eval
Browse files- README.md +91 -1
- scripts/TRAIN.md +4 -2
README.md
CHANGED
|
@@ -28,7 +28,9 @@ tags:
|
|
| 28 |
|
| 29 |
A pretrained language model based on the Llama model with about **33M** parameters. This model has been trained on **9.7B** (`9,782,206,713`) tokens from more than **5.2M** (`5,285,575`) dataset rows.
|
| 30 |
|
| 31 |
-
This model
|
|
|
|
|
|
|
| 32 |
|
| 33 |
[loss, val_loss](https://api.wandb.ai/links/mtasic85/z591qpyv)
|
| 34 |
|
|
@@ -40,6 +42,10 @@ This model is designed not for immediate use but rather for Continued Pretrainin
|
|
| 40 |
|
| 41 |
## lm-evaluation-harness
|
| 42 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 43 |
| Tasks |Version|Filter|n-shot| Metric | |Value | |Stderr|
|
| 44 |
|-----------------------------------------------------------|-------|------|-----:|-----------------------|---|-----:|---|------|
|
| 45 |
|leaderboard | N/A| | | | | | | |
|
|
@@ -90,3 +96,87 @@ This model is designed not for immediate use but rather for Continued Pretrainin
|
|
| 90 |
| - leaderboard_musr_object_placements | 1|none | 0|acc_norm |↑ |0.2930|± |0.0285|
|
| 91 |
| - leaderboard_musr_team_allocation | 1|none | 0|acc_norm |↑ |0.3720|± |0.0306|
|
| 92 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 28 |
|
| 29 |
A pretrained language model based on the Llama model with about **33M** parameters. This model has been trained on **9.7B** (`9,782,206,713`) tokens from more than **5.2M** (`5,285,575`) dataset rows.
|
| 30 |
|
| 31 |
+
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.
|
| 32 |
+
|
| 33 |
+
The objective is to streamline the cognitive or reasoning core, eliminating any redundant knowledge from the model.
|
| 34 |
|
| 35 |
[loss, val_loss](https://api.wandb.ai/links/mtasic85/z591qpyv)
|
| 36 |
|
|
|
|
| 42 |
|
| 43 |
## lm-evaluation-harness
|
| 44 |
|
| 45 |
+
```bash
|
| 46 |
+
litgpt evaluate --tasks 'leaderboard' --out_dir 'evaluate-0/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
|
| 47 |
+
```
|
| 48 |
+
|
| 49 |
| Tasks |Version|Filter|n-shot| Metric | |Value | |Stderr|
|
| 50 |
|-----------------------------------------------------------|-------|------|-----:|-----------------------|---|-----:|---|------|
|
| 51 |
|leaderboard | N/A| | | | | | | |
|
|
|
|
| 96 |
| - leaderboard_musr_object_placements | 1|none | 0|acc_norm |↑ |0.2930|± |0.0285|
|
| 97 |
| - leaderboard_musr_team_allocation | 1|none | 0|acc_norm |↑ |0.3720|± |0.0306|
|
| 98 |
|
| 99 |
+
```bash
|
| 100 |
+
litgpt evaluate --tasks 'hellaswag,gsm8k,truthfulqa_mc2,mmlu,winogrande,arc_challenge' --out_dir 'evaluate-1/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
|
| 101 |
+
```
|
| 102 |
+
|
| 103 |
+
| Tasks |Version| Filter |n-shot| Metric | |Value | |Stderr|
|
| 104 |
+
|---------------------------------------|------:|----------------|-----:|-----------|---|-----:|---|-----:|
|
| 105 |
+
|arc_challenge | 1|none | 0|acc |↑ |0.2048|± |0.0118|
|
| 106 |
+
| | |none | 0|acc_norm |↑ |0.2372|± |0.0124|
|
| 107 |
+
|gsm8k | 3|flexible-extract| 5|exact_match|↑ |0.0091|± |0.0026|
