diffusion_baseline / results /elf-progress.json
jyliuAI's picture
Add files using upload-large-folder tool
784f572 verified
Raw History Blame Contribute Delete
52 kB
{ "schema": "elf-progress-v1", "runs": [ { "run": "/home/ruizzhao/diffusion_baseline/results/elf-gpu01/quality/owt/unconditional/steps_1024/seed_42/compile_0", "status": "quality_valid", "task": "owt", "steps": 1024, "seed": 42, "compile": false, "split": "unconditional", "protocol": "elf_native_v1", "condition_policy": "none", "sample_count": 1024, "prompt_count": 1024, "metrics": { "gpt2_length_mean": 920.2958984375, "gpt2_length_min": 728, "empty_fraction": 0.0, "below_64_tokens_fraction": 0.0, "above_1024_tokens_fraction": 0.009765625, "entropy_gpt2_nats": 4.742035427877739, "generative_ppl": { "perplexity": 7.732375373221586, "total_nll": 1925010.6435546875, "valid_token_count": 941134, "sample_count": 1024, "batch_size": 8, "context_length": 1024, "truncation": "right", "model_id": "gpt2-large", "model_revision": "32b71b12589c2f8d625668d2335a01cac3249519", "tokenizer_id": "gpt2", "tokenizer_revision": "607a30d783dfa663caf39e06633721c8d4cfcd7e", "aggregation": "sum_nll_over_sum_valid_next_tokens" }, "entropy_policy": "mean per-sample natural-log entropy on retokenized GPT-2 suffix/text; same PPL length cap" }, "reason": null, "owt_main_grid": true }, { "run": "/home/ruizzhao/diffusion_baseline/results/elf-gpu01/quality/owt-prefix/c64-text-t5/steps_1/seed_42/compile_0", "status": "quality_valid", "task": "owt-prefix", "steps": 1, "seed": 42, "compile": false, "split": "c64-text-t5", "protocol": "c64_text_t5_v1", "condition_policy": "c64_raw_no_special_tokens_v1", "sample_count": 2048, "prompt_count": 1024, "metrics": { "gpt2_length_mean": 1196.43359375, "gpt2_length_min": 297, "empty_fraction": 0.0, "below_64_tokens_fraction": 0.0, "above_1024_tokens_fraction": 0.857421875, "entropy_gpt2_nats": 1.5581149953523385, "conditional_ppl": { "perplexity": 15.67715969847609, "total_nll": 180368.4974975586, "valid_token_count": 65536, "sample_count": 0, "batch_size": 0, "context_length": 1024, "truncation": "right", "model_id": "gpt2-large", "model_revision": "", "tokenizer_id": "gpt2", "tokenizer_revision": "", "aggregation": "sum_nll_over_sum_valid_next_tokens" }, "reference_conditional_ppl": { "perplexity": 17.09677422584074, "total_nll": 185955.798828125, "valid_token_count": 65503, "sample_count": 0, "batch_size": 0, "context_length": 1024, "truncation": "right", "model_id": "gpt2-large", "model_revision": "", "tokenizer_id": "gpt2", "tokenizer_revision": "", "aggregation": "sum_nll_over_sum_valid_next_tokens" }, "conditional_ppl_policy": "separate GPT-2 prefix/suffix tokenization; prefix targets masked", "entropy_all_completions_gpt2_nats": 1.5677205978966589, "grouped_self_bleu": 0.7343077781716898, "entropy_policy": "mean per-sample natural-log entropy on retokenized GPT-2 suffix/text; same PPL length cap" }, "reason": null, "owt_main_grid": false }, { "run": "/home/ruizzhao/diffusion_baseline/results/elf-gpu01/quality/owt-prefix/c64-text-t5/steps_1024/seed_42/compile_0", "status": "quality_valid", "task": "owt-prefix", "steps": 1024, "seed": 42, "compile": false, "split": "c64-text-t5", "protocol": "c64_text_t5_v1", "condition_policy": "c64_raw_no_special_tokens_v1", "sample_count": 2048, "prompt_count": 1024, "metrics": { "gpt2_length_mean": 840.10546875, "gpt2_length_min": 314, "empty_fraction": 0.0, "below_64_tokens_fraction": 0.0, "above_1024_tokens_fraction": 0.0, "entropy_gpt2_nats": 3.6359209661926672, "conditional_ppl": { "perplexity": 54.460543750001705, "total_nll": 261978.6180419922, "valid_token_count": 65536, "sample_count": 0, "batch_size": 0, "context_length": 1024, "truncation": "right", "model_id": "gpt2-large", "model_revision": "", "tokenizer_id": "gpt2", "tokenizer_revision": "", "aggregation": "sum_nll_over_sum_valid_next_tokens" }, "reference_conditional_ppl": { "perplexity": 17.09677422584074, "total_nll": 185955.798828125, "valid_token_count": 65503, "sample_count": 0, "batch_size": 0, "context_length": 1024, "truncation": "right", "model_id": "gpt2-large", "model_revision": "", "tokenizer_id": "gpt2", "tokenizer_revision": "", "aggregation": "sum_nll_over_sum_valid_next_tokens" }, "conditional_ppl_policy": "separate GPT-2 prefix/suffix tokenization; prefix targets masked", "entropy_all_completions_gpt2_nats": 3.6433539446522816, "grouped_self_bleu": 0.02856444539729572, "entropy_policy": "mean per-sample natural-log entropy on retokenized GPT-2 suffix/text; same PPL length cap" }, "reason": null, "owt_main_grid": false }, { "run": "/home/ruizzhao/diffusion_baseline/results/elf-gpu01/quality/owt-prefix/c64-text-t5/steps_16/seed_42/compile_0", "status": "quality_valid", "task": "owt-prefix", "steps": 16, "seed": 42, "compile": false, "split": "c64-text-t5", "protocol": "c64_text_t5_v1", "condition_policy": "c64_raw_no_special_tokens_v1", "sample_count": 2048, "prompt_count": 1024, "metrics": { "gpt2_length_mean": 889.7255859375, "gpt2_length_min": 547, "empty_fraction": 0.0, "below_64_tokens_fraction": 0.0, "above_1024_tokens_fraction": 0.0146484375, "entropy_gpt2_nats": 3.7182116660378637, "conditional_ppl": { "perplexity": 77.81100087401236, "total_nll": 285362.27893066406, "valid_token_count": 65536, "sample_count": 0, "batch_size": 0, "context_length": 1024, "truncation": "right", "model_id": "gpt2-large", "model_revision": "", "tokenizer_id": "gpt2", "tokenizer_revision": "", "aggregation": "sum_nll_over_sum_valid_next_tokens" }, "reference_conditional_ppl": { "perplexity": 17.09677422584074, "total_nll": 185955.798828125, "valid_token_count": 