Hugging Face's logo

tangledgroup
/
tangled-llama-q-128k-base-v0.1

Text Generation
Transformers
llama
litgpt
litdata
conversational
Inference Endpoints
Model card Files Community
mtasic85 commited on
Commit
f03b061
•
1 Parent(s): c9b9793

eval

Browse files
Files changed (2) hide show
  1. README.md +229 -0
  2. scripts/TRAIN.md +95 -0
README.md CHANGED
@@ -1,3 +1,232 @@
1
  ---
2
  license: apache-2.0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
  license: apache-2.0
3
+ pipeline_tag: text-generation
4
+ library_name: transformers
5
+ language: [
6
+ 'en', 'am', 'ar', 'as', 'az', 'be', 'bg', 'bn', 'br', 'bs', 'ca', 'cs', 'cy', 'da', 'de', 'el',
7
+ 'eo', 'es', 'et', 'eu', 'fa', 'ff', 'fi', 'fr', 'fy', 'ga', 'gd', 'gl', 'gn', 'gu', 'ha', 'he',
8
+ 'hi', 'hr', 'ht', 'hu', 'hy', 'id', 'ig', 'is', 'it', 'ja', 'jv', 'ka', 'kk', 'km', 'kn', 'ko',
9
+ 'ku', 'ky', 'la', 'lg', 'li', 'ln', 'lo', 'lt', 'lv', 'mg', 'mk', 'ml', 'mn', 'mr', 'ms', 'my',
10
+ 'ne', 'nl', 'no', 'ns', 'om', 'or', 'pa', 'pl', 'ps', 'pt', 'qu', 'rm', 'ro', 'ru', 'sa', 'si',
11
+ 'sc', 'sd', 'sk', 'sl', 'so', 'sq', 'sr', 'ss', 'su', 'sv', 'sw', 'ta', 'te', 'th', 'tl', 'tn',
12
+ 'tr', 'ug', 'uk', 'ur', 'uz', 'vi', 'wo', 'xh', 'yi', 'yo', 'zu',
13
+ ]
14
+ datasets: [
15
+ 'yahma/alpaca-cleaned',
16
+ 'saillab/taco-datasets',
17
+ 'xu-song/cc100-samples',
18
+ 'badrex/llm-emoji-dataset',
19
+ 'pszemraj/simple_wikipedia',
20
+ 'AtlasUnified/Atlas-Reasoning',
21
+ 'fblgit/simple-math',
22
+ 'AtlasUnified/atlas-math-sets',
23
+ 'rvv-karma/Math-QA',
24
+ 'microsoft/orca-math-word-problems-200k',
25
+ 'meta-math/MetaMathQA',
26
+ 'TIGER-Lab/MathInstruct',
27
+ 'ChuGyouk/WebInstructSub-only-socratic',
28
+ 'thesven/gsm8k-reasoning',
29
+ 'AlgorithmicResearchGroup/math_reasoning_autoformalization_track',
30
+ 'KingNish/reasoning-base-20k',
31
+ 'fmars/wiki_stem',
32
+ 'ChuGyouk/WebInstructSub-only-sciencestackexchange',
33
+ 'bigcode/the-stack-smol-xs',
34
+ 'cognitivecomputations/dolphin-coder',
35
+ 'HuggingFaceH4/CodeAlpaca_20K',
36
+ 'm-a-p/CodeFeedback-Filtered-Instruction',
37
+ 'NuclearAi/Nuke-X-Glaive-Python-Dataset',
38
+ 'iamtarun/python_code_instructions_18k_alpaca',
39
+ 'kloodia/html_200k',
40
+ 'kloodia/json_200k',
41
+ 'kloodia/javascript_200k',
42
+ 'bleugreen/typescript-chunks',
43
+ 'SkunkworksAI/reasoning-0.01',
44
+ 'Magpie-Align/Magpie-Reasoning-150K',
45
+ ]
46
+ tags:
47
+ - litgpt
48
+ - litdata
49
  ---
50
+
51
+ # tangled-llama-q-32k-base-v0.1
52
+
53
+ ![logo](./misc/logo.png)
54
+
55
+ A pretrained language model based on the Llama model with about **65M** parameters. This model has been trained on **16.7B** (`16,698,858,240`) tokens from more than **3.6M** (`3,597,088`) dataset rows.
