beeper-rose-v3 / config.json
AbstractPhil's picture
Upload folder using huggingface_hub
2dc9c8f verified
{
"name": "Rose-Beeper",
"context": 512,
"vocab_size": 8192,
"dim": 512,
"n_layers": 6,
"n_heads": 8,
"mlp_ratio": 4.0,
"dropout": 0.0,
"resid_dropout": 0.1,
"grad_checkpoint": false,
"compile_model": false,
"tokenizer_path": "beeper.tokenizer.json",
"add_bos_eos": true,
"span_corrupt_frac": 0.0,
"val_ratio": 0.01,
"test_ratio": 0.01,
"max_rows_per_dataset": null,
"dataset_cache_verbose": true,
"batch_size": 32,
"grad_accum_steps": 1,
"epochs": 10,
"lr": 0.0003,
"betas": [
0.9,
0.95
],
"weight_decay": 0.1,
"warmup_steps": 500,
"max_steps": null,
"clip_grad": 1.0,
"min_lr": 1e-06,
"label_smoothing": 0.0,
"mixed_precision": "bf16",
"log_dir": "./runs/rose_beeper",
"log_interval": 50,
"ckpt_dir": "./beeper_checkpoints",
"export_dir": "./beeper_export",
"temperature": 0.9,
"top_k": 40,
"top_p": 0.9,
"repetition_penalty": 1.1,
"presence_penalty": 0.6,
"frequency_penalty": 0.0,
"hf_repo": "AbstractPhil/beeper-rose-v3",
"upload_to_hub": true,
"pent_level": "medium",
"lambda_contrast": 0.25,
"pent_min_edge": 0.5,
"pent_temp": 0.1,
"contrast_warmup": 800,
"lambda_rose": 0.1,
"rose_scale": 1.8,
"lambda_geom_sep": 0.5,
"geom_sep_margin": 0.9,
"lambda_geom": 0.3,
"lambda_geom_angle": 0.8,
"lambda_geom_var": 0.3,
"lambda_geom_edge": 0.3,
"lambda_geom_vol": 0.6,
"lambda_geom_minrel": 1.0,
"geom_min_edge_rel": 0.6,
"geom_vol_lower_frac": 0.85,
"geom_sample_classes": 64,
"geom_sample_k": 64,
"seed": 1337,
"corpus": [
{
"name": "TinyStories",
"path": "roneneldan/TinyStories",
"split": "train[10%:20%]",
"weight": 0.1,
"dialect": [
0.6,
0.1,
0.05,
0.05,
0.2
]
},
{
"name": "WikipediaEN",
"path": "wikimedia/wikipedia",
"config": "20231101.en",
"split": "train[1%:2%]",
"weight": 0.6,
"dialect": [
0.12,
0.58,
0.1,
0.1,
0.1
]
},
{
"name": "AGNews",
"path": "ag_news",
"split": "train[:50%]",
"weight": 0.4,
"dialect": [
0.2,
0.5,
0.1,
0.1,
0.1
]
},
{
"name": "GSM8K",
"path": "openai/gsm8k",
"config": "main",
"split": "train[10%:20%]",
"weight": 0.6,
"dialect": [
0.1,
0.15,
0.5,
0.15,
0.1
]
},
{
"name": "AI2-ARC-Easy",
"path": "allenai/ai2_arc",
"config": "ARC-Easy",
"split": "train[10%:20%]",
"weight": 0.5,
"dialect": [
0.05,
0.15,
0.4,
0.25,
0.15
]
},
{
"name": "HH-RLHF",
"path": "Anthropic/hh-rlhf",
"split": "train[1%:2%]",
"weight": 0.6,
"dialect": [
0.1,
0.25,
0.2,
0.25,
0.2
]
},
{
"name": "SVAMP",
"path": "ChilleD/SVAMP",
"split": "train",
"weight": 0.45,
"dialect": [
0.1,
0.15,
0.55,
0.15,
0.05
]
},
{
"name": "MATH-500",
"path": "HuggingFaceH4/MATH-500",
"split": "test",
"weight": 0.15,
"dialect": [
0.05,
0.15,
0.6,
0.15,
0.05
]
},
{
"name": "SEP",
"path": "AiresPucrs/stanford-encyclopedia-philosophy",
"split": "train",
"weight": 0.3,
"dialect": [
0.05,
0.45,
0.18,
0.22,
0.1
]
},
{
"name": "ETHICS-commonsense",
"path": "hendrycks/ethics",
"config": "commonsense",
"split": "train",
"weight": 0.45,
"dialect": [
0.1,
0.3,
0.18,
0.24,
0.18
]
},
{
"name": "ETHICS-deontology",
"path": "hendrycks/ethics",
"config": "deontology",