65503, "sample_count": 0, "batch_size": 0, "context_length": 1024, "truncation": "right", "model_id": "gpt2-large", "model_revision": "", "tokenizer_id": "gpt2", "tokenizer_revision": "", "aggregation": "sum_nll_over_sum_valid_next_tokens" }, "conditional_ppl_policy": "separate GPT-2 prefix/suffix tokenization; prefix targets masked", "entropy_all_completions_gpt2_nats": 3.7202811412326118, "grouped_self_bleu": 0.029067700940403896, "entropy_policy": "mean per-sample natural-log entropy on retokenized GPT-2 suffix/text; same PPL length cap" }, "reason": null, "owt_main_grid": false }, { "run": "/home/ruizzhao/diffusion_baseline/results/elf-gpu01/quality/owt-prefix/c64-text-t5/steps_2/seed_42/compile_0", "status": "quality_valid", "task": "owt-prefix", "steps": 2, "seed": 42, "compile": false, "split": "c64-text-t5", "protocol": "c64_text_t5_v1", "condition_policy": "c64_raw_no_special_tokens_v1", "sample_count": 2048, "prompt_count": 1024, "metrics": { "gpt2_length_mean": 1710.9833984375, "gpt2_length_min": 378, "empty_fraction": 0.0, "below_64_tokens_fraction": 0.0, "above_1024_tokens_fraction": 0.9814453125, "entropy_gpt2_nats": 3.3800256392003765, "conditional_ppl": { "perplexity": 109.13122719759768, "total_nll": 307531.02740478516, "valid_token_count": 65536, "sample_count": 0, "batch_size": 0, "context_length": 1024, "truncation": "right", "model_id": "gpt2-large", "model_revision": "", "tokenizer_id": "gpt2", "tokenizer_revision": "", "aggregation": "sum_nll_over_sum_valid_next_tokens" }, "reference_conditional_ppl": { "perplexity": 17.09677422584074, "total_nll": 185955.798828125, "valid_token_count": 65503, "sample_count": 0, "batch_size": 0, "context_length": 1024, "truncation": "right", "model_id": "gpt2-large", "model_revision": "", "tokenizer_id": "gpt2", "tokenizer_revision": "", "aggregation": "sum_nll_over_sum_valid_next_tokens" }, "conditional_ppl_policy": "separate GPT-2 prefix/suffix tokenization; prefix targets masked", "entropy_all_completions_gpt2_nats": 3.381890864218911, "grouped_self_bleu": 0.09698413477700692, "entropy_policy": "mean per-sample natural-log entropy on retokenized GPT-2 suffix/text; same PPL length cap" }, "reason": null, "owt_main_grid": false }, { "run": "/home/ruizzhao/diffusion_baseline/results/elf-gpu01/quality/owt-prefix/c64-text-t5/steps_32/seed_42/compile_0", "status": "quality_valid", "task": "owt-prefix", "steps": 32, "seed": 42, "compile": false, "split": "c64-text-t5", "protocol": "c64_text_t5_v1", "condition_policy": "c64_raw_no_special_tokens_v1", "sample_count": 2048, "prompt_count": 1024, "metrics": { "gpt2_length_mean": 866.5859375, "gpt2_length_min": 631, "empty_fraction": 0.0, "below_64_tokens_fraction": 0.0, "above_1024_tokens_fraction": 0.0, "entropy_gpt2_nats": 3.720487705380751, "conditional_ppl": { "perplexity": 61.0453710353483, "total_nll": 269458.9562988281, "valid_token_count": 65536, "sample_count": 0, "batch_size": 0, "context_length": 1024, "truncation": "right", "model_id": "gpt2-large", "model_revision": "", "tokenizer_id": "gpt2", "tokenizer_revision": "", "aggregation": "sum_nll_over_sum_valid_next_tokens" }, "reference_conditional_ppl": { "perplexity": 17.09677422584074, "total_nll": 185955.798828125, "valid_token_count": 65503, "sample_count": 0, "batch_size": 0, "context_length": 1024, "truncation": "right", "model_id": "gpt2-large", "model_revision": "", "tokenizer_id": "gpt2", "tokenizer_revision": "", "aggregation": "sum_nll_over_sum_valid_next_tokens" }, "conditional_ppl_policy": "separate GPT-2 prefix/suffix tokenization; prefix targets masked", "entropy_all_completions_gpt2_nats": 3.725415810106105, "grouped_self_bleu": 0.03318571544649222, "entropy_policy": "mean per-sample natural-log entropy on retokenized GPT-2 suffix/text; same PPL length cap" }, "reason": null, "owt_main_grid": false }, { "run": "/home/ruizzhao/diffusion_baseline/results/elf-gpu01/quality/owt-prefix/c64-text-t5/steps_4/seed_42/compile_0", "status": "quality_valid", "task": "owt-prefix", "steps": 4, "seed": 42, "compile": false, "split": "c64-text-t5", "protocol": "c64_text_t5_v1", "condition_policy": "c64_raw_no_special_tokens_v1", "sample_count": 2048, "prompt_count": 1024, "metrics": { "gpt2_length_mean": 1468.8671875, "gpt2_length_min": 120, "empty_fraction": 0.0, "below_64_tokens_fraction": 0.0, "above_1024_tokens_fraction": 0.8134765625, "entropy_gpt2_nats": 3.7204756612412844, "conditional_ppl": { "perplexity": 279.8318645022191, "total_nll": 369242.2062988281, "valid_token_count": 65536, "sample_count": 0, "batch_size": 0, "context_length": 1024, "truncation": "right", "model_id": "gpt2-large", "model_revision": "", "tokenizer_id": "gpt2", "tokenizer_revision": "", "aggregation": "sum_nll_over_sum_valid_next_tokens" }, "reference_conditional_ppl": { "perplexity": 17.09677422584074, "total_nll": 185955.798828125, "valid_token_count": 65503, "sample_count": 0, "batch_size": 0, "context_length": 1024, "truncation": "right", "model_id": "gpt2-large", "model_revision": "", "tokenizer_id": "gpt2", "tokenizer_revision": "", "aggregation": "sum_nll_over_sum_valid_next_tokens" }, "conditional_ppl_policy": "separate GPT-2 prefix/suffix tokenization; prefix targets masked", "entropy_all_completions_gpt2_nats": 3.7329112460820646, "grouped_self_bleu": 0.023462587521361787, "entropy_policy": "mean per-sample natural-log entropy on retokenized GPT-2 suffix/text; same PPL length cap" }, "reason": null, "owt_main_grid": false }, { "run": "/home/ruizzhao/diffusion_baseline/results/elf-gpu01/quality/owt-prefix/c64-text-t5/steps_8/seed_42/compile_0", "status": "quality_valid", "task": "owt-prefix", "steps": 