vllm (pretrained=/root/autodl-tmp/Homunculus,add_bos_token=true,max_model_len=3096,dtype=bfloat16,trust_remote_code=true,gpu_memory_utilization=0.5), gen_kwargs: (None), limit: 250.0, num_fewshot: 5, batch_size: auto

Tasks Version Filter n-shot Metric Value Stderr
gsm8k 3 flexible-extract 5 exact_match ↑ 0.796 ± 0.0255
strict-match 5 exact_match ↑ 0.796 ± 0.0255

vllm (pretrained=/root/autodl-tmp/Homunculus,add_bos_token=true,max_model_len=3096,dtype=bfloat16,trust_remote_code=true,gpu_memory_utilization=0.5), gen_kwargs: (None), limit: 500.0, num_fewshot: 5, batch_size: auto

Tasks Version Filter n-shot Metric Value Stderr
gsm8k 3 flexible-extract 5 exact_match ↑ 0.796 ± 0.0180
strict-match 5 exact_match ↑ 0.792 ± 0.0182

vllm (pretrained=/root/autodl-tmp/Homunculus,add_bos_token=true,max_model_len=3048,dtype=bfloat16,trust_remote_code=true,gpu_memory_utilization=0.5), gen_kwargs: (None), limit: 15.0, num_fewshot: None, batch_size: auto

Groups Version Filter n-shot Metric Value Stderr
mmlu 2 none acc ↑ 0.6480 ± 0.0153
- humanities 2 none acc ↑ 0.6769 ± 0.0306
- other 2 none acc ↑ 0.6718 ± 0.0330
- social sciences 2 none acc ↑ 0.7444 ± 0.0315
- stem 2 none acc ↑ 0.5509 ± 0.0275

vllm (pretrained=/root/autodl-tmp/Homunculus-90-128-4096-9.9999,add_bos_token=true,max_model_len=3096,dtype=bfloat16,trust_remote_code=true,gpu_memory_utilization=0.4), gen_kwargs: (None), limit: 250.0, num_fewshot: 5, batch_size: auto

Tasks Version Filter n-shot Metric Value Stderr
gsm8k 3 flexible-extract 5 exact_match ↑ 0.796 ± 0.0255
strict-match 5 exact_match ↑ 0.796 ± 0.0255

vllm (pretrained=/root/autodl-tmp/Homunculus-90-128-4096-9.9999,add_bos_token=true,max_model_len=3096,dtype=bfloat16,trust_remote_code=true,gpu_memory_utilization=0.4), gen_kwargs: (None), limit: 250.0, num_fewshot: 5, batch_size: auto

Tasks Version Filter n-shot Metric Value Stderr
gsm8k 3 flexible-extract 5 exact_match ↑ 0.796 ± 0.0255
strict-match 5 exact_match ↑ 0.796 ± 0.0255

vllm (pretrained=/root/autodl-tmp/Homunculus-90-128-4096-9.9999,add_bos_token=true,max_model_len=3048,dtype=bfloat16,trust_remote_code=true,gpu_memory_utilization=0.4), gen_kwargs: (None), limit: 15.0, num_fewshot: None, batch_size: auto

Groups Version Filter n-shot Metric Value Stderr
mmlu 2 none acc ↑ 0.6538 ± 0.0152
- humanities 2 none acc ↑ 0.6872 ± 0.0301
- other 2 none acc ↑ 0.6769 ± 0.0322
- social sciences 2 none acc ↑ 0.7389 ± 0.0314
- stem 2 none acc ↑ 0.5614 ± 0.0277
Downloads last month
8
Safetensors
Model size
12.5B params
Tensor type
BF16
·
I8
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for noneUsername/Homunculus-W8A8

Quantized
(19)
this model