metadata
license: other
This is the merged hf tr version of llama 30B and OA's rlhf 30B xor weights:
https://huggingface.co/OpenAssistant/oasst-rlhf-2-llama-30b-7k-steps-xor
This the md5 checksum that I get locally, which matchs the original repo suggests
deb33dd4ffc3d2baddcce275a00b7c1b ./tokenizer.json
ed59bfee4e87b9193fea5897d610ab24 ./tokenizer_config.json
704373f0c0d62be75e5f7d41d39a7e57 ./special_tokens_map.json
4c5941b4ee12dc0d8e6b5ca3f6819f4d ./pytorch_model-00004-of-00007.bin
13a3641423840eb89f9a86507a90b2bf ./pytorch_model.bin.index.json
d08594778f00abe70b93899628e41246 ./pytorch_model-00007-of-00007.bin
9a4d2468ecf85bf07420b200faefb4af ./config.json
2c92d306969c427275f34b4ebf66f087 ./pytorch_model-00006-of-00007.bin
148bfd184af630a7633b4de2f41bfc49 ./generation_config.json
b6e90377103e9270cbe46b13aed288ec ./pytorch_model-00005-of-00007.bin
27b0dc092f99aa2efaf467b2d8026c3f ./added_tokens.json
ed991042b2a449123824f689bb94b29e ./pytorch_model-00002-of-00007.bin
f11acc069334434d68c45a80ee899fe5 ./pytorch_model-00003-of-00007.bin
9f41bd4d5720d28567b3e7820b4a8023 ./pytorch_model-00001-of-00007.bin
eeec4125e9c7560836b4873b6f8e3025 ./tokenizer.model
Open LLM Leaderboard Evaluation Results
Detailed results can be found here
| Metric | Value |
|---|---|
| Avg. | 53.18 |
| ARC (25-shot) | 61.35 |
| HellaSwag (10-shot) | 83.8 |
| MMLU (5-shot) | 57.89 |
| TruthfulQA (0-shot) | 51.18 |
| Winogrande (5-shot) | 78.77 |
| GSM8K (5-shot) | 31.46 |
| DROP (3-shot) | 7.78 |