8, "seed": 42, "compile": false, "split": "c64-text-t5", "protocol": "c64_text_t5_v1", "condition_policy": "c64_raw_no_special_tokens_v1", "sample_count": 2048, "prompt_count": 1024, "metrics": { "gpt2_length_mean": 1032.4716796875, "gpt2_length_min": 187, "empty_fraction": 0.0, "below_64_tokens_fraction": 0.0, "above_1024_tokens_fraction": 0.2255859375, "entropy_gpt2_nats": 3.732119776042421, "conditional_ppl": { "perplexity": 195.16295145471986, "total_nll": 345626.0412597656, "valid_token_count": 65536, "sample_count": 0, "batch_size": 0, "context_length": 1024, "truncation": "right", "model_id": "gpt2-large", "model_revision": "", "tokenizer_id": "gpt2", "tokenizer_revision": "", "aggregation": "sum_nll_over_sum_valid_next_tokens" }, "reference_conditional_ppl": { "perplexity": 17.09677422584074, "total_nll": 185955.798828125, "valid_token_count": 65503, "sample_count": 0, "batch_size": 0, "context_length": 1024, "truncation": "right", "model_id": "gpt2-large", "model_revision": "", "tokenizer_id": "gpt2", "tokenizer_revision": "", "aggregation": "sum_nll_over_sum_valid_next_tokens" }, "conditional_ppl_policy": "separate GPT-2 prefix/suffix tokenization; prefix targets masked", "entropy_all_completions_gpt2_nats": 3.734061037101676, "grouped_self_bleu": 0.023385133132632106, "entropy_policy": "mean per-sample natural-log entropy on retokenized GPT-2 suffix/text; same PPL length cap" }, "reason": null, "owt_main_grid": false }, { "run": "/home/ruizzhao/diffusion_baseline/results/elf-gpu01/quality/wmt14/validation/steps_1/seed_42/compile_0", "status": "quality_valid", "task": "wmt14", "steps": 1, "seed": 42, "compile": false, "split": "validation", "protocol": "elf_native_v1", "condition_policy": "released_ids_or_raw_plus_eos_then_source_cap_v2", "sample_count": 3000, "prompt_count": 3000, "metrics": { "bleu": 11.505873872162137, "bleu_signature": "nrefs:1|case:lc|eff:yes|tok:13a|smooth:exp|version:2.5.1", "empty_fraction": 0.0, "rouge1": 42.72518224597211, "rouge1_sem": 0.3376938096795512, "rouge2": 18.627999236000417, "rouge2_sem": 0.3288160344527289, "rougeL": 39.71823122057084, "rougeL_sem": 0.34136793858663844, "reference_policy": "full raw reference; no target truncation; ROUGE F1 stemmer=true" }, "reason": null, "owt_main_grid": false }, { "run": "/home/ruizzhao/diffusion_baseline/results/elf-gpu01/quality/wmt14/validation/steps_2/seed_42/compile_0", "status": "quality_valid", "task": "wmt14", "steps": 2, "seed": 42, "compile": false, "split": "validation", "protocol": "elf_native_v1", "condition_policy": "released_ids_or_raw_plus_eos_then_source_cap_v2", "sample_count": 3000, "prompt_count": 3000, "metrics": { "bleu": 18.20139347188345, "bleu_signature": "nrefs:1|case:lc|eff:yes|tok:13a|smooth:exp|version:2.5.1", "empty_fraction": 0.0, "rouge1": 53.169236264921295, "rouge1_sem": 0.3210215656436727, "rouge2": 26.71154624735424, "rouge2_sem": 0.358726198134827, "rougeL": 48.8386748015013, "rougeL_sem": 0.33459349211712974, "reference_policy": "full raw reference; no target truncation; ROUGE F1 stemmer=true" }, "reason": null, "owt_main_grid": false }, { "run": "/home/ruizzhao/diffusion_baseline/results/elf-gpu01/quality/wmt14/validation/steps_4/seed_42/compile_0", "status": "quality_valid", "task": "wmt14", "steps": 4, "seed": 42, "compile": false, "split": "validation", "protocol": "elf_native_v1", "condition_policy": "released_ids_or_raw_plus_eos_then_source_cap_v2", "sample_count": 3000, "prompt_count": 3000, "metrics": { "bleu": 22.179790549993754, "bleu_signature": "nrefs:1|case:lc|eff:yes|tok:13a|smooth:exp|version:2.5.1", "empty_fraction": 0.0, "rouge1": 57.05040287412753, "rouge1_sem": 0.3222957518145479, "rouge2": 30.51517872008967, "rouge2_sem": 0.3763067105857128, "rougeL": 52.20686019086576, "rougeL_sem": 0.33927305818515563, "reference_policy": "full raw reference; no target truncation; ROUGE F1 stemmer=true" }, "reason": null, "owt_main_grid": false }, { "run": "/home/ruizzhao/diffusion_baseline/results/elf-gpu01/quality/xsum/validation/steps_1/seed_42/compile_0", "status": "quality_valid", "task": "xsum", "steps": 1, "seed": 42, "compile": false, "split": "validation", "protocol": "elf_native_v1", "condition_policy": "released_ids_or_raw_plus_eos_then_source_cap_v2", "sample_count": 11332, "prompt_count": 11332, "metrics": { "bleu": 3.2517152186129366, "bleu_signature": "nrefs:1|case:lc|eff:yes|tok:13a|smooth:exp|version:2.5.1", "empty_fraction": 0.0, "rouge1": 23.71124486133378, "rouge1_sem": 0.11001167733369738, "rouge2": 5.741971168927821, "rouge2_sem": 0.07524688853635392, "rougeL": 19.426402931250138, "rougeL_sem": 0.10098594251946032, "reference_policy": "full raw reference; no target truncation; ROUGE F1 stemmer=true" }, "reason": null, "owt_main_grid": false }, { "run": "/home/ruizzhao/diffusion_baseline/results/elf-gpu01/quality/xsum/validation/steps_2/seed_42/compile_0", "status": "quality_valid", "task": "xsum", "steps": 2, "seed": 42, "compile": false, "split": "validation", "protocol": "elf_native_v1", "condition_policy": "released_ids_or_raw_plus_eos_then_source_cap_v2", "sample_count": 11332, "prompt_count": 11332, "metrics": { "bleu": 4.985643594403319, "bleu_signature": "nrefs:1|case:lc|eff:yes|tok:13a|smooth:exp|version:2.5.1", "empty_fraction": 0.0, "rouge1": 29.44908231699352, "rouge1_sem": 0.12013341103530582, "rouge2": 8.294996552944877, "rouge2_sem": 0.0911483100689226, "rougeL": 23.378245901295077, "rougeL_sem": 0.11191556963495196, "reference_policy": "full raw reference; no target truncation; ROUGE F1 stemmer=true" }, "reason": null, "owt_main_grid": false }, { "run": "/home/ruizzhao/diffusion_baseline/results/elf-gpu01/quality/xsum/validation/steps_4/seed_42/compile_0", "status": "quality_valid", "task": "xsum", "steps": 4, "seed": 42, "compile": false, "split": "validation", "protocol": "elf_native_v1", "condition_policy": "released_ids_or_raw_plus_eos_then_source_cap_v2", "sample_count": 11332, "prompt_count": 11332, "metrics": { "bleu": 6.333090223951085, "bleu_signature": "nrefs:1|case:lc|eff:yes|tok:13a|smooth:exp|version:2.5.1", "empty_fraction": 0.0, "rouge1": 32.629901406751415, "rouge1_sem": 0.12380493723417893, "rouge2": 10.042960139236806, "rouge2_sem": 0.09959482427692108, "rougeL": 25.492594404193664, "rougeL_sem": 0.11745609501449468, "reference_policy": "full raw reference; no target truncation; ROUGE F1 stemmer=true" }, "reason": null, "owt_main_grid": false }, { "run": "/home/ruizzhao/diffusion_baseline/results/elf-h200-gpu2/quality/owt/unconditional/steps_1/seed_42/compile_0", "status": "quality_valid", "task": "owt", "steps": 1, "seed": 42, "compile": false, "split": "unconditional", "protocol": "elf_native_v1", "condition_policy": "none", "sample_count": 1024, "prompt_count": 1024, "metrics": { "gpt2_length_mean": 1435.4521484375, "gpt2_length_min": 1153, "empty_fraction": 0.0, "below_64_tokens_fraction": 0.0, "above_1024_tokens_fraction": 1.0, "entropy_gpt2_nats": 1.9937805301450826, "generative_ppl": { "perplexity": 2.965652787484111, "total_nll": 1138790.8173828125, "valid_token_count": 1047552, "sample_count": 1024, "batch_size": 8, "context_length": 1024, "truncation": "right", "model_id": "gpt2-large", "model_revision": "32b71b12589c2f8d625668d2335a01cac3249519", "tokenizer_id": "gpt2", "tokenizer_revision": "607a30d783dfa663caf39e06633721c8d4cfcd7e", "aggregation": "sum_nll_over_sum_valid_next_tokens" }, "entropy_policy": "mean per-sample natural-log entropy on retokenized GPT-2 suffix/text; same PPL length cap" }, "reason": null, "owt_main_grid": true }, { "run": "/home/ruizzhao/diffusion_baseline/results/elf-h200-gpu2/quality/owt/unconditional/steps_16/seed_42/compile_0", "status": "quality_valid", "task": "owt", "steps": 16, "seed": 42, "compile": false, "split": "unconditional", "protocol": "elf_native_v1", "condition_policy": null, "sample_count": 1024, "prompt_count": 1024, "metrics": { "gpt2_length_mean": 970.0712890625, "gpt2_length_min": 90, "empty_fraction": 0.0, "below_64_tokens_fraction": 0.0, "above_1024_tokens_fraction": 0.029296875, "entropy_gpt2_nats": 5.160906091164785, "generative_ppl": { "perplexity": 32.176145854763625, "total_nll": 3411100.271484375, "valid_token_count": 982679, "sample_count": 1024, "batch_size": 8, "context_length": 1024, "truncation": "right", "model_id": "gpt2-large", "model_revision": "32b71b12589c2f8d625668d2335a01cac3249519", "tokenizer_id": "gpt2", "tokenizer_revision": "607a30d783dfa663caf39e06633721c8d4cfcd7e", "aggregation": "sum_nll_over_sum_valid_next_tokens" }, "entropy_policy": "mean per-sample natural-log entropy on retokenized GPT-2 suffix/text; same PPL length cap" }, "reason": null, "owt_main_grid": true }, { "run": "/home/ruizzhao/diffusion_baseline/results/elf-h200-gpu2/quality/owt/unconditional/steps_2/seed_42/compile_0", "status": "quality_valid", "task": "owt", "steps": 2, "seed": 42, "compile": false, "split": "unconditional", "protocol": "elf_native_v1", "condition_policy": "none", "sample_count": 1024, "prompt_count": 1024, "metrics": { "gpt2_length_mean": 1735.1103515625, "gpt2_length_min": 1070, "empty_fraction": 0.0, "below_64_tokens_fraction": 0.0, "above_1024_tokens_fraction": 1.0, "entropy_gpt2_nats": 4.545579040871858, "generative_ppl": { "perplexity": 29.373875907082294, "total_nll": 3540836.4912109375, "valid_token_count": 1047552, "sample_count": 1024, "batch_size": 8, "context_length": 1024, "truncation": "right", "model_id": "gpt2-large", "model_revision": "32b71b12589c2f8d625668d2335a01cac3249519", "tokenizer_id": "gpt2", "tokenizer_revision": "607a30d783dfa663caf39e06633721c8d4cfcd7e", "aggregation": "sum_nll_over_sum_valid_next_tokens" }, "entropy_policy": "mean per-sample natural-log entropy on retokenized GPT-2 suffix/text; same PPL length cap" }, "reason": null, "owt_main_grid": true }, { "run": "/home/ruizzhao/diffusion_baseline/results/elf-h200-gpu2/quality/owt/unconditional/steps_32/seed_42/compile_0", "status": "quality_valid", "task": "owt", "steps": 32, "seed": 42, "compile": false, "split": "unconditional", "protocol": "elf_native_v1", "condition_policy": null, "sample_count": 1024, "prompt_count": 1024, "metrics": { "gpt2_length_mean": 946.5576171875, "gpt2_length_min": 828, "empty_fraction": 0.0, "below_64_tokens_fraction": 0.0, "above_1024_tokens_fraction": 0.001953125, "entropy_gpt2_nats": 5.165104822228353, "generative_ppl": { "perplexity": 24.351687216961576, "total_nll": 3091172.19921875, "valid_token_count": 968230, "sample_count": 1024, "batch_size": 8, "context_length": 1024, "truncation": "right", "model_id": "gpt2-large", "model_revision": "32b71b12589c2f8d625668d2335a01cac3249519", "tokenizer_id": "gpt2", "tokenizer_revision": "607a30d783dfa663caf39e06633721c8d4cfcd7e", "aggregation": "sum_nll_over_sum_valid_next_tokens" }, "entropy_policy": "mean per-sample natural-log entropy on retokenized GPT-2 suffix/text; same PPL length cap" }, "reason": null, "owt_main_grid": true }, { "run": "/home/ruizzhao/diffusion_baseline/results/elf-h200-gpu2/quality/owt/unconditional/steps_4/seed_42/compile_0", "status": "quality_valid", "task": "owt", "steps": 4, "seed": 42, "compile": false, "split": "unconditional", "protocol": "elf_native_v1", "condition_policy": "none", "sample_count": 