-vst28-worker-0:10439:10692 [6] NCCL INFO misc/socket.cc:64 -> 3
+t-20260601014141-vst28-worker-0:10439:10692 [6] NCCL INFO misc/socket.cc:80 -> 3
+t-20260601014141-vst28-worker-0:10439:10692 [6] NCCL INFO misc/socket.cc:828 -> 3
+t-20260601014141-vst28-worker-0:10439:10692 [6] NCCL INFO misc/socket.cc:64 -> 3
+t-20260601014141-vst28-worker-0:10439:10692 [6] NCCL INFO misc/socket.cc:80 -> 3
+t-20260601014141-vst28-worker-0:10439:10692 [6] NCCL INFO misc/socket.cc:828 -> 3
+t-20260601014141-vst28-worker-0:10439:10634 [6] NCCL INFO misc/socket.cc:880 -> 3
+t-20260601014141-vst28-worker-0:10436:10705 [4] NCCL INFO misc/socket.cc:64 -> 3
+t-20260601014141-vst28-worker-0:10436:10705 [4] NCCL INFO misc/socket.cc:80 -> 3
+t-20260601014141-vst28-worker-0:10436:10705 [4] NCCL INFO misc/socket.cc:828 -> 3
+t-20260601014141-vst28-worker-0:10436:10705 [4] NCCL INFO misc/socket.cc:64 -> 3
+t-20260601014141-vst28-worker-0:10436:10705 [4] NCCL INFO misc/socket.cc:80 -> 3
+t-20260601014141-vst28-worker-0:10436:10705 [4] NCCL INFO misc/socket.cc:828 -> 3
+t-20260601014141-vst28-worker-0:10436:10705 [4] NCCL INFO misc/socket.cc:64 -> 3
+t-20260601014141-vst28-worker-0:10436:10705 [4] NCCL INFO misc/socket.cc:80 -> 3
+t-20260601014141-vst28-worker-0:10436:10705 [4] NCCL INFO misc/socket.cc:828 -> 3
+t-20260601014141-vst28-worker-0:10436:10705 [4] NCCL INFO misc/socket.cc:64 -> 3
+t-20260601014141-vst28-worker-0:10436:10705 [4] NCCL INFO misc/socket.cc:80 -> 3
+t-20260601014141-vst28-worker-0:10435:10644 [3] NCCL INFO misc/socket.cc:880 -> 3
+t-20260601014141-vst28-worker-0:10436:10705 [4] NCCL INFO misc/socket.cc:828 -> 3
+t-20260601014141-vst28-worker-0:10435:10706 [3] NCCL INFO misc/socket.cc:64 -> 3
+t-20260601014141-vst28-worker-0:10436:10635 [4] NCCL INFO misc/socket.cc:880 -> 3
+t-20260601014141-vst28-worker-0:10435:10706 [3] NCCL INFO misc/socket.cc:80 -> 3
+t-20260601014141-vst28-worker-0:10435:10706 [3] NCCL INFO misc/socket.cc:828 -> 3
+t-20260601014141-vst28-worker-0:10435:10706 [3] NCCL INFO misc/socket.cc:64 -> 3
+t-20260601014141-vst28-worker-0:10435:10706 [3] NCCL INFO misc/socket.cc:80 -> 3
+t-20260601014141-vst28-worker-0:10435:10706 [3] NCCL INFO misc/socket.cc:828 -> 3
+t-20260601014141-vst28-worker-0:10435:10706 [3] NCCL INFO misc/socket.cc:64 -> 3
+t-20260601014141-vst28-worker-0:10435:10706 [3] NCCL INFO misc/socket.cc:80 -> 3
+t-20260601014141-vst28-worker-0:10435:10706 [3] NCCL INFO misc/socket.cc:828 -> 3
+t-20260601014141-vst28-worker-0:10435:10644 [3] NCCL INFO misc/socket.cc:880 -> 3
+t-20260601014141-vst28-worker-0:10434:10646 [2] NCCL INFO misc/socket.cc:880 -> 3
+t-20260601014141-vst28-worker-0:10435:10706 [3] NCCL INFO misc/socket.cc:64 -> 3
+t-20260601014141-vst28-worker-0:10435:10706 [3] NCCL INFO misc/socket.cc:80 -> 3
+t-20260601014141-vst28-worker-0:10435:10706 [3] NCCL INFO misc/socket.cc:828 -> 3