|
| 108 |
+
| | |strict-match | 5|exact_match|↑ |0.0000|± |0.0000|
|
| 109 |
+
|hellaswag | 1|none | 0|acc |↑ |0.2613|± |0.0044|
|
| 110 |
+
| | |none | 0|acc_norm |↑ |0.2605|± |0.0044|
|
| 111 |
+
|mmlu | 2|none | |acc |↑ |0.2280|± |0.0035|
|
| 112 |
+
| - humanities | 2|none | |acc |↑ |0.2366|± |0.0062|
|
| 113 |
+
| - formal_logic | 1|none | 0|acc |↑ |0.2778|± |0.0401|
|
| 114 |
+
| - high_school_european_history | 1|none | 0|acc |↑ |0.2061|± |0.0316|
|
| 115 |
+
| - high_school_us_history | 1|none | 0|acc |↑ |0.2402|± |0.0300|
|
| 116 |
+
| - high_school_world_history | 1|none | 0|acc |↑ |0.2827|± |0.0293|
|
| 117 |
+
| - international_law | 1|none | 0|acc |↑ |0.2149|± |0.0375|
|
| 118 |
+
| - jurisprudence | 1|none | 0|acc |↑ |0.2315|± |0.0408|
|
| 119 |
+
| - logical_fallacies | 1|none | 0|acc |↑ |0.2209|± |0.0326|
|
| 120 |
+
| - moral_disputes | 1|none | 0|acc |↑ |0.2370|± |0.0229|
|
| 121 |
+
| - moral_scenarios | 1|none | 0|acc |↑ |0.2380|± |0.0142|
|
| 122 |
+
| - philosophy | 1|none | 0|acc |↑ |0.1833|± |0.0220|
|
| 123 |
+
| - prehistory | 1|none | 0|acc |↑ |0.2191|± |0.0230|
|
| 124 |
+
| - professional_law | 1|none | 0|acc |↑ |0.2360|± |0.0108|
|
| 125 |
+
| - world_religions | 1|none | 0|acc |↑ |0.3275|± |0.0360|
|
| 126 |
+
| - other | 2|none | |acc |↑ |0.2485|± |0.0077|
|
| 127 |
+
| - business_ethics | 1|none | 0|acc |↑ |0.3100|± |0.0465|
|
| 128 |
+
| - clinical_knowledge | 1|none | 0|acc |↑ |0.2226|± |0.0256|
|
| 129 |
+
| - college_medicine | 1|none | 0|acc |↑ |0.2486|± |0.0330|
|
| 130 |
+
| - global_facts | 1|none | 0|acc |↑ |0.1700|± |0.0378|
|
| 131 |
+
| - human_aging | 1|none | 0|acc |↑ |0.3229|± |0.0314|
|
| 132 |
+
| - management | 1|none | 0|acc |↑ |0.1748|± |0.0376|
|
| 133 |
+
| - marketing | 1|none | 0|acc |↑ |0.3034|± |0.0301|
|
| 134 |
+
| - medical_genetics | 1|none | 0|acc |↑ |0.3100|± |0.0465|
|
| 135 |
+
| - miscellaneous | 1|none | 0|acc |↑ |0.2414|± |0.0153|
|
| 136 |
+
| - nutrition | 1|none | 0|acc |↑ |0.2484|± |0.0247|
|
| 137 |
+
| - professional_accounting | 1|none | 0|acc |↑ |0.2411|± |0.0255|
|
| 138 |
+
| - professional_medicine | 1|none | 0|acc |↑ |0.1838|± |0.0235|
|
| 139 |
+
| - virology | 1|none | 0|acc |↑ |0.2831|± |0.0351|
|
| 140 |
+
| - social sciences | 2|none | |acc |↑ |0.2158|± |0.0074|
|
| 141 |
+
| - econometrics | 1|none | 0|acc |↑ |0.2368|± |0.0400|
|
| 142 |
+
| - high_school_geography | 1|none | 0|acc |↑ |0.1768|± |0.0272|
|
| 143 |
+
| - high_school_government_and_politics| 1|none | 0|acc |↑ |0.1969|± |0.0287|
|
| 144 |
+
| - high_school_macroeconomics | 1|none | 0|acc |↑ |0.2103|± |0.0207|
|
| 145 |
+
| - high_school_microeconomics | 1|none | 0|acc |↑ |0.2143|± |0.0267|
|
| 146 |
+
| - high_school_psychology | 1|none | 0|acc |↑ |0.1890|± |0.0168|
|
| 147 |
+
| - human_sexuality | 1|none | 0|acc |↑ |0.2672|± |0.0388|
|
| 148 |
+
| - professional_psychology | 1|none | 0|acc |↑ |0.2451|± |0.0174|
|
| 149 |
+
| - public_relations | 1|none | 0|acc |↑ |0.2091|± |0.0390|
|
| 150 |
+