1024, "prompt_count": 1024, "metrics": { "gpt2_length_mean": 1541.1455078125, "gpt2_length_min": 29, "empty_fraction": 0.0, "below_64_tokens_fraction": 0.0029296875, "above_1024_tokens_fraction": 0.8671875, "entropy_gpt2_nats": 5.084088689145975, "generative_ppl": { "perplexity": 60.12958963298963, "total_nll": 4204850.443359375, "valid_token_count": 1026449, "sample_count": 1024, "batch_size": 8, "context_length": 1024, "truncation": "right", "model_id": "gpt2-large", "model_revision": "32b71b12589c2f8d625668d2335a01cac3249519", "tokenizer_id": "gpt2", "tokenizer_revision": "607a30d783dfa663caf39e06633721c8d4cfcd7e", "aggregation": "sum_nll_over_sum_valid_next_tokens" }, "entropy_policy": "mean per-sample natural-log entropy on retokenized GPT-2 suffix/text; same PPL length cap" }, "reason": null, "owt_main_grid": true }, { "run": "/home/ruizzhao/diffusion_baseline/results/elf-h200-gpu2/quality/owt/unconditional/steps_64/seed_42/compile_0", "status": "quality_valid", "task": "owt", "steps": 64, "seed": 42, "compile": false, "split": "unconditional", "protocol": "elf_native_v1", "condition_policy": null, "sample_count": 1024, "prompt_count": 1024, "metrics": { "gpt2_length_mean": 941.1806640625, "gpt2_length_min": 776, "empty_fraction": 0.0, "below_64_tokens_fraction": 0.0, "above_1024_tokens_fraction": 0.0107421875, "entropy_gpt2_nats": 5.077369080826807, "generative_ppl": { "perplexity": 19.124748429177174, "total_nll": 2840430.541015625, "valid_token_count": 962537, "sample_count": 1024, "batch_size": 8, "context_length": 1024, "truncation": "right", "model_id": "gpt2-large", "model_revision": "32b71b12589c2f8d625668d2335a01cac3249519", "tokenizer_id": "gpt2", "tokenizer_revision": "607a30d783dfa663caf39e06633721c8d4cfcd7e", "aggregation": "sum_nll_over_sum_valid_next_tokens" }, "entropy_policy": "mean per-sample natural-log entropy on retokenized GPT-2 suffix/text; same PPL length cap" }, "reason": null, "owt_main_grid": false }, { "run": "/home/ruizzhao/diffusion_baseline/results/elf-h200-gpu2/quality/owt/unconditional/steps_8/seed_42/compile_0", "status": "quality_valid", "task": "owt", "steps": 8, "seed": 42, "compile": false, "split": "unconditional", "protocol": "elf_native_v1", "condition_policy": null, "sample_count": 1024, "prompt_count": 1024, "metrics": { "gpt2_length_mean": 1142.294921875, "gpt2_length_min": 83, "empty_fraction": 0.0, "below_64_tokens_fraction": 0.0, "above_1024_tokens_fraction": 0.376953125, "entropy_gpt2_nats": 5.252830425956166, "generative_ppl": { "perplexity": 70.94707709717905, "total_nll": 4277792.86328125, "valid_token_count": 1003721, "sample_count": 1024, "batch_size": 8, "context_length": 1024, "truncation": "right", "model_id": "gpt2-large", "model_revision": "32b71b12589c2f8d625668d2335a01cac3249519", "tokenizer_id": "gpt2", "tokenizer_revision": "607a30d783dfa663caf39e06633721c8d4cfcd7e", "aggregation": "sum_nll_over_sum_valid_next_tokens" }, "entropy_policy": "mean per-sample natural-log entropy on retokenized GPT-2 suffix/text; same PPL length cap" }, "reason": null, "owt_main_grid": true }, { "run": "/home/ruizzhao/diffusion_baseline/results/elf-h200-gpu2/quality/wmt14/validation/steps_16/seed_42/compile_0", "status": "quality_valid", "task": "wmt14", "steps": 16, "seed": 42, "compile": false, "split": "validation", "protocol": "elf_native_v1", "condition_policy": null, "sample_count": 3000, "prompt_count": 3000, "metrics": { "bleu": 13.401398197976961, "bleu_signature": "nrefs:1|case:lc|eff:yes|tok:13a|smooth:exp|version:2.5.1", "empty_fraction": 0.0026666666666666666, "rouge1": 42.10586726615773, "rouge1_sem": 0.39712024842128857, "rouge2": 20.214958587921828, "rouge2_sem": 0.31972426013372557, "rougeL": 37.08616804499092, "rougeL_sem": 0.37914262727794484, "reference_policy": "full raw reference; no target truncation; ROUGE F1 stemmer=true" }, "reason": null, "owt_main_grid": false }, { "run": "/home/ruizzhao/diffusion_baseline/results/elf-h200-gpu2/quality/wmt14/validation/steps_32/seed_42/compile_0", "status": "quality_valid", "task": "wmt14", "steps": 32, "seed": 42, "compile": false, "split": "validation", "protocol": "elf_native_v1", "condition_policy": null, "sample_count": 3000, "prompt_count": 3000, "metrics": { "bleu": 14.002077172974603, "bleu_signature": "nrefs:1|case:lc|eff:yes|tok:13a|smooth:exp|version:2.5.1", "empty_fraction": 0.0026666666666666666, "rouge1": 42.9331151761475, "rouge1_sem": 0.39226558900842284, "rouge2": 21.239180895947253, "rouge2_sem": 0.31796234076970126, "rougeL": 37.822469967124675, "rougeL_sem": 0.3746122865609226, "reference_policy": "full raw reference; no target truncation; ROUGE F1 stemmer=true" }, "reason": null, "owt_main_grid": false }, { "run": "/home/ruizzhao/diffusion_baseline/results/elf-h200-gpu2/quality/wmt14/validation/steps_64/seed_42/compile_0", "status": "quality_valid", "task": "wmt14", "steps": 64, "seed": 42, "compile": false, "split": "validation", "protocol": "elf_native_v1", "condition_policy": null, "sample_count": 3000, "prompt_count": 3000, "metrics": { "bleu": 14.166740978400506, "bleu_signature": "nrefs:1|case:lc|eff:yes|tok:13a|smooth:exp|version:2.5.1", "empty_fraction": 0.003, "rouge1": 43.23849851467062, "rouge1_sem": 0.3868893941941389, "rouge2": 21.77103708648444, "rouge2_sem": 0.31597545736717975, "rougeL": 38.020195846160334, "rougeL_sem": 0.3714766499770405, "reference_policy": "full raw reference; no target truncation; ROUGE F1 stemmer=true" }, "reason": null, "owt_main_grid": false }, { "run": "/home/ruizzhao/diffusion_baseline/results/elf-h200-gpu2/quality/wmt14/validation/steps_8/seed_42/compile_0", "status": "quality_valid", "task": "wmt14", "steps": 8, "seed": 42, "compile": false, "split": "validation", "protocol": "elf_native_v1", "condition_policy": null, "sample_count": 3000, "prompt_count": 3000, "metrics": { "bleu": 12.371998957336862, "bleu_signature": "nrefs:1|case:lc|eff:yes|tok:13a|smooth:exp|version:2.5.1", "empty_fraction": 0.0016666666666666668, "rouge1": 40.737713702371536, "rouge1_sem": 0.39718099505028703, "rouge2": 18.503359094937103, "rouge2_sem": 0.31636046822354447, "rougeL": 35.73263512485857, "rougeL_sem": 0.38040812372273436, "reference_policy": "full raw reference; no target truncation; ROUGE F1 stemmer=true" }, "reason": null, "owt_main_grid": false }, { "run": "/home/ruizzhao/diffusion_baseline/results/elf-h200-gpu2/quality/xsum/validation/steps_8/seed_42/compile_0", "status": "incomplete_generation", "task": "xsum", "steps": 8, "seed": 42, "compile": false, "split": "validation", "protocol": "elf_native_v1", "condition_policy": null }, { "run": "/home/ruizzhao/diffusion_baseline/results/elf-h200-gpu2-eos-v2/quality/wmt14/test/steps_32/seed_42/compile_0", "status": "quality_valid", "task": "wmt14", "steps": 32, "seed": 42, "compile": false, "split": "test", "protocol": "elf_native_v1", "condition_policy": "released_ids_or_raw_plus_eos_then_source_cap_v2", "sample_count": 3003, "prompt_count": 3003, "metrics": { "bleu": 25.866214222320252, "bleu_signature": "nrefs:1|case:lc|eff:yes|tok:13a|smooth:exp|version:2.5.1", "empty_fraction": 0.0, "rouge1": 61.13741235791628, "rouge1_sem": 0.29601964874970854, "rouge2": 35.120786021438725, "rouge2_sem": 0.35943648926011273, "rougeL": 56.174658046326265, "rougeL_sem": 0.3143682187275276, "reference_policy": "full raw reference; no target truncation; ROUGE F1 stemmer=true" }, "reason": null, "owt_main_grid": false }, { "run": "/home/ruizzhao/diffusion_baseline/results/elf-h200-gpu2-eos-v2/quality/wmt14/test/steps_64/seed_42/compile_0", "status": "quality_valid", "task": "wmt14", "steps": 64, "seed": 42, "compile": false, "split": "test", "protocol": "elf_native_v1", "condition_policy": "released_ids_or_raw_plus_eos_then_source_cap_v2", "sample_count": 3003, "prompt_count": 3003, "metrics": { "bleu": 26.34018303800428, "bleu_signature": "nrefs:1|case:lc|eff:yes|tok:13a|smooth:exp|version:2.5.1", "empty_fraction": 0.0, "rouge1": 61.583466649901546, "rouge1_sem": 0.29517765729412226, "rouge2": 35.67497246068508, "rouge2_sem": 0.3620526208155759, "rougeL": 56.64951939303449, "rougeL_sem": 0.31361281122446316, "reference_policy": "full raw reference; no target truncation; ROUGE F1 stemmer=true" }, "reason": null, "owt_main_grid": false }, { "run": "/home/ruizzhao/diffusion_baseline/results/elf-h200-gpu2-eos-v2/quality/wmt14/validation/steps_16/seed_42/compile_0", "status": "quality_valid", "task": "wmt14", "steps": 16, "seed": 42, "compile": false, "split": "validation", "protocol": "elf_native_v1", "condition_policy": "released_ids_or_raw_plus_eos_then_source_cap_v2", "sample_count": 3000, "prompt_count": 3000, "metrics": { "bleu": 25.331826445207003, "bleu_signature": "nrefs:1|case:lc|eff:yes|tok:13a|smooth:exp|version:2.5.1", "empty_fraction": 0.0, "rouge1": 59.377382971701095, "rouge1_sem": 0.3221949039548406, "rouge2": 33.420576225129246, "rouge2_sem": 0.38365593543635795, "rougeL": 54.55992649116913, "rougeL_sem": 0.33939428603113303, "reference_policy": "full raw reference; no target truncation; ROUGE F1 stemmer=true" }, "reason": null, "owt_main_grid": false }, { "run": "/home/ruizzhao/diffusion_baseline/results/elf-h200-gpu2-eos-v2/quality/wmt14/validation/steps_32/seed_42/compile_0", "status": "quality_valid", "task": "wmt14", "steps": 32, "seed": 42, "compile": false, "split": "validation", "protocol": "elf_native_v1", "condition_policy": "released_ids_or_raw_plus_eos_then_source_cap_v2", "sample_count": 3000, "prompt_count": 3000, "metrics": { "bleu": 26.142394757684823, "bleu_signature": "nrefs:1|case:lc|eff:yes|tok:13a|smooth:exp|version:2.5.1", "empty_fraction": 0.0, "rouge1": 60.10050570664322, "rouge1_sem": 0.3175978564046352, "rouge2": 34.1956402562575, "rouge2_sem": 0.3812216542018437, "rougeL": 55.2201139248711, "rougeL_sem": 0.336269109466868, "reference_policy": "full raw reference; no target truncation; ROUGE F1 stemmer=true" }, "reason": null, "owt_main_grid": false }, { "run": "/home/ruizzhao/diffusion_baseline/results/elf-h200-gpu2-eos-v2/quality/wmt14/validation/steps_64/seed_42/compile_0", "status": "quality_valid", "task": "wmt14", "steps": 64, "seed": 42, "compile": false, "split": "validation", "protocol": "elf_native_v1", "condition_policy": "released_ids_or_raw_plus_eos_then_source_cap_v2", "sample_count": 3000, "prompt_count": 3000, "metrics": { "bleu": 26.67410489997481, "bleu_signature": "nrefs:1|case:lc|eff:yes|tok:13a|smooth:exp|version:2.5.1", "empty_fraction": 0.0, "rouge1": 60.346232380335316, "rouge1_sem": 0.3173933819595248, "rouge2": 34.50467230871753, "rouge2_sem": 0.3814709446292345, "rougeL": 55.51389594651306, "rougeL_sem": 0.33536631844316006, "reference_policy": "full raw reference; no target truncation; ROUGE F1 stemmer=true" }, "reason": null, "owt_main_grid": false }, { "run": "/home/ruizzhao/diffusion_baseline/results/elf-h200-gpu2-eos-v2/quality/wmt14/validation/steps_8/seed_42/compile_0", "status": "quality_valid", "task": "wmt14", "steps": 8, "seed": 42, "compile": false, "split": "validation", "protocol": "elf_native_v1", "condition_policy": "released_ids_or_raw_plus_eos_then_source_cap_v2", "sample_count": 3000, "prompt_count": 3000, "metrics": { "bleu": 