"split": "train",
"weight": 0.35,
"dialect": [
0.1,
0.3,
0.18,
0.24,
0.18
]
},
{
"name": "ETHICS-justice",
"path": "hendrycks/ethics",
"config": "justice",
"split": "train",
"weight": 0.35,
"dialect": [
0.1,
0.3,
0.18,
0.24,
0.18
]
},
{
"name": "ETHICS-utilitarianism",
"path": "hendrycks/ethics",
"config": "utilitarianism",
"split": "train",
"weight": 0.35,
"dialect": [
0.1,
0.3,
0.18,
0.24,
0.18
]
},
{
"name": "ETHICS-virtue",
"path": "hendrycks/ethics",
"config": "virtue",
"split": "train",
"weight": 0.35,
"dialect": [
0.1,
0.3,
0.18,
0.24,
0.18
]
},
{
"name": "SocialChem101",
"path": "allenai/social-chemistry-101",
"split": "train",
"weight": 0.65,
"dialect": [
0.15,
0.25,
0.2,
0.2,
0.2
]
},
{
"name": "MoralStories",
"path": "demelin/moral_stories",
"split": "train",
"weight": 0.35,
"dialect": [
0.2,
0.2,
0.2,
0.2,
0.2
]
},
{
"name": "ART-AbductiveNLI",
"path": "allenai/art",
"split": "train",
"weight": 0.3,
"dialect": [
0.05,
0.2,
0.45,
0.2,
0.1
]
},
{
"name": "EntailmentBankV3",
"path": "ariesutiono/entailment-bank-v3",
"split": "train",
"weight": 0.35,
"dialect": [
0.05,
0.25,
0.45,
0.15,
0.1
]
},
{
"name": "LogiQA2.0NLI",
"path": "tasksource/logiqa-2.0-nli",
"split": "train",
"weight": 0.45,
"dialect": [
0.05,
0.25,
0.45,
0.15,
0.1
]
},
{
"name": "TruthfulQA-MC",
"path": "EleutherAI/truthful_qa_mc",
"split": "validation",
"weight": 0.25,
"dialect": [
0.05,
0.35,
0.25,
0.25,
0.1
]
},
{
"name": "VUA20-Metaphor",
"path": "CreativeLang/vua20_metaphor",
"split": "train[:5%]",
"weight": 0.3,
"dialect": [
0.3,
0.1,
0.1,
0.15,
0.35
]
}
],
"capoera": {
"enable": true,
"topic_bins": 512,
"mood_bins": 7
},
"_ok_entries": [
{
"name": "TinyStories",
"path": "roneneldan/TinyStories",
"split": "train[10%:20%]",
"weight": 0.1,
"dialect": [
0.6000000238418579,
0.10000000149011612,
0.05000000074505806,
0.05000000074505806,
0.20000000298023224
],
"class_id": 0,
"p": 0.012195121951219514
},
{
"name": "WikipediaEN",
"path": "wikimedia/wikipedia",
"config": "20231101.en",
"split": "train[1%:2%]",
"weight": 0.6,
"dialect": [
0.11999999731779099,
0.5799999833106995,
0.10000000149011612,
0.10000000149011612,
0.10000000149011612
],
"class_id": 1,
"p": 0.07317073170731708
},
{
"name": "AGNews",
"path": "ag_news",
"split": "train[:50%]",
"weight": 0.4,
"dialect": [
0.20000000298023224,
0.5,
0.10000000149011612,
0.10000000149011612,
0.10000000149011612
],
"class_id": 2,
"p": 0.04878048780487806
},
{
"name": "GSM8K",
"path": "openai/gsm8k",
"config": "main",
"split": "train[10%:20%]",
"weight": 0.6,
"dialect": [
0.10000000149011612,
0.15000000596046448,
0.5,
0.15000000596046448,
0.10000000149011612
],
"class_id": 3,
"p": 0.07317073170731708
},
{
"name": "AI2-ARC-Easy",
"path": "allenai/ai2_arc",
"config": "ARC-Easy",
"split": "train[10%:20%]",
"weight": 0.5,
"dialect": [
0.05000000074505806,
0.15000000596046448,
0.4000000059604645,
0.25,
0.15000000596046448
],
"class_id": 4,
"p": 0.06097560975609757
},
{
"name": "HH-RLHF",
"path": "Anthropic/hh-rlhf",
"split": "train[1%:2%]",
"weight": 0.6,
"dialect": [
0.10000000149011612,
0.25,
0.20000000298023224,
