Hugging Face's logo

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

Text Generation
Transformers
Safetensors
llama
litgpt
litdata
conversational
text-generation-inference
Inference Endpoints
Model card Files Community
mtasic85 commited on
Commit
9e90c28
•
1 Parent(s): 9a24edf

benchmark

Browse files
Files changed (3) hide show
  1. README.md +380 -0
  2. misc/logo.png +3 -0
  3. scripts/TRAIN.md +2 -2
README.md CHANGED
@@ -1,3 +1,383 @@
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
+ 'am', 'ar', 'as', 'az', 'be', 'bg', 'bn', 'bn_rom', 'br',
7
+ 'bs', 'ca', 'cs', 'cy', 'da', 'de', 'el', 'en', 'eo', 'es',
8
+ 'et', 'eu', 'fa', 'ff', 'fi', 'fr', 'fy', 'ga', 'gd', 'gl',
9
+ 'gn', 'gu', 'ha', 'he', 'hi', 'hi_rom', 'hr', 'ht', 'hu',
10
+ 'hy', 'id', 'ig', 'is', 'it', 'ja', 'jv', 'ka', 'kk', 'km',
11
+ 'kn', 'ko', 'ku', 'ky', 'la', 'lg', 'li', 'ln', 'lo', 'lt',
12
+ 'lv', 'mg', 'mk', 'ml', 'mn', 'mr', 'ms', 'my', 'my_zaw',
13
+ 'ne', 'nl', 'no', 'ns', 'om', 'or', 'pa', 'pl', 'ps', 'pt',
14
+ 'qu', 'rm', 'ro', 'ru', 'sa', 'si', 'sc', 'sd', 'sk', 'sl',
15
+ 'so', 'sq', 'sr', 'ss', 'su', 'sv', 'sw', 'ta', 'ta_rom',
16
+ 'te', 'te_rom', 'th', 'tl', 'tn', 'tr', 'ug', 'uk', 'ur',
17
+ 'ur_rom', 'uz', 'vi', 'wo', 'xh', 'yi', 'yo',
18
+ 'zh-Hans', 'zh-Hant', 'zu',
19
+ ]
20
+ datasets: [
21
+ 'yahma/alpaca-cleaned',
22
+ 'gbharti/wealth-alpaca_lora',
23
+ 'saillab/taco-datasets',
24
+ 'xu-song/cc100-samples',
25
+ 'ontocord/fineweb-permissive-multilingual-2m',
26
+ 'MuskumPillerum/General-Knowledge',
27
+ 'yirenc/general_knowledge_boolean',
28
+ 'nampdn-ai/tiny-textbooks',
29
+ 'nampdn-ai/tiny-codes',
30
+ 'bigcode/the-stack-smol-xs',
31
+ 'm-a-p/CodeFeedback-Filtered-Instruction',
32
+ 'jtatman/python-code-dataset-500k',
33
+ 'iamtarun/python_code_instructions_18k_alpaca',
34
+ 'HuggingFaceH4/CodeAlpaca_20K',
35
+ 'gair-prox/open-web-math-pro',
36
+ 'rvv-karma/Math-QA',
37
+ 'ajibawa-2023/Maths-College',
38
+ 'microsoft/orca-math-word-problems-200k',
39
+ 'fblgit/simple-math',
40
+ 'SkunkworksAI/reasoning-0.01',
41
+ 'badrex/llm-emoji-dataset',
42
+ ]
43
+ tags:
44
+ - litgpt
45
+ - litdata
46
  ---
47
+
48
+ # tangled-llama-154m-32k-base-v0.1
49
+
50
+ ![logo](./misc/logo.png)
51
+
52
+ A pretrained language model based on the Llama model with about **154M** parameters. This model has been trained on **11.4B** (`11,422,750,857`) tokens from more than **0.8M** (`796,399`) dataset rows.
53
+
54
+ 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.
55
+
56
+ The objective is to streamline the cognitive or reasoning core, eliminating any redundant knowledge from the model.
