Upload folder using huggingface_hub

beeper_final.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4695a25162370f3bb519f2419bd6d7f11f7b0b043b1eb77a2addc34f18372ce4
+oid sha256:102908f72f27df29d188876da62af337ab02151ee251dbeaac4fdf9ca9c5e20d
 size 116810550

beeper_final.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:058c19b4326e05820b694c1377235963dc099362bdb0a97baff8669997694eb6
+oid sha256:7bd81ee9ff6a23a8b3f370f19afdc43fc4991bed90b523a797a72d04e2ac4b4c
 size 116795584

config.json

@@ -17,7 +17,7 @@
   "test_ratio": 0.01,
   "max_rows_per_dataset": null,
   "dataset_cache_verbose": true,
-  "batch_size": 32,
+  "batch_size": 128,
   "grad_accum_steps": 1,
   "epochs": 10,
   "lr": 0.0003,
@@ -71,7 +71,7 @@
     {
       "name": "TinyStories",
       "path": "roneneldan/TinyStories",
-      "split": "train[20%:30%]",
+      "split": "train[30%:50%]",
       "weight": 0.1,
       "dialect": [
         0.6,
@@ -85,8 +85,8 @@
       "name": "WikipediaEN",
       "path": "wikimedia/wikipedia",
       "config": "20231101.en",
-      "split": "train[2%:5%]",
-      "weight": 0.4,
+      "split": "train[5%:15%]",
+      "weight": 0.5,
       "dialect": [
         0.12,
         0.58,
@@ -98,7 +98,7 @@
     {
       "name": "AGNews",
       "path": "ag_news",
-      "split": "train[50%:]",
+      "split": "train[:]",
       "weight": 0.1,
       "dialect": [
         0.2,
@@ -112,7 +112,7 @@
       "name": "GSM8K",
       "path": "openai/gsm8k",
       "config": "main",
-      "split": "train[20%:40%]",
+      "split": "train[40%:60%]",
       "weight": 0.6,
       "dialect": [
         0.1,
@@ -126,8 +126,8 @@
       "name": "AI2-ARC-Easy",
       "path": "allenai/ai2_arc",
       "config": "ARC-Easy",
-      "split": "train[20%:30%]",
-      "weight": 0.5,
+      "split": "train[30%:60%]",
+      "weight": 0.6,
       "dialect": [
         0.05,
         0.15,
@@ -139,8 +139,8 @@
     {
       "name": "HH-RLHF",
       "path": "Anthropic/hh-rlhf",
-      "split": "train[2%:5%]",
-      "weight": 0.4,
+      "split": "train[5%:10%]",
+      "weight": 0.5,
       "dialect": [
         0.1,
         0.25,
@@ -314,7 +314,7 @@
       "name": "LogiQA2.0NLI",
       "path": "tasksource/logiqa-2.0-nli",
       "split": "train",
-      "weight": 0.65,
+      "weight": 0.45,
       "dialect": [
         0.05,
         0.25,
@@ -327,7 +327,7 @@
       "name": "TruthfulQA-MC",
       "path": "EleutherAI/truthful_qa_mc",
       "split": "validation",
-      "weight": 0.65,
+      "weight": 0.45,
       "dialect": [
         0.05,
         0.35,
@@ -359,7 +359,7 @@
     {
       "name": "TinyStories",
       "path": "roneneldan/TinyStories",
-      "split": "train[20%:30%]",
+      "split": "train[30%:50%]",
       "weight": 0.1,
       "dialect": [
         0.6000000238418579,
@@ -369,14 +369,14 @@
         0.20000000298023224
       ],
       "class_id": 0,
-      "p": 0.01234567901234568
+      "p": 0.012500000000000002
     },
     {
       "name": "WikipediaEN",
       "path": "wikimedia/wikipedia",
       "config": "20231101.en",
-      "split": "train[2%:5%]",
-      "weight": 0.4,
+      "split": "train[5%:15%]",
+      "weight": 0.5,
       "dialect": [
         0.11999999731779099,
         0.5799999833106995,
@@ -385,12 +385,12 @@
         0.10000000149011612
       ],
       "class_id": 1,
-      "p": 0.04938271604938272
+      "p": 0.06250000000000001
     },
     {
       "name": "AGNews",