56
+
57
+ This model **isn't** designed for immediate use but rather for Continued Pretraining and Finetuning on a downstream task. While it can handle a context length of up to **128K** (`131,072`) tokens, it was pretrained with sequences of **2K** (`2048`) tokens.
58
+
59
+ The objective is to streamline the cognitive or reasoning core, eliminating any redundant knowledge from the model.
60
+
61
+ [loss, val_loss](https://api.wandb.ai/links/mtasic85/jo8uatgs)
62
+
63
+ [val_ppl](https://api.wandb.ai/links/mtasic85/f06bqhu6)
64
+
65
+ [epoch](https://api.wandb.ai/links/mtasic85/buj8l3pf)
66
+
67
+ [learning_rate](https://api.wandb.ai/links/mtasic85/h0ooy6sf)
68
+
69
+ ## lm-evaluation-harness
70
+
71
+ ```bash
72
+ litgpt evaluate --tasks 'hellaswag,gsm8k,truthfulqa_mc2,mmlu,winogrande,arc_challenge' --out_dir 'evaluate-quick/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
73
+ ```
74
+
75
+ | Tasks |Version| Filter |n-shot| Metric | |Value | |Stderr|
76
+ |---------------------------------------|------:|----------------|-----:|-----------|---|-----:|---|-----:|
77
+ |arc_challenge | 1|none | 0|acc |↑ |0.1962|± |0.0116|
78
+ | | |none | 0|acc_norm |↑ |0.2304|± |0.0123|
79
+ |gsm8k | 3|flexible-extract| 5|exact_match|↑ |0.0144|± |0.0033|
80
+ | | |strict-match | 5|exact_match|↑ |0.0015|± |0.0011|
81
+ |hellaswag | 1|none | 0|acc |↑ |0.2631|± |0.0044|
82
+ | | |none | 0|acc_norm |↑ |0.2758|± |0.0045|
83
+ |mmlu | 2|none | |acc |↑ |0.2473|± |0.0036|
84
+ | - humanities | 2|none | |acc |↑ |0.2351|± |0.0062|
85
+ | - formal_logic | 1|none | 0|acc |↑ |0.2857|± |0.0404|
86
+ | - high_school_european_history | 1|none | 0|acc |↑ |0.2667|± |0.0345|
87
+ | - high_school_us_history | 1|none | 0|acc |↑ |0.2696|± |0.0311|
88
+ | - high_school_world_history | 1|none | 0|acc |↑ |0.2110|± |0.0266|
89
+ | - international_law | 1|none | 0|acc |↑ |0.1653|± |0.0339|
90
+ | - jurisprudence | 1|none | 0|acc |↑ |0.2870|± |0.0437|
91
+ | - logical_fallacies | 1|none | 0|acc |↑ |0.2331|± |0.0332|
92
+ | - moral_disputes | 1|none | 0|acc |↑ |0.2283|± |0.0226|
93
+ | - moral_scenarios | 1|none | 0|acc |↑ |0.2425|± |0.0143|
94
+ | - philosophy | 1|none | 0|acc |↑ |0.2186|± |0.0235|
95
+ | - prehistory | 1|none | 0|acc |↑ |0.2099|± |0.0227|
96
+ | - professional_law | 1|none | 0|acc |↑ |0.2314|± |0.0108|
97
+ | - world_religions | 1|none | 0|acc |↑ |0.2632|± |0.0338|
98
+ | - other | 2|none | |acc |↑ |0.2485|± |0.0078|
99
+ | - business_ethics | 1|none | 0|acc |↑ |0.2600|± |0.0441|
100
+ | - clinical_knowledge | 1|none | 0|acc |↑ |0.2528|± |0.0267|
101
+ | - college_medicine | 1|none | 0|acc |↑ |0.2254|± |0.0319|
102
+ | - global_facts | 1|none | 0|acc |↑ |0.2700|± |0.0446|
103
+ | - human_aging | 1|none | 0|acc |↑ |0.2377|± |0.0286|
104