24.170511788509323, "bleu_signature": "nrefs:1|case:lc|eff:yes|tok:13a|smooth:exp|version:2.5.1", "empty_fraction": 0.0, "rouge1": 58.72488690372058, "rouge1_sem": 0.32242041054644116, "rouge2": 32.28881196492445, "rouge2_sem": 0.38058914383022696, "rougeL": 53.794457916073384, "rougeL_sem": 0.3399205830935264, "reference_policy": "full raw reference; no target truncation; ROUGE F1 stemmer=true" }, "reason": null, "owt_main_grid": false }, { "run": "/home/ruizzhao/diffusion_baseline/results/elf-h200-gpu2-eos-v2/quality/xsum/test/steps_32/seed_42/compile_0", "status": "quality_valid", "task": "xsum", "steps": 32, "seed": 42, "compile": false, "split": "test", "protocol": "elf_native_v1", "condition_policy": "released_ids_or_raw_plus_eos_then_source_cap_v2", "sample_count": 11334, "prompt_count": 11334, "metrics": { "bleu": 7.831639257982495, "bleu_signature": "nrefs:1|case:lc|eff:yes|tok:13a|smooth:exp|version:2.5.1", "empty_fraction": 0.0, "rouge1": 35.695546212670905, "rouge1_sem": 0.12592235515613065, "rouge2": 12.09416306826398, "rouge2_sem": 0.10862426941241783, "rougeL": 27.53447197401258, "rougeL_sem": 0.12135668682835216, "reference_policy": "full raw reference; no target truncation; ROUGE F1 stemmer=true" }, "reason": null, "owt_main_grid": false }, { "run": "/home/ruizzhao/diffusion_baseline/results/elf-h200-gpu2-eos-v2/quality/xsum/test/steps_64/seed_42/compile_0", "status": "quality_valid", "task": "xsum", "steps": 64, "seed": 42, "compile": false, "split": "test", "protocol": "elf_native_v1", "condition_policy": "released_ids_or_raw_plus_eos_then_source_cap_v2", "sample_count": 11334, "prompt_count": 11334, "metrics": { "bleu": 7.96428767569671, "bleu_signature": "nrefs:1|case:lc|eff:yes|tok:13a|smooth:exp|version:2.5.1", "empty_fraction": 0.0, "rouge1": 35.95022866839485, "rouge1_sem": 0.12588325780620363, "rouge2": 12.276741063589402, "rouge2_sem": 0.10902787335961733, "rougeL": 27.71153533477496, "rougeL_sem": 0.12096261329021324, "reference_policy": "full raw reference; no target truncation; ROUGE F1 stemmer=true" }, "reason": null, "owt_main_grid": false }, { "run": "/home/ruizzhao/diffusion_baseline/results/elf-h200-gpu2-eos-v2/quality/xsum/validation/steps_16/seed_42/compile_0", "status": "quality_valid", "task": "xsum", "steps": 16, "seed": 42, "compile": false, "split": "validation", "protocol": "elf_native_v1", "condition_policy": "released_ids_or_raw_plus_eos_then_source_cap_v2", "sample_count": 11332, "prompt_count": 11332, "metrics": { "bleu": 7.637624283432196, "bleu_signature": "nrefs:1|case:lc|eff:yes|tok:13a|smooth:exp|version:2.5.1", "empty_fraction": 0.0, "rouge1": 35.34846110226699, "rouge1_sem": 0.12540937524508475, "rouge2": 11.757763916859195, "rouge2_sem": 0.10711988085259284, "rougeL": 27.292067507641104, "rougeL_sem": 0.12083606443262047, "reference_policy": "full raw reference; no target truncation; ROUGE F1 stemmer=true" }, "reason": null, "owt_main_grid": false }, { "run": "/home/ruizzhao/diffusion_baseline/results/elf-h200-gpu2-eos-v2/quality/xsum/validation/steps_32/seed_42/compile_0", "status": "quality_valid", "task": "xsum", "steps": 32, "seed": 42, "compile": false, "split": "validation", "protocol": "elf_native_v1", "condition_policy": "released_ids_or_raw_plus_eos_then_source_cap_v2", "sample_count": 11332, "prompt_count": 11332, "metrics": { "bleu": 7.9401489561357845, "bleu_signature": "nrefs:1|case:lc|eff:yes|tok:13a|smooth:exp|version:2.5.1", "empty_fraction": 0.0, "rouge1": 35.84016178562892, "rouge1_sem": 0.12609865173225163, "rouge2": 12.17143430554345, "rouge2_sem": 0.10876330065212299, "rougeL": 27.65624888072885, "rougeL_sem": 0.12179568575390814, "reference_policy": "full raw reference; no target truncation; ROUGE F1 stemmer=true" }, "reason": null, "owt_main_grid": false }, { "run": "/home/ruizzhao/diffusion_baseline/results/elf-h200-gpu2-eos-v2/quality/xsum/validation/steps_64/seed_42/compile_0", "status": "quality_valid", "task": "xsum", "steps": 64, "seed": 42, "compile": false, "split": "validation", "protocol": "elf_native_v1", "condition_policy": "released_ids_or_raw_plus_eos_then_source_cap_v2", "sample_count": 11332, "prompt_count": 11332, "metrics": { "bleu": 8.077356963082332, "bleu_signature": "nrefs:1|case:lc|eff:yes|tok:13a|smooth:exp|version:2.5.1", "empty_fraction": 0.0, "rouge1": 36.02704578284659, "rouge1_sem": 0.12602619389183484, "rouge2": 12.352230792602231, "rouge2_sem": 0.1097850092292554, "rougeL": 27.816474304176275, "rougeL_sem": 0.12212453784989426, "reference_policy": "full raw reference; no target truncation; ROUGE F1 stemmer=true" }, "reason": null, "owt_main_grid": false }, { "run": "/home/ruizzhao/diffusion_baseline/results/elf-h200-gpu2-eos-v2/quality/xsum/validation/steps_8/seed_42/compile_0", "status": "quality_valid", "task": "xsum", "steps": 8, "seed": 42, "compile": false, "split": "validation", "protocol": "elf_native_v1", "condition_policy": "released_ids_or_raw_plus_eos_then_source_cap_v2", "sample_count": 11332, "prompt_count": 11332, "metrics": { "bleu": 7.158670323167527, "bleu_signature": "nrefs:1|case:lc|eff:yes|tok:13a|smooth:exp|version:2.5.1", "empty_fraction": 0.0, "rouge1": 34.37045819615643, "rouge1_sem": 0.12489606538226775, "rouge2": 11.156725532974084, "rouge2_sem": 0.10460478239947939, "rougeL": 26.68636066957463, "rougeL_sem": 0.1199117588273479, "reference_policy": "full raw reference; no target truncation; ROUGE F1 stemmer=true" }, "reason": null, "owt_main_grid": false }, { "run": "/home/ruizzhao/diffusion_baseline/results/elf-h200-gpu2-sanity/quality/owt/unconditional/steps_32/seed_42/compile_0", "status": "quality_valid", "task": "owt", "steps": 