| - security_studies | 1|none | 0|acc |↑ |0.1755|± |0.0244|
|
| 151 |
+
| - sociology | 1|none | 0|acc |↑ |0.2488|± |0.0306|
|
| 152 |
+
| - us_foreign_policy | 1|none | 0|acc |↑ |0.2700|± |0.0446|
|
| 153 |
+
| - stem | 2|none | |acc |↑ |0.2068|± |0.0072|
|
| 154 |
+
| - abstract_algebra | 1|none | 0|acc |↑ |0.1700|± |0.0378|
|
| 155 |
+
| - anatomy | 1|none | 0|acc |↑ |0.1778|± |0.0330|
|
| 156 |
+
| - astronomy | 1|none | 0|acc |↑ |0.1842|± |0.0315|
|
| 157 |
+
| - college_biology | 1|none | 0|acc |↑ |0.2569|± |0.0365|
|
| 158 |
+
| - college_chemistry | 1|none | 0|acc |↑ |0.1900|± |0.0394|
|
| 159 |
+
| - college_computer_science | 1|none | 0|acc |↑ |0.2600|± |0.0441|
|
| 160 |
+
| - college_mathematics | 1|none | 0|acc |↑ |0.2100|± |0.0409|
|
| 161 |
+
| - college_physics | 1|none | 0|acc |↑ |0.2059|± |0.0402|
|
| 162 |
+
| - computer_security | 1|none | 0|acc |↑ |0.2400|± |0.0429|
|
| 163 |
+
| - conceptual_physics | 1|none | 0|acc |↑ |0.2681|± |0.0290|
|
| 164 |
+
| - electrical_engineering | 1|none | 0|acc |↑ |0.2345|± |0.0353|
|
| 165 |
+
| - elementary_mathematics | 1|none | 0|acc |↑ |0.2011|± |0.0206|
|
| 166 |
+
| - high_school_biology | 1|none | 0|acc |↑ |0.1839|± |0.0220|
|
| 167 |
+
| - high_school_chemistry | 1|none | 0|acc |↑ |0.1527|± |0.0253|
|
| 168 |
+
| - high_school_computer_science | 1|none | 0|acc |↑ |0.2400|± |0.0429|
|
| 169 |
+
| - high_school_mathematics | 1|none | 0|acc |↑ |0.2111|± |0.0249|
|
| 170 |
+
| - high_school_physics | 1|none | 0|acc |↑ |0.1722|± |0.0308|
|
| 171 |
+
| - high_school_statistics | 1|none | 0|acc |↑ |0.1667|± |0.0254|
|
| 172 |
+
| - machine_learning | 1|none | 0|acc |↑ |0.2768|± |0.0425|
|
| 173 |
+
|truthfulqa_mc2 | 2|none | 0|acc |↑ |0.4971|± |0.0165|
|
| 174 |
+
|winogrande | 1|none | 0|acc |↑ |0.5114|± |0.0140|
|
| 175 |
+
|
| 176 |
+
| Groups |Version|Filter|n-shot|Metric| |Value | |Stderr|
|
| 177 |
+
|------------------|------:|------|------|------|---|-----:|---|-----:|
|
| 178 |
+
|mmlu | 2|none | |acc |↑ |0.2280|± |0.0035|
|
| 179 |
+
| - humanities | 2|none | |acc |↑ |0.2366|± |0.0062|
|
| 180 |
+
| - other | 2|none | |acc |↑ |0.2485|± |0.0077|
|
| 181 |
+
| - social sciences| 2|none | |acc |↑ |0.2158|± |0.0074|
|
| 182 |
+
| - stem | 2|none | |acc |↑ |0.2068|± |0.0072|
|
scripts/TRAIN.md
CHANGED
|
@@ -57,7 +57,9 @@ model.save_pretrained('out/converted_model/')
|
|
| 57 |
## Evaluate
|
| 58 |
|
| 59 |
```bash
|
| 60 |
-
|
| 61 |
|
| 62 |
-
litgpt evaluate --tasks 'hellaswag,gsm8k,truthfulqa_mc2,mmlu,
|
|
|
|
|
|
|
| 63 |
```
|
|
|
|
| 57 |
## Evaluate
|
| 58 |
|
| 59 |
```bash
|
| 60 |
+
litgpt evaluate --tasks 'leaderboard' --out_dir 'evaluate-0/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
|
| 61 |
|
| 62 |
+
litgpt evaluate --tasks 'hellaswag,gsm8k,truthfulqa_mc2,mmlu,winogrande,arc_challenge' --out_dir 'evaluate-1/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
|
| 63 |
+
|
| 64 |
+
litgpt evaluate --tasks 'mmlu_pro,ifeval,mgsm_direct,mathqa,gpqa' --out_dir 'evaluate-2/' --batch_size 4 --dtype 'bfloat16' out/pretrain/final/
|
| 65 |
```
|