57
+
58
+ [loss, val_loss](https://api.wandb.ai/links/mtasic85/88j5c76s)
59
+
60
+ [val_ppl](https://api.wandb.ai/links/mtasic85/slfz004v)
61
+
62
+ [epoch](https://api.wandb.ai/links/mtasic85/c2za3xna)
63
+
64
+ [learning_rate](https://api.wandb.ai/links/mtasic85/670vk6fg)
65
+
66
+ ## lm-evaluation-harness
67
+
68
+ ```bash
69
+ litgpt evaluate --tasks 'hellaswag,gsm8k,truthfulqa_mc2,mmlu,winogrande,arc_challenge' --out_dir 'evaluate-quick/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
70
+ ```
71
+
72
+ | Tasks |Version| Filter |n-shot| Metric | |Value | |Stderr|
73
+ |---------------------------------------|------:|----------------|-----:|-----------|---|-----:|---|-----:|
74
+ |arc_challenge | 1|none | 0|acc |↑ |0.1877|± |0.0114|
75
+ | | |none | 0|acc_norm |↑ |0.2389|± |0.0125|
76
+ |gsm8k | 3|flexible-extract| 5|exact_match|↑ |0.0136|± |0.0032|
77
+ | | |strict-match | 5|exact_match|↑ |0.0008|± |0.0008|
78
+ |hellaswag | 1|none | 0|acc |↑ |0.2660|± |0.0044|
79
+ | | |none | 0|acc_norm |↑ |0.2697|± |0.0044|
80
+ |mmlu | 2|none | |acc |↑ |0.2377|± |0.0036|
81
+ | - humanities | 2|none | |acc |↑ |0.2372|± |0.0062|
82
+ | - formal_logic | 1|none | 0|acc |↑ |0.2619|± |0.0393|
83
+ | - high_school_european_history | 1|none | 0|acc |↑ |0.2182|± |0.0323|
84
+ | - high_school_us_history | 1|none | 0|acc |↑ |0.2500|± |0.0304|
85
+ | - high_school_world_history | 1|none | 0|acc |↑ |0.2447|± |0.0280|
86
+ | - international_law | 1|none | 0|acc |↑ |0.2066|± |0.0370|
87
+ | - jurisprudence | 1|none | 0|acc |↑ |0.2778|± |0.0433|
88
+ | - logical_fallacies | 1|none | 0|acc |↑ |0.2025|± |0.0316|
89
+ | - moral_disputes | 1|none | 0|acc |↑ |0.2514|± |0.0234|
90
+ | - moral_scenarios | 1|none | 0|acc |↑ |0.2425|± |0.0143|
91
+ | - philosophy | 1|none | 0|acc |↑ |0.1768|± |0.0217|
92
+ | - prehistory | 1|none | 0|acc |↑ |0.2562|± |0.0243|
93
+ | - professional_law | 1|none | 0|acc |↑ |0.2379|± |0.0109|
94
+ | - world_religions | 1|none | 0|acc |↑ |0.2515|± |0.0333|
95
+ | - other | 2|none | |acc |↑ |0.2462|± |0.0077|
96
+ | - business_ethics | 1|none | 0|acc |↑ |0.2800|± |0.0451|
97
+ | - clinical_knowledge | 1|none | 0|acc |↑ |0.2377|± |0.0262|
98
+ | - college_medicine | 1|none | 0|acc |↑ |0.2370|± |0.0324|
99
+ | - global_facts | 1|none | 0|acc |↑ |0.2500|± |0.0435|
100
+ | - human_aging | 1|none | 0|acc |↑ |0.2108|± |0.0274|
101
+ | - management | 1|none | 0|acc |↑ |0.1942|± |0.0392|
102
+ | - marketing | 1|none | 0|acc |↑ |0.2436|± |0.0281|
103
+ | - medical_genetics | 1|none | 0|acc |↑ |0.2700|± |0.0446|
104
+ | - miscellaneous | 1|none | 0|acc |↑ |0.2554|± |0.0156|
105
+ | - nutrition | 1|none | 0|acc |↑ |0.2778|± |0.0256|
106
+ | - professional_accounting | 1|none | 0|acc |↑ |0.2411|± |0.0255|
107
+ | - professional_medicine | 1|none | 0|acc |↑ |0.2279|± |0.0255|
108
+ | - virology | 1|none | 0|acc |↑ |0.2530|± |0.0338|
109
+ | - social sciences | 2|none | |acc |↑ |0.2525|± |0.0078|
110
+ | - econometrics | 1|none | 0|acc |↑ |0.2281|± |0.0395|
111
+ | - high_school_geography | 1|none | 0|acc |↑ |0.1465|± |0.0252|
112
+ | - high_school_government_and_politics| 1|none | 0|acc |↑ |0.2539|± |0.0314|