32, "seed": 42, "compile": false, "split": "unconditional", "protocol": "elf_native_v1", "condition_policy": null, "sample_count": 1000, "prompt_count": 1000, "metrics": { "gpt2_length_mean": 946.265, "gpt2_length_min": 828, "empty_fraction": 0.0, "below_64_tokens_fraction": 0.0, "above_1024_tokens_fraction": 0.002, "entropy_gpt2_nats": 5.16711030983225, "generative_ppl": { "perplexity": 24.423866227405973, "total_nll": 3020584.65234375, "valid_token_count": 945244, "sample_count": 1000, "batch_size": 8, "context_length": 1024, "truncation": "right", "model_id": "gpt2-large", "model_revision": "32b71b12589c2f8d625668d2335a01cac3249519", "tokenizer_id": "gpt2", "tokenizer_revision": "607a30d783dfa663caf39e06633721c8d4cfcd7e", "aggregation": "sum_nll_over_sum_valid_next_tokens" }, "entropy_policy": "mean per-sample natural-log entropy on retokenized GPT-2 suffix/text; same PPL length cap" }, "reason": null, "owt_main_grid": false }, { "run": "/home/ruizzhao/diffusion_baseline/results/elf-h200-gpu2-sanity/quality/wmt14/official-validation/steps_64/seed_42/compile_0", "status": "quality_valid", "task": "wmt14", "steps": 64, "seed": 42, "compile": false, "split": "official-validation", "protocol": "elf_native_v1", "condition_policy": null, "sample_count": 1000, "prompt_count": 1000, "metrics": { "bleu": 25.79055916774286, "bleu_signature": "nrefs:1|case:lc|eff:yes|tok:13a|smooth:exp|version:2.5.1", "empty_fraction": 0.0, "rouge1": 59.23724443864586, "rouge1_sem": 0.5806150834970896, "rouge2": 33.28725815818107, "rouge2_sem": 0.675281292229775, "rougeL": 53.84961838511511, "rougeL_sem": 0.6090158681808834, "reference_policy": "full raw reference; no target truncation; ROUGE F1 stemmer=true" }, "reason": null, "owt_main_grid": false }, { "run": "/home/ruizzhao/diffusion_baseline/results/elf-h200-gpu2-sanity/quality/xsum/official-validation/steps_64/seed_42/compile_0", "status": "quality_valid", "task": "xsum", "steps": 64, "seed": 42, "compile": false, "split": "official-validation", "protocol": "elf_native_v1", "condition_policy": null, "sample_count": 1000, "prompt_count": 1000, "metrics": { "bleu": 8.084246853107375, "bleu_signature": "nrefs:1|case:lc|eff:yes|tok:13a|smooth:exp|version:2.5.1", "empty_fraction": 0.0, "rouge1": 36.31256063962687, "rouge1_sem": 0.43244218487081704, "rouge2": 12.346405181311322, "rouge2_sem": 0.366236795370104, "rougeL": 28.111495835498143, "rougeL_sem": 0.41366115435629475, "reference_policy": "full raw reference; no target truncation; ROUGE F1 stemmer=true" }, "reason": null, "owt_main_grid": false } ], "owt_steps_without_valid_result": [] }
{
"schema": "elf-progress-v1",
"runs": [
{
"run": "/home/ruizzhao/diffusion_baseline/results/elf-gpu01/quality/owt/unconditional/steps_1024/seed_42/compile_0",
"status": "quality_valid",
"task": "owt",
"steps": 1024,
"seed": 42,
"compile": false,
"split": "unconditional",
"protocol": "elf_native_v1",
"condition_policy": "none",
"sample_count": 1024,
"prompt_count": 1024,
"metrics": {
"gpt2_length_mean": 920.2958984375,
"gpt2_length_min": 728,
"empty_fraction": 0.0,
"below_64_tokens_fraction": 0.0,
"above_1024_tokens_fraction": 0.009765625,
"entropy_gpt2_nats": 4.742035427877739,
"generative_ppl": {
"perplexity": 7.732375373221586,
"total_nll": 1925010.6435546875,
"valid_token_count": 941134,
"sample_count": 1024,
"batch_size": 8,
"context_length": 1024,
"truncation": "right",
"model_id": "gpt2-large",
"model_revision": "32b71b12589c2f8d625668d2335a01cac3249519",
"tokenizer_id": "gpt2",
"tokenizer_revision": "607a30d783dfa663caf39e06633721c8d4cfcd7e",
"aggregation": "sum_nll_over_sum_valid_next_tokens"
},
"entropy_policy": "mean per-sample natural-log entropy on retokenized GPT-2 suffix/text; same PPL length cap"
},
"reason": null,
"owt_main_grid": true
},
{
"run": "/home/ruizzhao/diffusion_baseline/results/elf-gpu01/quality/owt-prefix/c64-text-t5/steps_1/seed_42/compile_0",
"status": "quality_valid",
"task": "owt-prefix",
"steps": 1,
"seed": 42,
"compile": false,
"split": "c64-text-t5",
"protocol": "c64_text_t5_v1",
"condition_policy": "c64_raw_no_special_tokens_v1",
"sample_count": 2048,
"prompt_count": 1024,
"metrics": {
"gpt2_length_mean": 1196.43359375,
"gpt2_length_min": 297,
"empty_fraction": 0.0,
"below_64_tokens_fraction": 0.0,
"above_1024_tokens_fraction": 0.857421875,
"entropy_gpt2_nats": 1.5581149953523385,
"conditional_ppl": {
"perplexity": 15.67715969847609,
"total_nll": 180368.4974975586,
"valid_token_count": 65536,
"sample_count": 0,
"batch_size": 0,
"context_length": 1024,
"truncation": "right",
"model_id": "gpt2-large",
"model_revision": "",
"tokenizer_id": "gpt2",
"tokenizer_revision": "",
"aggregation": "sum_nll_over_sum_valid_next_tokens"
},
"reference_conditional_ppl": {
"perplexity": 17.09677422584074,
"total_nll": 185955.798828125,
"valid_token_count": 65503,
"sample_count": 0,
"batch_size": 0,
"context_length": 1024,
"truncation": "right",
"model_id": "gpt2-large",
"model_revision": "",
"tokenizer_id": "gpt2",
"tokenizer_revision": "",
"aggregation": "sum_nll_over_sum_valid_next_tokens"
},
"conditional_ppl_policy": "separate GPT-2 prefix/suffix tokenization; prefix targets masked",
"entropy_all_completions_gpt2_nats": 1.5677205978966589,
"grouped_self_bleu": 0.7343077781716898,
"entropy_policy": "mean per-sample natural-log entropy on retokenized GPT-2 suffix/text; same PPL length cap"
},
"reason": null,
"owt_main_grid": false
},
{
"run": "/home/ruizzhao/diffusion_baseline/results/elf-gpu01/quality/owt-prefix/c64-text-t5/steps_1024/seed_42/compile_0",
"status": "quality_valid",
"task": "owt-prefix",