       "path": "ag_news",
-      "split": "train[50%:]",
+      "split": "train[:]",
       "weight": 0.1,
       "dialect": [
         0.20000000298023224,
@@ -400,13 +400,13 @@
         0.10000000149011612
       ],
       "class_id": 2,
-      "p": 0.01234567901234568
+      "p": 0.012500000000000002
     },
     {
       "name": "GSM8K",
       "path": "openai/gsm8k",
       "config": "main",
-      "split": "train[20%:40%]",
+      "split": "train[40%:60%]",
       "weight": 0.6,
       "dialect": [
         0.10000000149011612,
@@ -416,14 +416,14 @@
         0.10000000149011612
       ],
       "class_id": 3,
-      "p": 0.07407407407407407
+      "p": 0.07500000000000001
     },
     {
       "name": "AI2-ARC-Easy",
       "path": "allenai/ai2_arc",
       "config": "ARC-Easy",
-      "split": "train[20%:30%]",
-      "weight": 0.5,
+      "split": "train[30%:60%]",
+      "weight": 0.6,
       "dialect": [
         0.05000000074505806,
         0.15000000596046448,
@@ -432,13 +432,13 @@
         0.15000000596046448
       ],
       "class_id": 4,
-      "p": 0.0617283950617284
+      "p": 0.07500000000000001
     },
     {
       "name": "HH-RLHF",
       "path": "Anthropic/hh-rlhf",
-      "split": "train[2%:5%]",
-      "weight": 0.4,
+      "split": "train[5%:10%]",
+      "weight": 0.5,
       "dialect": [
         0.10000000149011612,
         0.25,
@@ -447,7 +447,7 @@
         0.20000000298023224
       ],
       "class_id": 5,
-      "p": 0.04938271604938272
+      "p": 0.06250000000000001
     },
     {
       "name": "SVAMP",
@@ -462,7 +462,7 @@
         0.05000000074505806
       ],
       "class_id": 6,
-      "p": 0.0308641975308642
+      "p": 0.03125000000000001
     },
     {
       "name": "MATH-500",
@@ -477,7 +477,7 @@
         0.05000000074505806
       ],
       "class_id": 7,
-      "p": 0.0308641975308642
+      "p": 0.03125000000000001
     },
     {
       "name": "SEP",
@@ -492,7 +492,7 @@
         0.10000000149011612
       ],
       "class_id": 8,
-      "p": 0.037037037037037035
+      "p": 0.037500000000000006
     },
     {
       "name": "ETHICS-commonsense",
@@ -508,7 +508,7 @@
         0.18000000715255737
       ],
       "class_id": 9,
-      "p": 0.05555555555555556
+      "p": 0.05625000000000001
     },
     {
       "name": "ETHICS-deontology",
@@ -524,7 +524,7 @@
         0.18000000715255737
       ],
       "class_id": 10,
-      "p": 0.043209876543209874
+      "p": 0.043750000000000004
     },
     {
       "name": "ETHICS-justice",
@@ -540,7 +540,7 @@
         0.18000000715255737
       ],
       "class_id": 11,
-      "p": 0.043209876543209874
+      "p": 0.043750000000000004
     },
     {
       "name": "ETHICS-utilitarianism",
@@ -556,7 +556,7 @@
         0.18000000715255737
       ],
       "class_id": 12,
-      "p": 0.043209876543209874
+      "p": 0.043750000000000004
     },
     {
       "name": "ETHICS-virtue",
@@ -572,7 +572,7 @@
         0.18000000715255737
       ],
       "class_id": 13,
-      "p": 0.043209876543209874
+      "p": 0.043750000000000004
     },
     {
       "name": "SocialChem101",
@@ -587,7 +587,7 @@
         0.20000000298023224
       ],
       "class_id": 14,
-      "p": 0.08024691358024692
+      "p": 0.08125000000000002
     },
     {
       "name": "MoralStories",
@@ -602,7 +602,7 @@
         0.20000000298023224
       ],
       "class_id": 15,
-      "p": 0.043209876543209874