113
+ | - high_school_macroeconomics | 1|none | 0|acc |↑ |0.2333|± |0.0214|
114
+ | - high_school_microeconomics | 1|none | 0|acc |↑ |0.2269|± |0.0272|
115
+ | - high_school_psychology | 1|none | 0|acc |↑ |0.2330|± |0.0181|
116
+ | - human_sexuality | 1|none | 0|acc |↑ |0.2824|± |0.0395|
117
+ | - professional_psychology | 1|none | 0|acc |↑ |0.2859|± |0.0183|
118
+ | - public_relations | 1|none | 0|acc |↑ |0.2364|± |0.0407|
119
+ | - security_studies | 1|none | 0|acc |↑ |0.3388|± |0.0303|
120
+ | - sociology | 1|none | 0|acc |↑ |0.2637|± |0.0312|
121
+ | - us_foreign_policy | 1|none | 0|acc |↑ |0.2700|± |0.0446|
122
+ | - stem | 2|none | |acc |↑ |0.2157|± |0.0073|
123
+ | - abstract_algebra | 1|none | 0|acc |↑ |0.2000|± |0.0402|
124
+ | - anatomy | 1|none | 0|acc |↑ |0.2222|± |0.0359|
125
+ | - astronomy | 1|none | 0|acc |↑ |0.1842|± |0.0315|
126
+ | - college_biology | 1|none | 0|acc |↑ |0.2639|± |0.0369|
127
+ | - college_chemistry | 1|none | 0|acc |↑ |0.2100|± |0.0409|
128
+ | - college_computer_science | 1|none | 0|acc |↑ |0.2400|± |0.0429|
129
+ | - college_mathematics | 1|none | 0|acc |↑ |0.2200|± |0.0416|
130
+ | - college_physics | 1|none | 0|acc |↑ |0.2059|± |0.0402|
131
+ | - computer_security | 1|none | 0|acc |↑ |0.2400|± |0.0429|
132
+ | - conceptual_physics | 1|none | 0|acc |↑ |0.2553|± |0.0285|
133
+ | - electrical_engineering | 1|none | 0|acc |↑ |0.2414|± |0.0357|
134
+ | - elementary_mathematics | 1|none | 0|acc |↑ |0.2328|± |0.0218|
135
+ | - high_school_biology | 1|none | 0|acc |↑ |0.1839|± |0.0220|
136
+ | - high_school_chemistry | 1|none | 0|acc |↑ |0.1626|�� |0.0260|
137
+ | - high_school_computer_science | 1|none | 0|acc |↑ |0.2300|± |0.0423|
138
+ | - high_school_mathematics | 1|none | 0|acc |↑ |0.2037|± |0.0246|
139
+ | - high_school_physics | 1|none | 0|acc |↑ |0.1921|± |0.0322|
140
+ | - high_school_statistics | 1|none | 0|acc |↑ |0.1852|± |0.0265|
141
+ | - machine_learning | 1|none | 0|acc |↑ |0.2857|± |0.0429|
142
+ |truthfulqa_mc2 | 2|none | 0|acc |↑ |0.4650|± |0.0161|
143
+ |winogrande | 1|none | 0|acc |↑ |0.4957|± |0.0141|
144
+
145
+ | Groups |Version|Filter|n-shot|Metric| |Value | |Stderr|
146
+ |------------------|------:|------|------|------|---|-----:|---|-----:|
147
+ |mmlu | 2|none | |acc |↑ |0.2377|± |0.0036|
148
+ | - humanities | 2|none | |acc |↑ |0.2372|± |0.0062|
149
+ | - other | 2|none | |acc |↑ |0.2462|± |0.0077|
150
+ | - social sciences| 2|none | |acc |↑ |0.2525|± |0.0078|
151
+ | - stem | 2|none | |acc |↑ |0.2157|± |0.0073|
152
+
153
+
154
+ ```bash
155
+ litgpt evaluate --tasks 'leaderboard' --out_dir 'evaluate-leaderboard/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
156
+ ```
157
+
158
+ | Tasks |Version|Filter|n-shot| Metric | |Value | |Stderr|
159
+ |-----------------------------------------------------------|-------|------|-----:|-----------------------|---|-----:|---|------|
160
+ |leaderboard | N/A| | | | | | | |
161
+ | - leaderboard_bbh | N/A| | | | | | | |
162
+ | - leaderboard_bbh_boolean_expressions | 1|none | 3|acc_norm |↑ |0.4720|± |0.0316|
163
+ | - leaderboard_bbh_causal_judgement | 1|none | 3|acc_norm |↑ |0.5187|± |0.0366|