0.25,
0.20000000298023224
],
"class_id": 5,
"p": 0.07317073170731708
},
{
"name": "SVAMP",
"path": "ChilleD/SVAMP",
"split": "train",
"weight": 0.45,
"dialect": [
0.10000000149011612,
0.15000000596046448,
0.550000011920929,
0.15000000596046448,
0.05000000074505806
],
"class_id": 6,
"p": 0.05487804878048781
},
{
"name": "MATH-500",
"path": "HuggingFaceH4/MATH-500",
"split": "test",
"weight": 0.15,
"dialect": [
0.05000000074505806,
0.15000000596046448,
0.6000000238418579,
0.15000000596046448,
0.05000000074505806
],
"class_id": 7,
"p": 0.01829268292682927
},
{
"name": "SEP",
"path": "AiresPucrs/stanford-encyclopedia-philosophy",
"split": "train",
"weight": 0.3,
"dialect": [
0.05000000074505806,
0.44999998807907104,
0.18000000715255737,
0.2199999988079071,
0.10000000149011612
],
"class_id": 8,
"p": 0.03658536585365854
},
{
"name": "ETHICS-commonsense",
"path": "hendrycks/ethics",
"config": "commonsense",
"split": "train",
"weight": 0.45,
"dialect": [
0.10000000149011612,
0.30000001192092896,
0.18000000715255737,
0.23999999463558197,
0.18000000715255737
],
"class_id": 9,
"p": 0.05487804878048781
},
{
"name": "ETHICS-deontology",
"path": "hendrycks/ethics",
"config": "deontology",
"split": "train",
"weight": 0.35,
"dialect": [
0.10000000149011612,
0.30000001192092896,
0.18000000715255737,
0.23999999463558197,
0.18000000715255737
],
"class_id": 10,
"p": 0.042682926829268296
},
{
"name": "ETHICS-justice",
"path": "hendrycks/ethics",
"config": "justice",
"split": "train",
"weight": 0.35,
"dialect": [
0.10000000149011612,
0.30000001192092896,
0.18000000715255737,
0.23999999463558197,
0.18000000715255737
],
"class_id": 11,
"p": 0.042682926829268296
},
{
"name": "ETHICS-utilitarianism",
"path": "hendrycks/ethics",
"config": "utilitarianism",
"split": "train",
"weight": 0.35,
"dialect": [
0.10000000149011612,
0.30000001192092896,
0.18000000715255737,
0.23999999463558197,
0.18000000715255737
],
"class_id": 12,
"p": 0.042682926829268296
},
{
"name": "ETHICS-virtue",
"path": "hendrycks/ethics",
"config": "virtue",
"split": "train",
"weight": 0.35,
"dialect": [
0.10000000149011612,
0.30000001192092896,
0.18000000715255737,
0.23999999463558197,
0.18000000715255737
],
"class_id": 13,
"p": 0.042682926829268296
},
{
"name": "SocialChem101",
"path": "allenai/social-chemistry-101",
"split": "train",
"weight": 0.65,
"dialect": [
0.15000000596046448,
0.25,
0.20000000298023224,
0.20000000298023224,
0.20000000298023224
],
"class_id": 14,
"p": 0.07926829268292684
},
{
"name": "MoralStories",
"path": "demelin/moral_stories",
"split": "train",
"weight": 0.35,
"dialect": [
0.20000000298023224,
0.20000000298023224,
0.20000000298023224,
0.20000000298023224,
0.20000000298023224
],
"class_id": 15,
"p": 0.042682926829268296
},
{
"name": "ART-AbductiveNLI",
"path": "allenai/art",
"split": "train",
"weight": 0.3,
"dialect": [
0.05000000074505806,
0.20000000298023224,
0.44999998807907104,
0.20000000298023224,
0.10000000149011612
],
"class_id": 16,
"p": 0.03658536585365854
},
{
"name": "EntailmentBankV3",
"path": "ariesutiono/entailment-bank-v3",