+ | - management | 1|none | 0|acc |↑ |0.2816|± |0.0445|
105
+ | - marketing | 1|none | 0|acc |↑ |0.2692|± |0.0291|
106
+ | - medical_genetics | 1|none | 0|acc |↑ |0.2600|± |0.0441|
107
+ | - miscellaneous | 1|none | 0|acc |↑ |0.2350|± |0.0152|
108
+ | - nutrition | 1|none | 0|acc |↑ |0.2549|± |0.0250|
109
+ | - professional_accounting | 1|none | 0|acc |↑ |0.2801|± |0.0268|
110
+ | - professional_medicine | 1|none | 0|acc |↑ |0.2610|± |0.0267|
111
+ | - virology | 1|none | 0|acc |↑ |0.1807|± |0.0300|
112
+ | - social sciences | 2|none | |acc |↑ |0.2658|± |0.0080|
113
+ | - econometrics | 1|none | 0|acc |↑ |0.1930|± |0.0371|
114
+ | - high_school_geography | 1|none | 0|acc |↑ |0.2172|± |0.0294|
115
+ | - high_school_government_and_politics| 1|none | 0|acc |↑ |0.3212|± |0.0337|
116
+ | - high_school_macroeconomics | 1|none | 0|acc |↑ |0.2923|± |0.0231|
117
+ | - high_school_microeconomics | 1|none | 0|acc |↑ |0.3025|± |0.0298|
118
+ | - high_school_psychology | 1|none | 0|acc |↑ |0.2752|± |0.0191|
119
+ | - human_sexuality | 1|none | 0|acc |↑ |0.2290|± |0.0369|
120
+ | - professional_psychology | 1|none | 0|acc |↑ |0.2386|± |0.0172|
121
+ | - public_relations | 1|none | 0|acc |↑ |0.2636|± |0.0422|
122
+ | - security_studies | 1|none | 0|acc |↑ |0.3143|± |0.0297|
123
+ | - sociology | 1|none | 0|acc |↑ |0.2338|± |0.0299|
124
+ | - us_foreign_policy | 1|none | 0|acc |↑ |0.2600|± |0.0441|
125
+ | - stem | 2|none | |acc |↑ |0.2464|± |0.0077|
126
+ | - abstract_algebra | 1|none | 0|acc |↑ |0.2500|± |0.0435|
127
+ | - anatomy | 1|none | 0|acc |↑ |0.2148|± |0.0355|
128
+ | - astronomy | 1|none | 0|acc |↑ |0.1908|± |0.0320|
129
+ | - college_biology | 1|none | 0|acc |↑ |0.2569|± |0.0365|
130
+ | - college_chemistry | 1|none | 0|acc |↑ |0.2700|± |0.0446|
131
+ | - college_computer_science | 1|none | 0|acc |↑ |0.3500|± |0.0479|
132
+ | - college_mathematics | 1|none | 0|acc |↑ |0.2700|± |0.0446|
133
+ | - college_physics | 1|none | 0|acc |↑ |0.2745|± |0.0444|
134
+ | - computer_security | 1|none | 0|acc |↑ |0.3000|± |0.0461|
135
+ | - conceptual_physics | 1|none | 0|acc |↑ |0.2766|± |0.0292|
136
+ | - electrical_engineering | 1|none | 0|acc |↑ |0.2345|± |0.0353|
137
+ | - elementary_mathematics | 1|none | 0|acc |↑ |0.2566|± |0.0225|
138
+ | - high_school_biology | 1|none | 0|acc |↑ |0.2226|± |0.0237|
139
+ | - high_school_chemistry | 1|none | 0|acc |↑ |0.2217|± |0.0292|
140
+ | - high_school_computer_science | 1|none | 0|acc |↑ |0.2000|± |0.0402|
141
+ | - high_school_mathematics | 1|none | 0|acc |↑ |0.2370|± |0.0259|
142
+ | - high_school_physics | 1|none | 0|acc |↑ |0.2517|± |0.0354|
143
+ | - high_school_statistics | 1|none | 0|acc |↑ |0.2685|± |0.0302|
144
+ | - machine_learning | 1|none | 0|acc |↑ |0.1786|± |0.0364|
145