164
+ | - leaderboard_bbh_date_understanding | 1|none | 3|acc_norm |↑ |0.2000|± |0.0253|
165
+ | - leaderboard_bbh_disambiguation_qa | 1|none | 3|acc_norm |↑ |0.3560|± |0.0303|
166
+ | - leaderboard_bbh_formal_fallacies | 1|none | 3|acc_norm |↑ |0.4640|± |0.0316|
167
+ | - leaderboard_bbh_geometric_shapes | 1|none | 3|acc_norm |↑ |0.0800|± |0.0172|
168
+ | - leaderboard_bbh_hyperbaton | 1|none | 3|acc_norm |↑ |0.5160|± |0.0317|
169
+ | - leaderboard_bbh_logical_deduction_five_objects | 1|none | 3|acc_norm |↑ |0.1760|± |0.0241|
170
+ | - leaderboard_bbh_logical_deduction_seven_objects | 1|none | 3|acc_norm |↑ |0.1600|± |0.0232|
171
+ | - leaderboard_bbh_logical_deduction_three_objects | 1|none | 3|acc_norm |↑ |0.3320|± |0.0298|
172
+ | - leaderboard_bbh_movie_recommendation | 1|none | 3|acc_norm |↑ |0.2640|± |0.0279|
173
+ | - leaderboard_bbh_navigate | 1|none | 3|acc_norm |↑ |0.5840|± |0.0312|
174
+ | - leaderboard_bbh_object_counting | 1|none | 3|acc_norm |↑ |0.0840|± |0.0176|
175
+ | - leaderboard_bbh_penguins_in_a_table | 1|none | 3|acc_norm |↑ |0.1986|± |0.0331|
176
+ | - leaderboard_bbh_reasoning_about_colored_objects | 1|none | 3|acc_norm |↑ |0.1720|± |0.0239|
177
+ | - leaderboard_bbh_ruin_names | 1|none | 3|acc_norm |↑ |0.2360|± |0.0269|
178
+ | - leaderboard_bbh_salient_translation_error_detection | 1|none | 3|acc_norm |↑ |0.1560|± |0.0230|
179
+ | - leaderboard_bbh_snarks | 1|none | 3|acc_norm |↑ |0.4663|± |0.0375|
180
+ | - leaderboard_bbh_sports_understanding | 1|none | 3|acc_norm |↑ |0.4520|± |0.0315|
181
+ | - leaderboard_bbh_temporal_sequences | 1|none | 3|acc_norm |↑ |0.2160|± |0.0261|
182
+ | - leaderboard_bbh_tracking_shuffled_objects_five_objects | 1|none | 3|acc_norm |↑ |0.2200|± |0.0263|
183
+ | - leaderboard_bbh_tracking_shuffled_objects_seven_objects| 1|none | 3|acc_norm |↑ |0.1480|± |0.0225|
184
+ | - leaderboard_bbh_tracking_shuffled_objects_three_objects| 1|none | 3|acc_norm |↑ |0.3200|± |0.0296|
185
+ | - leaderboard_bbh_web_of_lies | 1|none | 3|acc_norm |↑ |0.4880|± |0.0317|
186
+ | - leaderboard_gpqa | N/A| | | | | | | |
187
+ | - leaderboard_gpqa_diamond | 1|none | 0|acc_norm |↑ |0.2020|± |0.0286|
188
+ | - leaderboard_gpqa_extended | 1|none | 0|acc_norm |↑ |0.2656|± |0.0189|
189
+ | - leaderboard_gpqa_main | 1|none | 0|acc_norm |↑ |0.2656|± |0.0209|
190
+ | - leaderboard_ifeval | 3|none | 0|inst_level_loose_acc |↑ |0.2290|± | N/A|
191
+ | | |none | 0|inst_level_strict_acc |↑ |0.1990|± | N/A|
192
+ | | |none | 0|prompt_level_loose_acc |↑ |0.1128|± |0.0136|
193
+ | | |none | 0|prompt_level_strict_acc|↑ |0.1017|± |0.0130|
194
+ | - leaderboard_math_hard | N/A| | | | | | | |
195
+ | - leaderboard_math_algebra_hard | 1|none | 4|exact_match |↑ |0.0000|± | 0|
196
+ | - leaderboard_math_counting_and_prob_hard | 1|none | 4|exact_match |↑ |0.0000|± | 0|
197
+ | - leaderboard_math_geometry_hard | 1|none | 4|exact_match |↑ |0.0000|± | 0|
198
+ | - leaderboard_math_intermediate_algebra_hard | 1|none | 4|exact_match |↑ |0.0000|± | 0|
199
+ | - leaderboard_math_num_theory_hard | 1|none | 4|exact_match |↑ |0.0000|± | 0|
200