"split": "train",
"weight": 0.35,
"dialect": [
0.05000000074505806,
0.25,
0.44999998807907104,
0.15000000596046448,
0.10000000149011612
],
"class_id": 17,
"p": 0.042682926829268296
},
{
"name": "LogiQA2.0NLI",
"path": "tasksource/logiqa-2.0-nli",
"split": "train",
"weight": 0.45,
"dialect": [
0.05000000074505806,
0.25,
0.44999998807907104,
0.15000000596046448,
0.10000000149011612
],
"class_id": 18,
"p": 0.05487804878048781
},
{
"name": "TruthfulQA-MC",
"path": "EleutherAI/truthful_qa_mc",
"split": "validation",
"weight": 0.25,
"dialect": [
0.05000000074505806,
0.3499999940395355,
0.25,
0.25,
0.10000000149011612
],
"class_id": 19,
"p": 0.030487804878048783
},
{
"name": "VUA20-Metaphor",
"path": "CreativeLang/vua20_metaphor",
"split": "train[:5%]",
"weight": 0.3,
"dialect": [
0.30000001192092896,
0.10000000149011612,
0.10000000149011612,
0.15000000596046448,
0.3499999940395355
],
"class_id": 20,
"p": 0.03658536585365854
}
],
"_alive_entries": [
{
"name": "TinyStories",
"path": "roneneldan/TinyStories",
"split": "train[10%:20%]",
"weight": 0.1,
"dialect": [
0.6000000238418579,
0.10000000149011612,
0.05000000074505806,
0.05000000074505806,
0.20000000298023224
],
"class_id": 0,
"p": 0.012195121951219514
},
{
"name": "WikipediaEN",
"path": "wikimedia/wikipedia",
"config": "20231101.en",
"split": "train[1%:2%]",
"weight": 0.6,
"dialect": [
0.11999999731779099,
0.5799999833106995,
0.10000000149011612,
0.10000000149011612,
0.10000000149011612
],
"class_id": 1,
"p": 0.07317073170731708
},
{
"name": "AGNews",
"path": "ag_news",
"split": "train[:50%]",
"weight": 0.4,
"dialect": [
0.20000000298023224,
0.5,
0.10000000149011612,
0.10000000149011612,
0.10000000149011612
],
"class_id": 2,
"p": 0.04878048780487806
},
{
"name": "GSM8K",
"path": "openai/gsm8k",
"config": "main",
"split": "train[10%:20%]",
"weight": 0.6,
"dialect": [
0.10000000149011612,
0.15000000596046448,
0.5,
0.15000000596046448,
0.10000000149011612
],
"class_id": 3,
"p": 0.07317073170731708
},
{
"name": "AI2-ARC-Easy",
"path": "allenai/ai2_arc",
"config": "ARC-Easy",
"split": "train[10%:20%]",
"weight": 0.5,
"dialect": [
0.05000000074505806,
0.15000000596046448,
0.4000000059604645,
0.25,
0.15000000596046448
],
"class_id": 4,
"p": 0.06097560975609757
},
{
"name": "HH-RLHF",
"path": "Anthropic/hh-rlhf",
"split": "train[1%:2%]",
"weight": 0.6,
"dialect": [
0.10000000149011612,
0.25,
0.20000000298023224,
0.25,
0.20000000298023224
],
"class_id": 5,
"p": 0.07317073170731708
},
{
"name": "SVAMP",
"path": "ChilleD/SVAMP",
"split": "train",
"weight": 0.45,
"dialect": [
0.10000000149011612,
0.15000000596046448,
0.550000011920929,
0.15000000596046448,
0.05000000074505806
],
"class_id": 6,
"p": 0.05487804878048781
},
{
"name": "MATH-500",
"path": "HuggingFaceH4/MATH-500",
"split": "test",
"weight": 0.15,
"dialect": [
0.05000000074505806,
0.15000000596046448,
0.6000000238418579,
0.15000000596046448,
0.05000000074505806
],
"class_id": 7,
"p": 0.01829268292682927
},
{
"name": "SEP",
"path": "AiresPucrs/stanford-encyclopedia-philosophy",
"split": "train",
"weight": 0.3,
"dialect": [
