neurips-2023-llm-efficiency
Collection
Fine-tune models, datasets and artifacts used for llm efficiency competition.
https://llm-efficiency-challenge.github.io/challenge
•
15 items
•
Updated
This model is a fine-tuned version of mistralai/Mistral-7B-v0.1 on the None dataset. It achieves the following results on the evaluation set:
More information needed
More information needed
More information needed
The following hyperparameters were used during training:
Training Loss | Epoch | Step | Validation Loss |
---|---|---|---|
0.8335 | 0.06 | 20 | 0.6429 |
0.6725 | 0.12 | 40 | 0.5888 |
0.5927 | 0.18 | 60 | 0.5603 |
0.5847 | 0.24 | 80 | 0.5362 |
0.5552 | 0.3 | 100 | 0.5256 |
0.5511 | 0.36 | 120 | 0.5243 |
0.5466 | 0.42 | 140 | 0.5102 |
0.4395 | 0.48 | 160 | 0.5065 |
0.6854 | 0.54 | 180 | 0.4971 |
0.7326 | 0.6 | 200 | 0.5150 |
0.8204 | 0.66 | 220 | 0.5008 |
0.6009 | 0.72 | 240 | 0.4972 |
0.4471 | 0.78 | 260 | 0.4944 |
0.5934 | 0.84 | 280 | 0.5146 |
0.6574 | 0.9 | 300 | 0.5057 |
0.4566 | 0.96 | 320 | 0.4880 |
0.6119 | 1.02 | 340 | 0.5442 |
0.3779 | 1.08 | 360 | 0.5540 |
0.4431 | 1.14 | 380 | 0.5375 |
0.38 | 1.2 | 400 | 0.5541 |
0.4542 | 1.26 | 420 | 0.5359 |
0.5392 | 1.32 | 440 | 0.5394 |
0.2573 | 1.38 | 460 | 0.5318 |
0.5441 | 1.44 | 480 | 0.5201 |
0.3758 | 1.5 | 500 | 0.5147 |
0.4403 | 1.56 | 520 | 0.5134 |
0.3308 | 1.62 | 540 | 0.5289 |
0.4604 | 1.68 | 560 | 0.5205 |
0.4479 | 1.74 | 580 | 0.5340 |
0.521 | 1.8 | 600 | 0.5094 |
0.32 | 1.86 | 620 | 0.4995 |
0.3984 | 1.92 | 640 | 0.4878 |
0.3799 | 1.98 | 660 | 0.4826 |
0.1484 | 2.04 | 680 | 0.7261 |
0.3305 | 2.1 | 700 | 0.6187 |
0.1477 | 2.16 | 720 | 0.5499 |
0.176 | 2.22 | 740 | 0.5796 |
0.1892 | 2.28 | 760 | 0.5717 |
0.1921 | 2.34 | 780 | 0.5416 |
0.1366 | 2.4 | 800 | 0.5866 |
0.1726 | 2.46 | 820 | 0.5562 |
0.1264 | 2.51 | 840 | 0.5621 |
0.2054 | 2.57 | 860 | 0.5678 |
0.1722 | 2.63 | 880 | 0.5573 |
0.2399 | 2.69 | 900 | 0.5553 |
0.229 | 2.75 | 920 | 0.5565 |
0.1876 | 2.81 | 940 | 0.5609 |
0.2281 | 2.87 | 960 | 0.5633 |
0.1727 | 2.93 | 980 | 0.5645 |
0.3536 | 2.99 | 1000 | 0.5631 |
Base model
mistralai/Mistral-7B-v0.1