+ |truthfulqa_mc2 | 2|none | 0|acc |↑ |0.4668|± |0.0161|
146
+ |winogrande | 1|none | 0|acc |↑ |0.5012|± |0.0141|
147
+
148
+ | Groups |Version|Filter|n-shot|Metric| |Value | |Stderr|
149
+ |------------------|------:|------|------|------|---|-----:|---|-----:|
150
+ |mmlu | 2|none | |acc |↑ |0.2473|± |0.0036|
151
+ | - humanities | 2|none | |acc |↑ |0.2351|± |0.0062|
152
+ | - other | 2|none | |acc |↑ |0.2485|± |0.0078|
153
+ | - social sciences| 2|none | |acc |↑ |0.2658|± |0.0080|
154
+ | - stem | 2|none | |acc |↑ |0.2464|± |0.0077|
155
+
156
+ ```bash
157
+ litgpt evaluate --tasks 'leaderboard' --out_dir 'evaluate-leaderboard/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
158
+ ```
159
+
160
+ Tasks |Version|Filter|n-shot| Metric | |Value | |Stderr|
161
+ |-----------------------------------------------------------|-------|------|-----:|-----------------------|---|-----:|---|------|
162
+ |leaderboard | N/A| | | | | | | |
163
+ | - leaderboard_bbh | N/A| | | | | | | |
164
+ | - leaderboard_bbh_boolean_expressions | 1|none | 3|acc_norm |↑ |0.4600|± |0.0316|
165
+ | - leaderboard_bbh_causal_judgement | 1|none | 3|acc_norm |↑ |0.5187|± |0.0366|
166
+ | - leaderboard_bbh_date_understanding | 1|none | 3|acc_norm |↑ |0.1840|± |0.0246|
167
+ | - leaderboard_bbh_disambiguation_qa | 1|none | 3|acc_norm |↑ |0.3880|± |0.0309|
168
+ | - leaderboard_bbh_formal_fallacies | 1|none | 3|acc_norm |↑ |0.4680|± |0.0316|
169
+ | - leaderboard_bbh_geometric_shapes | 1|none | 3|acc_norm |↑ |0.1000|± |0.0190|
170
+ | - leaderboard_bbh_hyperbaton | 1|none | 3|acc_norm |↑ |0.5160|± |0.0317|
171
+ | - leaderboard_bbh_logical_deduction_five_objects | 1|none | 3|acc_norm |↑ |0.2080|± |0.0257|
172
+ | - leaderboard_bbh_logical_deduction_seven_objects | 1|none | 3|acc_norm |↑ |0.1720|± |0.0239|
173
+ | - leaderboard_bbh_logical_deduction_three_objects | 1|none | 3|acc_norm |↑ |0.3280|± |0.0298|
174
+ | - leaderboard_bbh_movie_recommendation | 1|none | 3|acc_norm |↑ |0.2640|± |0.0279|
175
+ | - leaderboard_bbh_navigate | 1|none | 3|acc_norm |↑ |0.5760|± |0.0313|
176
+ | - leaderboard_bbh_object_counting | 1|none | 3|acc_norm |↑ |0.0520|± |0.0141|
177
+ | - leaderboard_bbh_penguins_in_a_table | 1|none | 3|acc_norm |↑ |0.2260|± |0.0347|
178
+ | - leaderboard_bbh_reasoning_about_colored_objects | 1|none | 3|acc_norm |↑ |0.0720|± |0.0164|
179
+ | - leaderboard_bbh_ruin_names | 1|none | 3|acc_norm |↑ |0.2280|± |0.0266|
180
+ | - leaderboard_bbh_salient_translation_error_detection | 1|none | 3|acc_norm |↑ |0.1920|± |0.0250|
181
+ | - leaderboard_bbh_snarks | 1|none | 3|acc_norm |↑ |0.4831|± |0.0376|
182
+ | - leaderboard_bbh_sports_understanding | 1|none | 3|acc_norm |↑ |0.4600|± |0.0316|
183
+ | - leaderboard_bbh_temporal_sequences | 1|none | 3|acc_norm |↑ |0.2360|± |0.0269|