+t-20260601014141-vst28-worker-0:10436:10635 [4] NCCL INFO misc/socket.cc:880 -> 3
+t-20260601014141-vst28-worker-0:10438:10708 [5] NCCL INFO misc/socket.cc:64 -> 3
+t-20260601014141-vst28-worker-0:10438:10708 [5] NCCL INFO misc/socket.cc:80 -> 3
+t-20260601014141-vst28-worker-0:10438:10708 [5] NCCL INFO misc/socket.cc:828 -> 3
+t-20260601014141-vst28-worker-0:10438:10708 [5] NCCL INFO misc/socket.cc:64 -> 3
+t-20260601014141-vst28-worker-0:10438:10708 [5] NCCL INFO misc/socket.cc:80 -> 3
+t-20260601014141-vst28-worker-0:10438:10708 [5] NCCL INFO misc/socket.cc:828 -> 3
+t-20260601014141-vst28-worker-0:10438:10708 [5] NCCL INFO misc/socket.cc:64 -> 3
+t-20260601014141-vst28-worker-0:10438:10708 [5] NCCL INFO misc/socket.cc:80 -> 3
+t-20260601014141-vst28-worker-0:10438:10708 [5] NCCL INFO misc/socket.cc:828 -> 3
+t-20260601014141-vst28-worker-0:10438:10708 [5] NCCL INFO misc/socket.cc:64 -> 3
+t-20260601014141-vst28-worker-0:10438:10708 [5] NCCL INFO misc/socket.cc:80 -> 3
+t-20260601014141-vst28-worker-0:10438:10708 [5] NCCL INFO misc/socket.cc:828 -> 3
+t-20260601014141-vst28-worker-0:10438:10638 [5] NCCL INFO misc/socket.cc:880 -> 3
+t-20260601014141-vst28-worker-0:10439:10634 [6] NCCL INFO misc/socket.cc:880 -> 3
+t-20260601014141-vst28-worker-0:10440:10710 [7] NCCL INFO misc/socket.cc:64 -> 3
+t-20260601014141-vst28-worker-0:10440:10710 [7] NCCL INFO misc/socket.cc:80 -> 3
+t-20260601014141-vst28-worker-0:10440:10710 [7] NCCL INFO misc/socket.cc:828 -> 3
+t-20260601014141-vst28-worker-0:10440:10710 [7] NCCL INFO misc/socket.cc:64 -> 3
+t-20260601014141-vst28-worker-0:10440:10710 [7] NCCL INFO misc/socket.cc:80 -> 3
+t-20260601014141-vst28-worker-0:10440:10710 [7] NCCL INFO misc/socket.cc:828 -> 3
+t-20260601014141-vst28-worker-0:10440:10710 [7] NCCL INFO misc/socket.cc:64 -> 3
+t-20260601014141-vst28-worker-0:10440:10710 [7] NCCL INFO misc/socket.cc:80 -> 3
+t-20260601014141-vst28-worker-0:10440:10710 [7] NCCL INFO misc/socket.cc:828 -> 3
+t-20260601014141-vst28-worker-0:10440:10632 [7] NCCL INFO misc/socket.cc:880 -> 3
+t-20260601014141-vst28-worker-0:10433:10685 [1] NCCL INFO comm 0x99599b0 rank 1 nranks 8 cudaDev 1 busId 67020 - Abort COMPLETE
+t-20260601014141-vst28-worker-0:10434:10687 [2] NCCL INFO comm 0x9a70eb0 rank 2 nranks 8 cudaDev 2 busId 69020 - Abort COMPLETE
+t-20260601014141-vst28-worker-0:10439:10692 [6] NCCL INFO comm 0xa3f72f0 rank 6 nranks 8 cudaDev 6 busId 73020 - Abort COMPLETE
+W0531 17:42:36.378000 10365 torch/distributed/elastic/multiprocessing/api.py:898] Sending process 10432 closing signal SIGTERM
+W0531 17:42:36.379000 10365 torch/distributed/elastic/multiprocessing/api.py:898] Sending process 10435 closing signal SIGTERM
+W0531 17:42:36.380000 10365 torch/distributed/elastic/multiprocessing/api.py:898] Sending process 10436 closing signal SIGTERM
