metadata
dataset_info:
features:
- name: description
dtype: string
- name: source_repo
dtype: string
- name: task_type
dtype: string
- name: seed
dtype: int64
- name: needle_template
dtype: string
- name: question
dtype: string
- name: scoring
dtype: string
- name: tokenizer
dtype: string
- name: haystack_source
dtype: string
- name: prompt_format
dtype: string
- name: grid
struct:
- name: context_length
dtype: int64
- name: depth_percent
dtype: float64
- name: seed
dtype: int64
- name: needle
dtype: string
- name: question
dtype: string
- name: expected_answer
dtype: string
- name: actual_depth_percent
dtype: float64
- name: insertion_token_index
dtype: int64
- name: final_context_token_count
dtype: int64
- name: full_prompt_token_count
dtype: int64
- name: full_prompt_char_count
dtype: int64
- name: full_prompt
dtype: string
splits:
- name: train
num_bytes: 534558430
num_examples: 1225
download_size: 534154626
dataset_size: 534558430
configs:
- config_name: default
data_files:
- split: train
path: data/train-*