184
+ | - leaderboard_bbh_tracking_shuffled_objects_five_objects | 1|none | 3|acc_norm |↑ |0.2080|± |0.0257|
185
+ | - leaderboard_bbh_tracking_shuffled_objects_seven_objects| 1|none | 3|acc_norm |↑ |0.1680|± |0.0237|
186
+ | - leaderboard_bbh_tracking_shuffled_objects_three_objects| 1|none | 3|acc_norm |↑ |0.3040|± |0.0292|
187
+ | - leaderboard_bbh_web_of_lies | 1|none | 3|acc_norm |↑ |0.4880|± |0.0317|
188
+ | - leaderboard_gpqa | N/A| | | | | | | |
189
+ | - leaderboard_gpqa_diamond | 1|none | 0|acc_norm |↑ |0.2121|± |0.0291|
190
+ | - leaderboard_gpqa_extended | 1|none | 0|acc_norm |↑ |0.2619|± |0.0188|
191
+ | - leaderboard_gpqa_main | 1|none | 0|acc_norm |↑ |0.2589|± |0.0207|
192
+ | - leaderboard_ifeval | 3|none | 0|inst_level_loose_acc |↑ |0.1966|± | N/A|
193
+ | | |none | 0|inst_level_strict_acc |↑ |0.1835|± | N/A|
194
+ | | |none | 0|prompt_level_loose_acc |↑ |0.1017|± |0.0130|
195
+ | | |none | 0|prompt_level_strict_acc|↑ |0.0998|± |0.0129|
196
+ | - leaderboard_math_hard | N/A| | | | | | | |
197
+ | - leaderboard_math_algebra_hard | 1|none | 4|exact_match |↑ |0.0000|± | 0|
198
+ | - leaderboard_math_counting_and_prob_hard | 1|none | 4|exact_match |↑ |0.0000|± | 0|
199
+ | - leaderboard_math_geometry_hard | 1|none | 4|exact_match |↑ |0.0000|± | 0|
200
+ | - leaderboard_math_intermediate_algebra_hard | 1|none | 4|exact_match |↑ |0.0000|± | 0|
201
+ | - leaderboard_math_num_theory_hard | 1|none | 4|exact_match |↑ |0.0000|± | 0|
202
+ | - leaderboard_math_prealgebra_hard | 1|none | 4|exact_match |↑ |0.0000|± | 0|
203
+ | - leaderboard_math_precalculus_hard | 1|none | 4|exact_match |↑ |0.0000|± | 0|
204
+ | - leaderboard_mmlu_pro | 0.1|none | 5|acc |↑ |0.1155|± |0.0029|
205
+ | - leaderboard_musr | N/A| | | | | | | |
206
+ | - leaderboard_musr_murder_mysteries | 1|none | 0|acc_norm |↑ |0.5040|± |0.0317|
207
+ | - leaderboard_musr_object_placements | 1|none | 0|acc_norm |↑ |0.3086|± |0.0289|
208
+ | - leaderboard_musr_team_allocation | 1|none | 0|acc_norm |↑ |0.3400|± |0.0300|
209
+
210
+ ```bash
211
+ litgpt evaluate --tasks 'bbh_zeroshot,bbh_fewshot,bbh_cot_fewshot,bbh_cot_zeroshot' --out_dir 'evaluate-bigbenchhard/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
212
+ ```
213
+
214
+ ```bash
215
+ litgpt evaluate --tasks 'mmlu,mmlu_pro' --out_dir 'evaluate-mmlu/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
216
+ ```
217
+
218
+ ```bash
219
+ litgpt evaluate --tasks 'arc_challenge,boolq,gpqa,hellaswag,openbookqa,piqa,truthfulqa_mc2,winogrande' --out_dir 'evaluate-reasoning/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
220
+ ```
221
+
222
+ ```bash
223
+ litgpt evaluate --tasks 'mmlu_multilingual,mgsm' --out_dir 'evaluate-multilinguals/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