+ | - leaderboard_math_prealgebra_hard | 1|none | 4|exact_match |↑ |0.0000|± | 0|
201
+ | - leaderboard_math_precalculus_hard | 1|none | 4|exact_match |↑ |0.0000|± | 0|
202
+ | - leaderboard_mmlu_pro | 0.1|none | 5|acc |↑ |0.1104|± |0.0029|
203
+ | - leaderboard_musr | N/A| | | | | | | |
204
+ | - leaderboard_musr_murder_mysteries | 1|none | 0|acc_norm |↑ |0.4920|± |0.0317|
205
+ | - leaderboard_musr_object_placements | 1|none | 0|acc_norm |↑ |0.2891|± |0.0284|
206
+ | - leaderboard_musr_team_allocation | 1|none | 0|acc_norm |↑ |0.3440|± |0.0301|
207
+
208
+ ```bash
209
+ 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/
210
+ ```
211
+
212
+ None
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
+ | Tasks |Version| Filter |n-shot| Metric | |Value | |Stderr|
219
+ |---------------------------------------|------:|--------------|-----:|-----------|---|-----:|---|-----:|
220
+ |mmlu | 2|none | |acc |↑ |0.2377|± |0.0036|
221
+ | - humanities | 2|none | |acc |↑ |0.2372|± |0.0062|
222
+ | - formal_logic | 1|none | 0|acc |↑ |0.2619|± |0.0393|
223
+ | - high_school_european_history | 1|none | 0|acc |↑ |0.2182|± |0.0323|
224
+ | - high_school_us_history | 1|none | 0|acc |↑ |0.2500|± |0.0304|
225
+ | - high_school_world_history | 1|none | 0|acc |↑ |0.2447|± |0.0280|
226
+ | - international_law | 1|none | 0|acc |↑ |0.2066|± |0.0370|
227
+ | - jurisprudence | 1|none | 0|acc |↑ |0.2778|± |0.0433|
228
+ | - logical_fallacies | 1|none | 0|acc |↑ |0.2025|± |0.0316|
229
+ | - moral_disputes | 1|none | 0|acc |↑ |0.2514|± |0.0234|
230
+ | - moral_scenarios | 1|none | 0|acc |↑ |0.2425|± |0.0143|
231
+ | - philosophy | 1|none | 0|acc |↑ |0.1768|± |0.0217|
232
+ | - prehistory | 1|none | 0|acc |↑ |0.2562|± |0.0243|
233
+ | - professional_law | 1|none | 0|acc |↑ |0.2379|± |0.0109|
234
+ | - world_religions | 1|none | 0|acc |↑ |0.2515|± |0.0333|
235
+ | - other | 2|none | |acc |↑ |0.2462|± |0.0077|
236
+ | - business_ethics | 1|none | 0|acc |↑ |0.2800|± |0.0451|
237
+ | - clinical_knowledge | 1|none | 0|acc |↑ |0.2377|± |0.0262|
238
+ | - college_medicine | 1|none | 0|acc |↑ |0.2370|± |0.0324|
239
+ | - global_facts | 1|none | 0|acc |↑ |0.2500|± |0.0435|
240
+ | - human_aging | 1|none | 0|acc |↑ |0.2108|± |0.0274|
241
+ | - management | 1|none | 0|acc |↑ |0.1942|± |0.0392|
242
+ | - marketing | 1|none | 0|acc |↑ |0.2436|± |0.0281|
243
+ | - medical_genetics | 1|none | 0|acc |↑ |0.2700|± |0.0446|
244
+ | - miscellaneous | 1|none | 0|acc |↑ |0.2554|± |0.0156|
245
+ | - nutrition | 1|none | 0|acc |↑ |0.2778|± |0.0256|
246
+ | - professional_accounting | 1|none | 0|acc |↑ |0.2411|± |0.0255|
247
+ | - professional_medicine | 1|none | 0|acc |↑ |0.2279|± |0.0255|
248
+ | - virology | 1|none | 0|acc |↑ |0.2530|± |0.0338|
249
+ | - social sciences | 2|none | |acc |↑ |0.2525|± |0.0078|
250
+ | - econometrics | 1|none | 0|acc |↑ |0.2281|± |0.0395|
251
+ | - high_school_geography | 1|none | 0|acc |↑ |0.1465|± |0.0252|
252
+ | - high_school_government_and_politics| 1|none | 0|acc |↑ |0.2539|± |0.0314|
253
+ | - high_school_macroeconomics | 1|none | 0|acc |↑ |0.2333|± |0.0214|
254
+ | - high_school_microeconomics | 1|none | 0|acc |↑ |0.2269|± |0.0272|
