multstage abliteration
vllm (pretrained=/mnt/g/abliterated/Orca-2-13b-f16,add_bos_token=true,max_model_len=2048,dtype=float16,tensor_parallel_size=2,gpu_memory_utilization=0.8,max_num_seqs=3), gen_kwargs: (None), limit: 250.0, num_fewshot: 5, batch_size: 1
| Tasks | Version | Filter | n-shot | Metric | | Value | | Stderr |
|---|
| gsm8k | 3 | flexible-extract | 5 | exact_match | ↑ | 0.496 | ± | 0.0317 |
| | strict-match | 5 | exact_match | ↑ | 0.488 | ± | 0.0317 |
vllm (pretrained=/mnt/g/abliterated/Orca-2-13b-abliterated-round1,add_bos_token=true,max_model_len=2048,dtype=float16,tensor_parallel_size=2,gpu_memory_utilization=0.8,max_num_seqs=3), gen_kwargs: (None), limit: 250.0, num_fewshot: 5, batch_size: 2
| Tasks | Version | Filter | n-shot | Metric | | Value | | Stderr |
|---|
| gsm8k | 3 | flexible-extract | 5 | exact_match | ↑ | 0.536 | ± | 0.0316 |
| | strict-match | 5 | exact_match | ↑ | 0.528 | ± | 0.0316 |
vllm (pretrained=/mnt/g/abliterated/Orca-2-13b-abliterated-round2,add_bos_token=true,max_model_len=2048,dtype=float16,tensor_parallel_size=2,gpu_memory_utilization=0.85,max_num_seqs=4), gen_kwargs: (None), limit: 250.0, num_fewshot: 5, batch_size: 4
| Tasks | Version | Filter | n-shot | Metric | | Value | | Stderr |
|---|
| gsm8k | 3 | flexible-extract | 5 | exact_match | ↑ | 0.512 | ± | 0.0317 |
| | strict-match | 5 | exact_match | ↑ | 0.500 | ± | 0.0317 |