224
+ ```
225
+
226
+ ```bash
227
+ litgpt evaluate --tasks 'gsm8k,mathqa' --out_dir 'evaluate-math/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
228
+ ```
229
+
230
+ ```bash
231
+ litgpt evaluate --tasks 'wikitext,qasper' --out_dir 'evaluate-long/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
232
+ ```
scripts/TRAIN.md CHANGED
@@ -58,18 +58,113 @@ save_file(state_dict, 'out/converted_model/model.safetensors')
58
 
59
  ```bash
60
  litgpt evaluate --tasks 'hellaswag,gsm8k,truthfulqa_mc2,mmlu,winogrande,arc_challenge' --out_dir 'evaluate-quick/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
61
 
 
62
  litgpt evaluate --tasks 'leaderboard' --out_dir 'evaluate-leaderboard/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
 
63
 
 
64
  litgpt evaluate --tasks 'bbh_zeroshot,bbh_fewshot,bbh_cot_fewshot,bbh_cot_zeroshot' --out_dir 'evaluate-bigbenchhard/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
 
65
 
 
66
  litgpt evaluate --tasks 'mmlu,mmlu_pro' --out_dir 'evaluate-mmlu/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
 
67
 
 
68
  litgpt evaluate --tasks 'arc_challenge,boolq,gpqa,hellaswag,openbookqa,piqa,truthfulqa_mc2,winogrande' --out_dir 'evaluate-reasoning/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
 
69
 
 
70
  litgpt evaluate --tasks 'mmlu_multilingual,mgsm' --out_dir 'evaluate-multilinguals/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
 
71
 
 
72
  litgpt evaluate --tasks 'gsm8k,mathqa' --out_dir 'evaluate-math/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
 
73
 
 
74
  litgpt evaluate --tasks 'qasper' --out_dir 'evaluate-long/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
75
  ```
 
58
 
59
  ```bash
60
  litgpt evaluate --tasks 'hellaswag,gsm8k,truthfulqa_mc2,mmlu,winogrande,arc_challenge' --out_dir 'evaluate-quick/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
61
+ ```
62
+
63
+ | Tasks |Version| Filter |n-shot| Metric | |Value | |Stderr|
64
+ |---------------------------------------|------:|----------------|-----:|-----------|---|-----:|---|-----:|
65
+ |arc_challenge | 1|none | 0|acc |↑ |0.1962|± |0.0116|
66
+ | | |none | 0|acc_norm |↑ |0.2304|± |0.0123|
67
+ |gsm8k | 3|flexible-extract| 5|exact_match|↑ |0.0144|± |0.0033|
68
+ | | |strict-match | 5|exact_match|↑ |0.0015|± |0.0011|
69
+ |hellaswag | 1|none | 0|acc |↑ |0.2631|± |0.0044|
70
+ | | |none | 0|acc_norm |↑ |0.2758|± |0.0045|
71
+ |mmlu | 2|none | |acc |↑ |0.2473|± |0.0036|
72
+ | - humanities | 2|none | |acc |↑ |0.2351|± |0.0062|
73
+ | - formal_logic | 1|none | 0|acc |↑ |0.2857|± |0.0404|
74
+ | - high_school_european_history | 1|none | 0|acc |↑ |0.2667|± |0.0345|
75
+ | - high_school_us_history | 1|none | 0|acc |↑ |0.2696|± |0.0311|