255
+ | - high_school_psychology | 1|none | 0|acc |↑ |0.2330|± |0.0181|
256
+ | - human_sexuality | 1|none | 0|acc |↑ |0.2824|± |0.0395|
257
+ | - professional_psychology | 1|none | 0|acc |↑ |0.2859|± |0.0183|
258
+ | - public_relations | 1|none | 0|acc |↑ |0.2364|± |0.0407|
259
+ | - security_studies | 1|none | 0|acc |↑ |0.3388|± |0.0303|
260
+ | - sociology | 1|none | 0|acc |↑ |0.2637|± |0.0312|
261
+ | - us_foreign_policy | 1|none | 0|acc |↑ |0.2700|± |0.0446|
262
+ | - stem | 2|none | |acc |↑ |0.2157|± |0.0073|
263
+ | - abstract_algebra | 1|none | 0|acc |↑ |0.2000|± |0.0402|
264
+ | - anatomy | 1|none | 0|acc |↑ |0.2222|± |0.0359|
265
+ | - astronomy | 1|none | 0|acc |↑ |0.1842|± |0.0315|
266
+ | - college_biology | 1|none | 0|acc |↑ |0.2639|± |0.0369|
267
+ | - college_chemistry | 1|none | 0|acc |↑ |0.2100|± |0.0409|
268
+ | - college_computer_science | 1|none | 0|acc |↑ |0.2400|± |0.0429|
269
+ | - college_mathematics | 1|none | 0|acc |↑ |0.2200|± |0.0416|
270
+ | - college_physics | 1|none | 0|acc |↑ |0.2059|± |0.0402|
271
+ | - computer_security | 1|none | 0|acc |↑ |0.2400|± |0.0429|
272
+ | - conceptual_physics | 1|none | 0|acc |↑ |0.2553|± |0.0285|
273
+ | - electrical_engineering | 1|none | 0|acc |↑ |0.2414|± |0.0357|
274
+ | - elementary_mathematics | 1|none | 0|acc |↑ |0.2328|± |0.0218|
275
+ | - high_school_biology | 1|none | 0|acc |↑ |0.1839|± |0.0220|
276
+ | - high_school_chemistry | 1|none | 0|acc |↑ |0.1626|± |0.0260|
277
+ | - high_school_computer_science | 1|none | 0|acc |↑ |0.2300|± |0.0423|
278
+ | - high_school_mathematics | 1|none | 0|acc |↑ |0.2037|± |0.0246|
279
+ | - high_school_physics | 1|none | 0|acc |↑ |0.1921|± |0.0322|
280
+ | - high_school_statistics | 1|none | 0|acc |↑ |0.1852|± |0.0265|
281
+ | - machine_learning | 1|none | 0|acc |↑ |0.2857|± |0.0429|
282
+ |mmlu_pro | 2|custom-extract| |exact_match|↑ |0.0000|± |0.0000|
283
+ | - biology | 1|custom-extract| 5|exact_match|↑ |0.0000|± |0.0000|
284
+ | - business | 1|custom-extract| 5|exact_match|↑ |0.0000|± |0.0000|
285
+ | - chemistry | 1|custom-extract| 5|exact_match|↑ |0.0000|± |0.0000|
286
+ | - computer_science | 1|custom-extract| 5|exact_match|↑ |0.0000|± |0.0000|
287
+ | - economics | 1|custom-extract| 5|exact_match|↑ |0.0000|± |0.0000|
288
+ | - engineering | 1|custom-extract| 5|exact_match|↑ |0.0000|± |0.0000|
289
+ | - health | 1|custom-extract| 5|exact_match|↑ |0.0000|± |0.0000|
290
+ | - history | 1|custom-extract| 5|exact_match|↑ |0.0000|± |0.0000|
291
+ | - law | 1|custom-extract| 5|exact_match|↑ |0.0000|± |0.0000|
292
+ | - math | 1|custom-extract| 5|exact_match|↑ |0.0000|± |0.0000|
293
+ | - other | 1|custom-extract| 5|exact_match|↑ |0.0000|± |0.0000|
294
+ | - philosophy | 1|custom-extract| 5|exact_match|↑ |0.0000|± |0.0000|
295
+ | - physics | 1|custom-extract| 5|exact_match|↑ |0.0000|± |0.0000|
296
+ | - psychology | 1|custom-extract| 5|exact_match|↑ |0.0000|± |0.0000|
297
+
298
+ | Groups |Version| Filter |n-shot| Metric | |Value | |Stderr|
299
+ |------------------|------:|--------------|------|-----------|---|-----:|---|-----:|
300