+      "p": 0.043750000000000004
     },
     {
       "name": "ART-AbductiveNLI",
@@ -617,7 +617,7 @@
         0.10000000149011612
       ],
       "class_id": 16,
-      "p": 0.037037037037037035
+      "p": 0.037500000000000006
     },
     {
       "name": "EntailmentBankV3",
@@ -632,13 +632,13 @@
         0.10000000149011612
       ],
       "class_id": 17,
-      "p": 0.043209876543209874
+      "p": 0.043750000000000004
     },
     {
       "name": "LogiQA2.0NLI",
       "path": "tasksource/logiqa-2.0-nli",
       "split": "train",
-      "weight": 0.65,
+      "weight": 0.45,
       "dialect": [
         0.05000000074505806,
         0.25,
@@ -647,13 +647,13 @@
         0.10000000149011612
       ],
       "class_id": 18,
-      "p": 0.08024691358024692
+      "p": 0.05625000000000001
     },
     {
       "name": "TruthfulQA-MC",
       "path": "EleutherAI/truthful_qa_mc",
       "split": "validation",
-      "weight": 0.65,
+      "weight": 0.45,
       "dialect": [
         0.05000000074505806,
         0.3499999940395355,
@@ -662,7 +662,7 @@
         0.10000000149011612
       ],
       "class_id": 19,
-      "p": 0.08024691358024692
+      "p": 0.05625000000000001
     },
     {
       "name": "VUA20-Metaphor",
@@ -677,14 +677,14 @@
         0.3499999940395355
       ],
       "class_id": 20,
-      "p": 0.04938271604938272
+      "p": 0.05000000000000001
     }
   ],
   "_alive_entries": [
     {
       "name": "TinyStories",
       "path": "roneneldan/TinyStories",
-      "split": "train[20%:30%]",
+      "split": "train[30%:50%]",
       "weight": 0.1,
       "dialect": [
         0.6000000238418579,
@@ -694,14 +694,14 @@
         0.20000000298023224
       ],
       "class_id": 0,
-      "p": 0.01234567901234568
+      "p": 0.012500000000000002
     },
     {
       "name": "WikipediaEN",
       "path": "wikimedia/wikipedia",
       "config": "20231101.en",
-      "split": "train[2%:5%]",
-      "weight": 0.4,
+      "split": "train[5%:15%]",
+      "weight": 0.5,
       "dialect": [
         0.11999999731779099,
         0.5799999833106995,
@@ -710,12 +710,12 @@
         0.10000000149011612
       ],
       "class_id": 1,
-      "p": 0.04938271604938272
+      "p": 0.06250000000000001
     },
     {
       "name": "AGNews",
       "path": "ag_news",
-      "split": "train[50%:]",
+      "split": "train[:]",
       "weight": 0.1,
       "dialect": [
         0.20000000298023224,
@@ -725,13 +725,13 @@
         0.10000000149011612
       ],
       "class_id": 2,
-      "p": 0.01234567901234568
+      "p": 0.012500000000000002
     },
     {
       "name": "GSM8K",
       "path": "openai/gsm8k",
       "config": "main",
-      "split": "train[20%:40%]",
+      "split": "train[40%:60%]",
       "weight": 0.6,
       "dialect": [
         0.10000000149011612,
@@ -741,14 +741,14 @@
         0.10000000149011612
       ],
       "class_id": 3,
-      "p": 0.07407407407407407
+      "p": 0.07500000000000001
     },
     {
       "name": "AI2-ARC-Easy",
       "path": "allenai/ai2_arc",
       "config": "ARC-Easy",
-      "split": "train[20%:30%]",
-      "weight": 0.5,
+      "split": "train[30%:60%]",
+      "weight": 0.6,
       "dialect": [
         0.05000000074505806,
         0.15000000596046448,
@@ -757,13 +757,13 @@
         0.15000000596046448
       ],
       "class_id": 4,
-      "p": 0.0617283950617284
+      "p": 0.07500000000000001
     },
     {
       "name": "HH-RLHF",