76
+ | - high_school_world_history | 1|none | 0|acc |↑ |0.2110|± |0.0266|
77
+ | - international_law | 1|none | 0|acc |↑ |0.1653|± |0.0339|
78
+ | - jurisprudence | 1|none | 0|acc |↑ |0.2870|± |0.0437|
79
+ | - logical_fallacies | 1|none | 0|acc |↑ |0.2331|± |0.0332|
80
+ | - moral_disputes | 1|none | 0|acc |↑ |0.2283|± |0.0226|
81
+ | - moral_scenarios | 1|none | 0|acc |↑ |0.2425|± |0.0143|
82
+ | - philosophy | 1|none | 0|acc |↑ |0.2186|± |0.0235|
83
+ | - prehistory | 1|none | 0|acc |↑ |0.2099|± |0.0227|
84
+ | - professional_law | 1|none | 0|acc |↑ |0.2314|± |0.0108|
85
+ | - world_religions | 1|none | 0|acc |↑ |0.2632|± |0.0338|
86
+ | - other | 2|none | |acc |↑ |0.2485|± |0.0078|
87
+ | - business_ethics | 1|none | 0|acc |↑ |0.2600|± |0.0441|
88
+ | - clinical_knowledge | 1|none | 0|acc |↑ |0.2528|± |0.0267|
89
+ | - college_medicine | 1|none | 0|acc |↑ |0.2254|± |0.0319|
90
+ | - global_facts | 1|none | 0|acc |↑ |0.2700|± |0.0446|
91
+ | - human_aging | 1|none | 0|acc |↑ |0.2377|± |0.0286|
92
+ | - management | 1|none | 0|acc |↑ |0.2816|± |0.0445|
93
+ | - marketing | 1|none | 0|acc |↑ |0.2692|± |0.0291|
94
+ | - medical_genetics | 1|none | 0|acc |↑ |0.2600|± |0.0441|
95
+ | - miscellaneous | 1|none | 0|acc |↑ |0.2350|± |0.0152|
96
+ | - nutrition | 1|none | 0|acc |↑ |0.2549|± |0.0250|
97
+ | - professional_accounting | 1|none | 0|acc |↑ |0.2801|± |0.0268|
98
+ | - professional_medicine | 1|none | 0|acc |↑ |0.2610|± |0.0267|
99
+ | - virology | 1|none | 0|acc |↑ |0.1807|± |0.0300|
100
+ | - social sciences | 2|none | |acc |↑ |0.2658|± |0.0080|
101
+ | - econometrics | 1|none | 0|acc |↑ |0.1930|± |0.0371|
102
+ | - high_school_geography | 1|none | 0|acc |↑ |0.2172|± |0.0294|
103
+ | - high_school_government_and_politics| 1|none | 0|acc |↑ |0.3212|± |0.0337|
104
+ | - high_school_macroeconomics | 1|none | 0|acc |↑ |0.2923|± |0.0231|
105
+ | - high_school_microeconomics | 1|none | 0|acc |↑ |0.3025|± |0.0298|
106
+ | - high_school_psychology | 1|none | 0|acc |↑ |0.2752|± |0.0191|
107
+ | - human_sexuality | 1|none | 0|acc |↑ |0.2290|± |0.0369|
108
+ | - professional_psychology | 1|none | 0|acc |↑ |0.2386|± |0.0172|
109
+ | - public_relations | 1|none | 0|acc |↑ |0.2636|± |0.0422|
110
+ | - security_studies | 1|none | 0|acc |↑ |0.3143|± |0.0297|
111
+ | - sociology | 1|none | 0|acc |↑ |0.2338|± |0.0299|
112
+ | - us_foreign_policy | 1|none | 0|acc |↑ |0.2600|± |0.0441|
113
+ | - stem | 2|none | |acc |↑ |0.2464|± |0.0077|
114
+ | - abstract_algebra | 1|none | 0|acc |↑ |0.2500|± |0.0435|
115
+ | - anatomy | 1|none | 0|acc |↑ |0.2148|± |0.0355|
116
+ | - astronomy | 1|none | 0|acc |↑ |0.1908|± |0.0320|
117
+ | - college_biology | 1|none | 0|acc |↑ |0.2569|± |0.0365|
118
+ | - college_chemistry | 1|none | 0|acc |↑ |0.2700|± |0.0446|