+ |mmlu | 2|none | |acc |↑ |0.2377|± |0.0036|
301
+ | - humanities | 2|none | |acc |↑ |0.2372|± |0.0062|
302
+ | - other | 2|none | |acc |↑ |0.2462|± |0.0077|
303
+ | - social sciences| 2|none | |acc |↑ |0.2525|± |0.0078|
304
+ | - stem | 2|none | |acc |↑ |0.2157|± |0.0073|
305
+ |mmlu_pro | 2|custom-extract| |exact_match|↑ |0.0000|± |0.0000|
306
+
307
+
308
+ ```bash
309
+ 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/
310
+ ```
311
+
312
+ | Tasks |Version| Filter |n-shot| Metric | |Value | |Stderr|
313
+ |-------------------------------|------:|----------------|-----:|-----------|---|-----:|---|-----:|
314
+ |arc_challenge | 1|none | 0|acc |↑ |0.1903|± |0.0115|
315
+ | | |none | 0|acc_norm |↑ |0.2406|± |0.0125|
316
+ |boolq | 2|none | 0|acc |↑ |0.5838|± |0.0086|
317
+ |gpqa_diamond_cot_n_shot | 2|flexible-extract| 0|exact_match|↑ |0.1212|± |0.0233|
318
+ | | |strict-match | 0|exact_match|↑ |0.0000|± |0.0000|
319
+ |gpqa_diamond_cot_zeroshot | 1|flexible-extract| 0|exact_match|↑ |0.1465|± |0.0252|
320
+ | | |strict-match | 0|exact_match|↑ |0.0000|± |0.0000|
321
+ |gpqa_diamond_generative_n_shot | 2|flexible-extract| 0|exact_match|↑ |0.2273|± |0.0299|
322
+ | | |strict-match | 0|exact_match|↑ |0.0000|± |0.0000|
323
+ |gpqa_diamond_n_shot | 2|none | 0|acc |↑ |0.2475|± |0.0307|
324
+ | | |none | 0|acc_norm |↑ |0.2475|± |0.0307|
325
+ |gpqa_diamond_zeroshot | 1|none | 0|acc |↑ |0.2273|± |0.0299|
326
+ | | |none | 0|acc_norm |↑ |0.2273|± |0.0299|
327
+ |gpqa_extended_cot_n_shot | 2|flexible-extract| 0|exact_match|↑ |0.1392|± |0.0148|
328
+ | | |strict-match | 0|exact_match|↑ |0.0000|± |0.0000|
329
+ |gpqa_extended_cot_zeroshot | 1|flexible-extract| 0|exact_match|↑ |0.1502|± |0.0153|
330
+ | | |strict-match | 0|exact_match|↑ |0.0000|± |0.0000|
331
+ |gpqa_extended_generative_n_shot| 2|flexible-extract| 0|exact_match|↑ |0.2289|± |0.0180|
332
+ | | |strict-match | 0|exact_match|↑ |0.0000|± |0.0000|
333
+ |gpqa_extended_n_shot | 2|none | 0|acc |↑ |0.2344|± |0.0181|
334
+ | | |none | 0|acc_norm |↑ |0.2344|± |0.0181|
335
+ |gpqa_extended_zeroshot | 1|none | 0|acc |↑ |0.2582|± |0.0187|
336
+ | | |none | 0|acc_norm |↑ |0.2582|± |0.0187|
337
+ |gpqa_main_cot_n_shot | 2|flexible-extract| 0|exact_match|↑ |0.1429|± |0.0166|
338
+ | | |strict-match | 0|exact_match|↑ |0.0000|± |0.0000|
339
+ |gpqa_main_cot_zeroshot | 1|flexible-extract| 0|exact_match|↑ |0.1629|± |0.0175|
340
+ | | |strict-match | 0|exact_match|↑ |0.0000|± |0.0000|
341
+ |gpqa_main_generative_n_shot | 2|flexible-extract| 0|exact_match|↑ |0.2366|± |0.0201|
342
+ | | |strict-match | 0|exact_match|↑ |0.0000|± |0.0000|
343
+ |gpqa_main_n_shot | 2|none | 0|acc |↑ |0.2500|± |0.0205|
344
+ | | |none | 0|acc_norm |↑ |0.2500|± |0.0205|
345
+ |gpqa_main_zeroshot | 1|none | 0|acc |↑ |0.2746|± |0.0211|
346
+ | | |none | 0|acc_norm |↑ |0.2746|± |0.0211|
347
+ |hellaswag | 1|none | 0|acc |↑ |0.2658|± |0.0044|
348
+ | | |none | 0|acc_norm |↑ |0.2690|± |0.0044|
349
+ |openbookqa | 1|none | 0|acc |↑ |0.1380|± |0.0154|
350
+ | | |none | 0|acc_norm |↑ |0.2740|± |0.0200|
351
+ |piqa | 1|none | 0|acc |↑ |0.5555|± |0.0116|
352
+ | | |none | 0|acc_norm |↑ |0.5571|± |0.0116|
