tangled-llama-108m-32k-base-v0.1

logo

A pretrained language model based on the Llama model with about 108M parameters. This model has been trained on 9.7B (9,782,206,713) tokens from more than 5.2M (5,285,575) dataset rows.

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 32K (32,768) tokens, it was pretrained with sequences of 2K (2048) tokens.

The objective is to streamline the cognitive or reasoning core, eliminating any redundant knowledge from the model.

loss, val_loss

val_ppl

epoch

learning_rate

lm-evaluation-harness

litgpt evaluate --tasks 'leaderboard' --out_dir 'evaluate-0/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
Tasks Version Filter n-shot Metric Value Stderr
leaderboard N/A
- leaderboard_bbh N/A
- leaderboard_bbh_boolean_expressions 1 none 3 acc_norm 0.5040 ± 0.0317
- leaderboard_bbh_causal_judgement 1 none 3 acc_norm 0.5134 ± 0.0366
- leaderboard_bbh_date_understanding 1 none 3 acc_norm 0.1880 ± 0.0248
- leaderboard_bbh_disambiguation_qa 1 none 3 acc_norm 0.3000 ± 0.0290
- leaderboard_bbh_formal_fallacies 1 none 3 acc_norm 0.4680 ± 0.0316
- leaderboard_bbh_geometric_shapes 1 none 3 acc_norm 0.0560 ± 0.0146
- leaderboard_bbh_hyperbaton 1 none 3 acc_norm 0.4960 ± 0.0317
- leaderboard_bbh_logical_deduction_five_objects 1 none 3 acc_norm 0.1680 ± 0.0237
- leaderboard_bbh_logical_deduction_seven_objects 1 none 3 acc_norm 0.1400 ± 0.0220
- leaderboard_bbh_logical_deduction_three_objects 1 none 3 acc_norm 0.3360 ± 0.0299
- leaderboard_bbh_movie_recommendation 1 none 3 acc_norm 0.2280 ± 0.0266
- leaderboard_bbh_navigate 1 none 3 acc_norm 0.4720 ± 0.0316
- leaderboard_bbh_object_counting 1 none 3 acc_norm 0.1000 ± 0.0190
- leaderboard_bbh_penguins_in_a_table 1 none 3 acc_norm 0.1918 ± 0.0327
- leaderboard_bbh_reasoning_about_colored_objects 1 none 3 acc_norm 0.1480 ± 0.0225
- leaderboard_bbh_ruin_names 1 none 3 acc_norm 0.2920 ± 0.0288
- leaderboard_bbh_salient_translation_error_detection 1 none 3 acc_norm 0.1520 ± 0.0228
- leaderboard_bbh_snarks 1 none 3 acc_norm 0.4494 ± 0.0374
- leaderboard_bbh_sports_understanding 1 none 3 acc_norm 0.4600 ± 0.0316
- leaderboard_bbh_temporal_sequences 1 none 3 acc_norm 0.2480 ± 0.0274
- leaderboard_bbh_tracking_shuffled_objects_five_objects 1 none 3 acc_norm 0.1920 ± 0.0250
- leaderboard_bbh_tracking_shuffled_objects_seven_objects 1 none 3 acc_norm 0.1560 ± 0.0230
- leaderboard_bbh_tracking_shuffled_objects_three_objects 1 none 3 acc_norm 0.3000 ± 0.0290
- leaderboard_bbh_web_of_lies 1 none 3 acc_norm 0.5040 ± 0.0317
- leaderboard_gpqa N/A
- leaderboard_gpqa_diamond 1 none 0 acc_norm 0.2222 ± 0.0296
- leaderboard_gpqa_extended 1 none 0 acc_norm 0.2711 ± 0.0190
- leaderboard_gpqa_main 1 none 0 acc_norm 0.2589 ± 0.0207
- leaderboard_ifeval 3 none 0 inst_level_loose_acc 0.2050 ± N/A
none 0 inst_level_strict_acc 0.1966 ± N/A
none 0 prompt_level_loose_acc 0.1072 ± 0.0133
none 0 prompt_level_strict_acc 0.1035 ± 0.0131
- leaderboard_math_hard N/A
- leaderboard_math_algebra_hard 1 none 4 exact_match 0.0000 ± 0
- leaderboard_math_counting_and_prob_hard 1 none 4 exact_match 0.0000 ± 0
- leaderboard_math_geometry_hard 1 none 4 exact_match 0.0000 ± 0
- leaderboard_math_intermediate_algebra_hard 1 none 4 exact_match 0.0000 ± 0
- leaderboard_math_num_theory_hard 1 none 4 exact_match 0.0000 ± 0
- leaderboard_math_prealgebra_hard 1 none 4 exact_match 0.0000 ± 0
- leaderboard_math_precalculus_hard 1 none 4 exact_match 0.0000 ± 0
- leaderboard_mmlu_pro 0.1 none 5 acc 0.1151 ± 0.0029
- leaderboard_musr N/A
- leaderboard_musr_murder_mysteries 1 none 0 acc_norm 0.4840 ± 0.0317
- leaderboard_musr_object_placements 1 none 0 acc_norm 0.3125 ± 0.0290
- leaderboard_musr_team_allocation 1 none 0 acc_norm 0.3840 ± 0.0308
litgpt evaluate --tasks 'hellaswag,gsm8k,truthfulqa_mc2,mmlu,winogrande,arc_challenge' --out_dir 'evaluate-1/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
