metadata
dataset_info:
features:
- name: input_ids_1
sequence: int64
- name: attention_mask_1
sequence: int64
- name: labels_1
sequence: int64
- name: input_ids_2
sequence: int64
- name: attention_mask_2
sequence: int64
- name: labels_2
sequence: int64
splits:
- name: train
num_bytes: 145200
num_examples: 50
download_size: 25156
dataset_size: 145200
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
From Differential Adaptive Merging: Created via:
!python dam/data/create_merge_dataset.py \
--dataset_names "tatsu-lab/alpaca,microsoft/orca-math-word-problems-200k" \
--model_ids "meta-llama/Llama-3.2-1B,meta-llama/Llama-3.2-1B-Instruct" \
--base_model_name "meta-llama/Llama-3.2-1B" \
--cache_dir "/content/cache" \
--compute_logits False \
--dataset_id "your_dataset_id" \
--example_count 50 \
--max_length 60 \
--add_top_k_logits False