Text Generation
Transformers
PyTorch
bloom
Eval Results
text-generation-inference
Inference Endpoints
bloomz-7b1-p3 / evaluation_bloomz-7b1-p3 /evaluation_val /agg.limited=3000.model=p31lossseqglobal_step1000.task=wmt14_hi_en.templates=version-en-hi-target.fewshot=0.batchsize=16.seed=1234.timestamp=2022-09-09T23:48:38.json
Muennighoff's picture
Organize eval
8634ce5
raw
history blame
921 Bytes
{"results": [{"task_name": "wmt14_hi_en", "prompt_name": "version-en-hi-target", "bleu": 0.1858574511075315, "fixed_answer_choice_list": null, "dataset_path": "wmt14", "dataset_name": "hi-en", "subset": null, "prompt_id": "b5952cac-9388-4901-98ed-c45cccfed5de", "prompt_jinja": "If the original version says: {{translation[\"en\"]}}; then the Hindi version should say:\n||| {{translation[\"hi\"]}}", "prompt_original_task": true, "comment": "", "bleu_stderr": 0.029122685049572238}], "config": {"model": "hf-causal", "model_args": "pretrained=/gpfsscratch/rech/six/commun/commun/experiments/muennighoff/bloomckpt/6b3t0/p31lossseqglobal_step1000,use_accelerate=True,tokenizer=/gpfsscratch/rech/six/commun/commun/experiments/muennighoff/bloomckpt/6b3t0/p31lossseqglobal_step1000,dtype=float16", "num_fewshot": 0, "batch_size": 16, "device": "cuda", "use_cache": false, "limit": 3000, "bootstrap_iters": 10, "seed": 1234}}