Tasks Version Filter n-shot Metric Value Stderr
arc_challenge 1 none 0 acc 0.1911 ± 0.0115
none 0 acc_norm 0.2355 ± 0.0124
gsm8k 3 flexible-extract 5 exact_match 0.0152 ± 0.0034
strict-match 5 exact_match 0.0000 ± 0.0000
hellaswag 1 none 0 acc 0.2661 ± 0.0044
none 0 acc_norm 0.2708 ± 0.0044
mmlu 2 none acc 0.2315 ± 0.0036
- humanities 2 none acc 0.2372 ± 0.0062
- formal_logic 1 none 0 acc 0.2937 ± 0.0407
- high_school_european_history 1 none 0 acc 0.2424 ± 0.0335
- high_school_us_history 1 none 0 acc 0.2451 ± 0.0302
- high_school_world_history 1 none 0 acc 0.2321 ± 0.0275
- international_law 1 none 0 acc 0.1983 ± 0.0364
- jurisprudence 1 none 0 acc 0.2315 ± 0.0408
- logical_fallacies 1 none 0 acc 0.1840 ± 0.0304
- moral_disputes 1 none 0 acc 0.2110 ± 0.0220
- moral_scenarios 1 none 0 acc 0.2380 ± 0.0142
- philosophy 1 none 0 acc 0.1961 ± 0.0226
- prehistory 1 none 0 acc 0.2315 ± 0.0235
- professional_law 1 none 0 acc 0.2503 ± 0.0111
- world_religions 1 none 0 acc 0.2865 ± 0.0347
- other 2 none acc 0.2385 ± 0.0076
- business_ethics 1 none 0 acc 0.2900 ± 0.0456
- clinical_knowledge 1 none 0 acc 0.2113 ± 0.0251
- college_medicine 1 none 0 acc 0.1965 ± 0.0303
- global_facts 1 none 0 acc 0.1900 ± 0.0394
- human_aging 1 none 0 acc 0.3004 ± 0.0308
- management 1 none 0 acc 0.1748 ± 0.0376
- marketing 1 none 0 acc 0.2863 ± 0.0296
- medical_genetics 1 none 0 acc 0.2800 ± 0.0451
- miscellaneous 1 none 0 acc 0.2350 ± 0.0152
- nutrition 1 none 0 acc 0.2255 ± 0.0239
- professional_accounting 1 none 0 acc 0.2482 ± 0.0258
- professional_medicine 1 none 0 acc 0.1985 ± 0.0242
- virology 1 none 0 acc 0.2771 ± 0.0348
- social sciences 2 none acc 0.2281 ± 0.0076
- econometrics 1 none 0 acc 0.2105 ± 0.0384
- high_school_geography 1 none 0 acc 0.1818 ± 0.0275
- high_school_government_and_politics 1 none 0 acc 0.2280 ± 0.0303
- high_school_macroeconomics 1 none 0 acc 0.2410 ± 0.0217
- high_school_microeconomics 1 none 0 acc 0.2353 ± 0.0276
- high_school_psychology 1 none 0 acc 0.2055 ± 0.0173
- human_sexuality 1 none 0 acc 0.2595 ± 0.0384
- professional_psychology 1 none 0 acc 0.2418 ± 0.0173
- public_relations 1 none 0 acc 0.2091 ± 0.0390
- security_studies 1 none 0 acc 0.2408 ± 0.0274
- sociology 1 none 0 acc 0.2040 ± 0.0285
- us_foreign_policy 1 none 0 acc 0.3100 ± 0.0465
- stem 2 none acc 0.2195 ± 0.0074
- abstract_algebra 1 none 0 acc 0.2700 ± 0.0446
- anatomy 1 none 0 acc 0.1630 ± 0.0319
- astronomy 1 none 0 acc 0.2303 ± 0.0343
- college_biology 1 none 0 acc 0.2569 ± 0.0365
- college_chemistry 1 none 0 acc 0.2400 ± 0.0429
- college_computer_science 1 none 0 acc 0.2100 ± 0.0409
- college_mathematics 1 none 0 acc 0.2100 ± 0.0409
- college_physics 1 none 0 acc 0.2745 ± 0.0444
- computer_security 1 none 0 acc 0.3000 ± 0.0461
- conceptual_physics 1 none 0 acc 0.1957 ± 0.0259
- electrical_engineering 1 none 0 acc 0.2276 ± 0.0349
- elementary_mathematics 1 none 0 acc 0.2302 ± 0.0217
- high_school_biology 1 none 0 acc 0.1968 ± 0.0226
- high_school_chemistry 1 none 0 acc 0.1527 ± 0.0253
- high_school_computer_science 1 none 0 acc 0.2500 ± 0.0435
- high_school_mathematics 1 none 0 acc 0.1963 ± 0.0242
- high_school_physics 1 none 0 acc 0.2053 ± 0.0330
- high_school_statistics 1 none 0 acc 0.2269 ± 0.0286
- machine_learning 1 none 0 acc 0.2768 ± 0.0425
truthfulqa_mc2 2 none 0 acc 0.4681 ± 0.0159
winogrande 1 none 0 acc 0.5146 ± 0.0140
Groups Version Filter n-shot Metric Value Stderr
mmlu 2 none acc 0.2315 ± 0.0036
- humanities 2 none acc 0.2372 ± 0.0062
- other 2 none acc 0.2385 ± 0.0076
- social sciences 2 none acc 0.2281 ± 0.0076
- stem 2 none acc 0.2195 ± 0.0074
Downloads last month
207
Safetensors
Model size
108M params
Tensor type
BF16
·
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.

Datasets used to train tangledgroup/tangled-llama-108m-32k-base-v0.1