0.05000000074505806,
0.44999998807907104,
0.18000000715255737,
0.2199999988079071,
0.10000000149011612
],
"class_id": 8,
"p": 0.03658536585365854
},
{
"name": "ETHICS-commonsense",
"path": "hendrycks/ethics",
"config": "commonsense",
"split": "train",
"weight": 0.45,
"dialect": [
0.10000000149011612,
0.30000001192092896,
0.18000000715255737,
0.23999999463558197,
0.18000000715255737
],
"class_id": 9,
"p": 0.05487804878048781
},
{
"name": "ETHICS-deontology",
"path": "hendrycks/ethics",
"config": "deontology",
"split": "train",
"weight": 0.35,
"dialect": [
0.10000000149011612,
0.30000001192092896,
0.18000000715255737,
0.23999999463558197,
0.18000000715255737
],
"class_id": 10,
"p": 0.042682926829268296
},
{
"name": "ETHICS-justice",
"path": "hendrycks/ethics",
"config": "justice",
"split": "train",
"weight": 0.35,
"dialect": [
0.10000000149011612,
0.30000001192092896,
0.18000000715255737,
0.23999999463558197,
0.18000000715255737
],
"class_id": 11,
"p": 0.042682926829268296
},
{
"name": "ETHICS-utilitarianism",
"path": "hendrycks/ethics",
"config": "utilitarianism",
"split": "train",
"weight": 0.35,
"dialect": [
0.10000000149011612,
0.30000001192092896,
0.18000000715255737,
0.23999999463558197,
0.18000000715255737
],
"class_id": 12,
"p": 0.042682926829268296
},
{
"name": "ETHICS-virtue",
"path": "hendrycks/ethics",
"config": "virtue",
"split": "train",
"weight": 0.35,
"dialect": [
0.10000000149011612,
0.30000001192092896,
0.18000000715255737,
0.23999999463558197,
0.18000000715255737
],
"class_id": 13,
"p": 0.042682926829268296
},
{
"name": "SocialChem101",
"path": "allenai/social-chemistry-101",
"split": "train",
"weight": 0.65,
"dialect": [
0.15000000596046448,
0.25,
0.20000000298023224,
0.20000000298023224,
0.20000000298023224
],
"class_id": 14,
"p": 0.07926829268292684
},
{
"name": "MoralStories",
"path": "demelin/moral_stories",
"split": "train",
"weight": 0.35,
"dialect": [
0.20000000298023224,
0.20000000298023224,
0.20000000298023224,
0.20000000298023224,
0.20000000298023224
],
"class_id": 15,
"p": 0.042682926829268296
},
{
"name": "ART-AbductiveNLI",
"path": "allenai/art",
"split": "train",
"weight": 0.3,
"dialect": [
0.05000000074505806,
0.20000000298023224,
0.44999998807907104,
0.20000000298023224,
0.10000000149011612
],
"class_id": 16,
"p": 0.03658536585365854
},
{
"name": "LogiQA2.0NLI",
"path": "tasksource/logiqa-2.0-nli",
"split": "train",
"weight": 0.45,
"dialect": [
0.05000000074505806,
0.25,
0.44999998807907104,
0.15000000596046448,
0.10000000149011612
],
"class_id": 18,
"p": 0.05487804878048781
},
{
"name": "TruthfulQA-MC",
"path": "EleutherAI/truthful_qa_mc",
"split": "validation",
"weight": 0.25,
"dialect": [
0.05000000074505806,
0.3499999940395355,
0.25,
0.25,
0.10000000149011612
],
"class_id": 19,
"p": 0.030487804878048783
},
{
"name": "VUA20-Metaphor",
"path": "CreativeLang/vua20_metaphor",
"split": "train[:5%]",
"weight": 0.3,
"dialect": [
0.30000001192092896,
0.10000000149011612,
0.10000000149011612,
0.15000000596046448,
0.3499999940395355
],
"class_id": 20,
"p": 0.03658536585365854
}
]
}