119
+ | - college_computer_science | 1|none | 0|acc |↑ |0.3500|± |0.0479|
120
+ | - college_mathematics | 1|none | 0|acc |↑ |0.2700|± |0.0446|
121
+ | - college_physics | 1|none | 0|acc |↑ |0.2745|± |0.0444|
122
+ | - computer_security | 1|none | 0|acc |↑ |0.3000|± |0.0461|
123
+ | - conceptual_physics | 1|none | 0|acc |↑ |0.2766|± |0.0292|
124
+ | - electrical_engineering | 1|none | 0|acc |↑ |0.2345|± |0.0353|
125
+ | - elementary_mathematics | 1|none | 0|acc |↑ |0.2566|± |0.0225|
126
+ | - high_school_biology | 1|none | 0|acc |↑ |0.2226|± |0.0237|
127
+ | - high_school_chemistry | 1|none | 0|acc |↑ |0.2217|± |0.0292|
128
+ | - high_school_computer_science | 1|none | 0|acc |↑ |0.2000|± |0.0402|
129
+ | - high_school_mathematics | 1|none | 0|acc |↑ |0.2370|± |0.0259|
130
+ | - high_school_physics | 1|none | 0|acc |↑ |0.2517|± |0.0354|
131
+ | - high_school_statistics | 1|none | 0|acc |↑ |0.2685|± |0.0302|
132
+ | - machine_learning | 1|none | 0|acc |↑ |0.1786|± |0.0364|
133
+ |truthfulqa_mc2 | 2|none | 0|acc |↑ |0.4668|± |0.0161|
134
+ |winogrande | 1|none | 0|acc |↑ |0.5012|± |0.0141|
135
+
136
+ | Groups |Version|Filter|n-shot|Metric| |Value | |Stderr|
137
+ |------------------|------:|------|------|------|---|-----:|---|-----:|
138
+ |mmlu | 2|none | |acc |↑ |0.2473|± |0.0036|
139
+ | - humanities | 2|none | |acc |↑ |0.2351|± |0.0062|
140
+ | - other | 2|none | |acc |↑ |0.2485|± |0.0078|
141
+ | - social sciences| 2|none | |acc |↑ |0.2658|± |0.0080|
142
+ | - stem | 2|none | |acc |↑ |0.2464|± |0.0077|
143
 
144
+ ```bash
145
  litgpt evaluate --tasks 'leaderboard' --out_dir 'evaluate-leaderboard/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
146
+ ```
147
 
148
+ ```bash
149
  litgpt evaluate --tasks 'bbh_zeroshot,bbh_fewshot,bbh_cot_fewshot,bbh_cot_zeroshot' --out_dir 'evaluate-bigbenchhard/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
150
+ ```
151
 
152
+ ```bash
153
  litgpt evaluate --tasks 'mmlu,mmlu_pro' --out_dir 'evaluate-mmlu/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
154
+ ```
155
 
156
+ ```bash
157
  litgpt evaluate --tasks 'arc_challenge,boolq,gpqa,hellaswag,openbookqa,piqa,truthfulqa_mc2,winogrande' --out_dir 'evaluate-reasoning/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
158
+ ```
159
 
160
+ ```bash
161
  litgpt evaluate --tasks 'mmlu_multilingual,mgsm' --out_dir 'evaluate-multilinguals/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
162
+ ```
163
 
164
+ ```bash
165
  litgpt evaluate --tasks 'gsm8k,mathqa' --out_dir 'evaluate-math/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
166
+ ```
167
 
168
+ ```bash
169
  litgpt evaluate --tasks 'qasper' --out_dir 'evaluate-long/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
170
  ```