+W0531 17:42:36.380000 10365 torch/distributed/elastic/multiprocessing/api.py:898] Sending process 10438 closing signal SIGTERM
+W0531 17:42:36.381000 10365 torch/distributed/elastic/multiprocessing/api.py:898] Sending process 10440 closing signal SIGTERM
+E0531 17:42:36.909000 10365 torch/distributed/elastic/multiprocessing/api.py:870] failed (exitcode: 1) local_rank: 1 (pid: 10433) of binary: /usr/bin/python
+Traceback (most recent call last):
+  File "/usr/local/bin/torchrun", line 33, in <module>
+    sys.exit(load_entry_point('torch==2.7.0a0+ecf3bae40a.nv25.2', 'console_scripts', 'torchrun')())
+             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+  File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/multiprocessing/errors/__init__.py", line 355, in wrapper
+    return f(*args, **kwargs)
+           ^^^^^^^^^^^^^^^^^^
+  File "/usr/local/lib/python3.12/dist-packages/torch/distributed/run.py", line 918, in main
+    run(args)
+  File "/usr/local/lib/python3.12/dist-packages/torch/distributed/run.py", line 909, in run
+    elastic_launch(
+  File "/usr/local/lib/python3.12/dist-packages/torch/distributed/launcher/api.py", line 139, in __call__
+    return launch_agent(self._config, self._entrypoint, list(args))
+           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+  File "/usr/local/lib/python3.12/dist-packages/torch/distributed/launcher/api.py", line 270, in launch_agent
+    raise ChildFailedError(
+torch.distributed.elastic.multiprocessing.errors.ChildFailedError: 
+============================================================
+train.py FAILED
+------------------------------------------------------------
+Failures:
+[1]:
+  time      : 2026-05-31_17:42:36
+  host      : t-20260601014141-vst28-worker-0.t-20260601014141-vst28-worker.mlplatform-customtask.svc.cluster.local
+  rank      : 2 (local_rank: 2)
+  exitcode  : 1 (pid: 10434)
+  error_file: <N/A>
+  traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html
+[2]:
+  time      : 2026-05-31_17:42:36
+  host      : t-20260601014141-vst28-worker-0.t-20260601014141-vst28-worker.mlplatform-customtask.svc.cluster.local
+  rank      : 6 (local_rank: 6)
+  exitcode  : 1 (pid: 10439)
+  error_file: <N/A>
+  traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html
+------------------------------------------------------------
+Root Cause (first observed failure):
+[0]:
+  time      : 2026-05-31_17:42:36
+  host      : t-20260601014141-vst28-worker-0.t-20260601014141-vst28-worker.mlplatform-customtask.svc.cluster.local
+  rank      : 1 (local_rank: 1)
+  exitcode  : 1 (pid: 10433)
+  error_file: <N/A>
+  traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html
+============================================================
+[exit] 2026-05-31T17:42:37+00:00 rc=1 run=owt_t5_elftokenized_full_len1024_C1_to_1024_pow1_d768_l12_h12_gbs512_8gpu_5ep_elfopt_t5embed_unfixed_probadd_selfcond_ce_20260531_174225