       "path": "Anthropic/hh-rlhf",
-      "split": "train[2%:5%]",
-      "weight": 0.4,
+      "split": "train[5%:10%]",
+      "weight": 0.5,
       "dialect": [
         0.10000000149011612,
         0.25,
@@ -772,7 +772,7 @@
         0.20000000298023224
       ],
       "class_id": 5,
-      "p": 0.04938271604938272
+      "p": 0.06250000000000001
     },
     {
       "name": "SVAMP",
@@ -787,7 +787,7 @@
         0.05000000074505806
       ],
       "class_id": 6,
-      "p": 0.0308641975308642
+      "p": 0.03125000000000001
     },
     {
       "name": "MATH-500",
@@ -802,7 +802,7 @@
         0.05000000074505806
       ],
       "class_id": 7,
-      "p": 0.0308641975308642
+      "p": 0.03125000000000001
     },
     {
       "name": "SEP",
@@ -817,7 +817,7 @@
         0.10000000149011612
       ],
       "class_id": 8,
-      "p": 0.037037037037037035
+      "p": 0.037500000000000006
     },
     {
       "name": "ETHICS-commonsense",
@@ -833,7 +833,7 @@
         0.18000000715255737
       ],
       "class_id": 9,
-      "p": 0.05555555555555556
+      "p": 0.05625000000000001
     },
     {
       "name": "ETHICS-deontology",
@@ -849,7 +849,7 @@
         0.18000000715255737
       ],
       "class_id": 10,
-      "p": 0.043209876543209874
+      "p": 0.043750000000000004
     },
     {
       "name": "ETHICS-justice",
@@ -865,7 +865,7 @@
         0.18000000715255737
       ],
       "class_id": 11,
-      "p": 0.043209876543209874
+      "p": 0.043750000000000004
     },
     {
       "name": "ETHICS-utilitarianism",
@@ -881,7 +881,7 @@
         0.18000000715255737
       ],
       "class_id": 12,
-      "p": 0.043209876543209874
+      "p": 0.043750000000000004
     },
     {
       "name": "ETHICS-virtue",
@@ -897,7 +897,7 @@
         0.18000000715255737
       ],
       "class_id": 13,
-      "p": 0.043209876543209874
+      "p": 0.043750000000000004
     },
     {
       "name": "SocialChem101",
@@ -912,7 +912,7 @@
         0.20000000298023224
       ],
       "class_id": 14,
-      "p": 0.08024691358024692
+      "p": 0.08125000000000002
     },
     {
       "name": "MoralStories",
@@ -927,7 +927,7 @@
         0.20000000298023224
       ],
       "class_id": 15,
-      "p": 0.043209876543209874
+      "p": 0.043750000000000004
     },
     {
       "name": "ART-AbductiveNLI",
@@ -942,13 +942,13 @@
         0.10000000149011612
       ],
       "class_id": 16,
-      "p": 0.037037037037037035
+      "p": 0.037500000000000006
     },
     {
       "name": "LogiQA2.0NLI",
       "path": "tasksource/logiqa-2.0-nli",
       "split": "train",
-      "weight": 0.65,
+      "weight": 0.45,
       "dialect": [
         0.05000000074505806,
         0.25,
@@ -957,13 +957,13 @@
         0.10000000149011612
       ],
       "class_id": 18,
-      "p": 0.08024691358024692
+      "p": 0.05625000000000001
     },
     {
       "name": "TruthfulQA-MC",
       "path": "EleutherAI/truthful_qa_mc",
       "split": "validation",
-      "weight": 0.65,
+      "weight": 0.45,
       "dialect": [
         0.05000000074505806,
         0.3499999940395355,
@@ -972,7 +972,7 @@
         0.10000000149011612
       ],
       "class_id": 19,
-      "p": 0.08024691358024692
+      "p": 0.05625000000000001
     },
     {
       "name": "VUA20-Metaphor",
@@ -987,7 +987,7 @@
         0.3499999940395355
       ],
       "class_id": 20,
-      "p": 0.04938271604938272
+      "p": 0.05000000000000001
     }
   ]
 }