353
+ |truthfulqa_mc2 | 2|none | 0|acc |↑ |0.4650|± |0.0160|
354
+ |winogrande | 1|none | 0|acc |↑ |0.4949|± |0.0141|
355
+
356
+ ```bash
357
+ litgpt evaluate --tasks 'mmlu_multilingual,mgsm' --out_dir 'evaluate-multilinguals/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
358
+ ```
359
+
360
+ None
361
+
362
+ ```bash
363
+ litgpt evaluate --tasks 'gsm8k,mathqa' --out_dir 'evaluate-math/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
364
+ ```
365
+
366
+ |Tasks |Version| Filter |n-shot| Metric | |Value | |Stderr|
367
+ |------|------:|----------------|-----:|-----------|---|-----:|---|-----:|
368
+ |gsm8k | 3|flexible-extract| 5|exact_match|↑ |0.0136|± |0.0032|
369
+ | | |strict-match | 5|exact_match|↑ |0.0008|± |0.0008|
370
+ |mathqa| 1|none | 0|acc |↑ |0.2191|± |0.0076|
371
+ | | |none | 0|acc_norm |↑ |0.2181|± |0.0076|
372
+
373
+ ```bash
374
+ litgpt evaluate --tasks 'wikitext,qasper' --out_dir 'evaluate-long/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
375
+ ```
376
+
377
+ | Tasks |Version|Filter|n-shot| Metric | | Value | |Stderr|
378
+ |---------------|------:|------|-----:|---------------|---|---------:|---|------|
379
+ |qasper_bool | 1|none | 0|f1 |↑ | 0.8215|± |0.0222|
380
+ |qasper_freeform| 2|none | 0|f1_abstractive |↑ | 0.0390|± |0.0045|
381
+ |wikitext | 2|none | 0|bits_per_byte |↓ | 2.6525|± | N/A|
382
+ | | |none | 0|byte_perplexity|↓ | 6.2874|± | N/A|
383
+ | | |none | 0|word_perplexity|↓ |18611.9448|± | N/A|
misc/logo.png ADDED

Git LFS Details

  • SHA256: ca2c2d49dbce4c454798df5969c8328c68580f6fa171680b1221a63f86209bf5
  • Pointer size: 132 Bytes
  • Size of remote file: 3.96 MB
scripts/TRAIN.md CHANGED
@@ -54,11 +54,11 @@ litgpt evaluate --tasks 'bbh_zeroshot,bbh_fewshot,bbh_cot_fewshot,bbh_cot_zerosh
54
 
55
  litgpt evaluate --tasks 'mmlu,mmlu_pro' --out_dir 'evaluate-mmlu/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
56
 
57
- litgpt evaluate --tasks 'arc_challenge,boolq,gpqa,hellaswag,openbookqa,piqa,siqa,truthfulqa_mc2,winogrande' --out_dir 'evaluate-reasoning/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
58
 
59
  litgpt evaluate --tasks 'mmlu_multilingual,mgsm' --out_dir 'evaluate-multilinguals/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
60
 
61
  litgpt evaluate --tasks 'gsm8k,mathqa' --out_dir 'evaluate-math/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
62
 
63
- litgpt evaluate --tasks 'qasper' --out_dir 'evaluate-long/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
64
  ```
 
54
 
55
  litgpt evaluate --tasks 'mmlu,mmlu_pro' --out_dir 'evaluate-mmlu/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
56
 
57
+ 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/
58
 
59
  litgpt evaluate --tasks 'mmlu_multilingual,mgsm' --out_dir 'evaluate-multilinguals/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
60
 
61
  litgpt evaluate --tasks 'gsm8k,mathqa' --out_dir 'evaluate-math/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
62
 
63
+ litgpt evaluate --tasks 'wikitext,qasper' --out_dir 'evaluate-long/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
64
  ```