apple/DCLM-7B

Model Card for DCLM-Baseline-7B

DCLM-Baseline-7B is a 7 billion parameter language model trained on the DCLM-Baseline dataset, which was curated as part of the DataComp for Language Models (DCLM) benchmark. This model is designed to showcase the effectiveness of systematic data curation techniques for improving language model performance.

Model Details

Size	Training Tokens	Layers	Hidden Size	Attention Heads	Context Length
7B	2.5T	32	4096	32	2048

Model Description

• Developed by: DataComp for Language Models (DCLM) Team
• Model type: Decoder-only Transformer language model
• Language(s): English (primarily)
• License: Apple Sample Code License
• Contact: [email protected]
• Date: June 2024

Model Sources

• Repository: https://github.com/mlfoundations/dclm
• Dataset: https://huggingface.co/datasets/mlfoundations/dclm-baseline-1.0
• Paper: DataComp-LM: In search of the next generation of training sets for language models

Using Model

First install open_lm

pip install git+https://github.com/mlfoundations/open_lm.git

Then:

from open_lm.hf import *
from transformers import AutoTokenizer, AutoModelForCausalLM
tokenizer = AutoTokenizer.from_pretrained("apple/DCLM-Baseline-7B")
model = AutoModelForCausalLM.from_pretrained("apple/DCLM-Baseline-7B")

inputs = tokenizer(["Machine learning is"], return_tensors="pt")
gen_kwargs = {"max_new_tokens": 50, "top_p": 0.8, "temperature": 0.8, "do_sample": True, "repetition_penalty": 1.1}
output = model.generate(inputs['input_ids'], **gen_kwargs)
output = tokenizer.decode(output[0].tolist(), skip_special_tokens=True)
print(output)

Training Details

The model was trained using the following setup:

• Architecture: Decoder-only Transformer
• Framework: PyTorch with OpenLM
• Optimizer: AdamW
• Learning Rate: 2e-3 (peak)
• Weight Decay: 0.05
• Batch Size: 2048 sequences
• Sequence Length: 2048 tokens
• Total Training Tokens: 2.5T
• Hardware: Trained on H100 GPUs

For more detailed training information, please refer to Section 3.4 and Appendix F of the DCLM paper. To ensure our trained model is broadly useful, including for math and coding tasks, we combine our 3.8T DCLM-BASELINE with the StarCoder and ProofPile2 data to arrive at a 4.1T token dataset.

Evaluation

Here are the evaluation results for DCLM-Baseline-7B on various tasks (using llm-foundry eval suite)

Task	Score
MMLU (zero-shot)	0.5766
MMLU (few-shot)	0.6372
HellaSwag (zero-shot)	0.7987
HellaSwag	0.8043
Jeopardy	0.4745
TriviaQA	0.5270
GSM8K (CoT)	0.0250
AGI Eval SAT Math (CoT)	0.0136
AQuA (CoT)	0.0490
SVAMP (CoT)	0.4900
BigBench QA Wikidata	0.7120
ARC Easy	0.8220
ARC Challenge	0.5990
BigBench Misconceptions	0.6986
COPA	0.8500
SIQA	0.8291
CommonsenseQA	0.8018
PIQA	0.8128
OpenBookQA	0.4540
BigBench Novel Concepts	0.7188
BigBench Strange Stories	0.7586
BigBench Strategy QA	0.6173
LAMBADA	0.8220
Winograd	0.8828
Winogrande	0.7269
BigBench Conlang Translation	0.0244
BigBench Language Identification	0.5219
BigBench Conceptual Combinations	0.6990
BigBench Elementary Math QA	0.3431
BigBench Dyck Languages	0.4930
AGI Eval LSAT AR	0.2435
BigBench CS Algorithms	0.6121
BigBench Logical Deduction	0.3620
BigBench Operators	0.4857
BigBench Repeat Copy Logic	0.4063
Simple Arithmetic (no spaces)	0.2940
Simple Arithmetic (with spaces)	0.3110
MathQA	0.3098
LogiQA	0.4132
PubMedQA	0.7060
SQuAD	0.5856
AGI Eval LSAT RC	0.6716
AGI Eval LSAT LR	0.5392
CoQA	0.4074
BigBench Understanding Fables	0.6825
BoolQ	0.8343
AGI Eval SAT EN	0.7670
Winogender MC (Female)	0.6000
Winogender MC (Male)	0.5500
Enterprise PII Classification	0.7676
BBQ	0.6912
GPQA Main	0.2612
GPQA Diamond	0.2475

Note: All scores are presented as decimal values between 0 and 1, representing the proportion of correct answers or the model's performance on each task.

Comparison

Below are comparisions of this model with other models in the 7B regime.

Model	Params	Tokens	Open dataset?	CORE	MMLU	EXTENDED
Open weights, closed datasets
Llama2	7B	2T	❌	49.2	45.8	34.1
DeepSeek	7B	2T	❌	50.7	48.5	35.3
Mistral-0.3	7B	?	❌	57.0	62.7	45.1
QWEN-2	7B	?	❌	57.5	71.9	50.5
Llama3	8B	15T	❌	57.6	66.2	46.3
Gemma	8B	6T	❌	57.8	64.3	44.6
Phi-3	7B	?	❌	61.0	69.9	57.9
Open weights, open datasets
Falcon	7B	1T	✅	44.1	27.4	25.1
OLMo-1.7	7B	2.1T	✅	47.0	54.0	34.2
MAP-Neo	7B	4.5T	✅	50.2	57.1	40.4
DCLM-7B	7B	2.5T	✅	56.1	63.7	43.6

Limitations and Biases

While DCLM-Baseline-7B demonstrates strong performance across a range of tasks, it's important to note:

1. The model may exhibit biases present in its training data, which is derived from web crawl data.
2. It has not undergone specific alignment or safety fine-tuning, so outputs should be used with caution.
3. Performance on tasks not included in the evaluation suite may vary.
4. The model's knowledge is limited to its training data cutoff date.

Ethical Considerations

Users should be aware that this model, like all large language models, can potentially generate harmful or biased content. It should not be used for making decisions about individuals or in sensitive applications without appropriate safeguards and human oversight.

Citation

If you use this model in your research, please cite:

@article{Li2024DataCompLM,
  title={DataComp-LM: In search of the next generation of training sets for language models},
  author={Jeffrey Li and Alex Fang and Georgios Smyrnis and Maor Ivgi and Matt Jordan and Samir Gadre and Hritik Bansal and Etash Guha and Sedrick Keh and Kushal Arora and [... full author list]},
  journal={arXiv preprint arXiv:2406.11794},
  year={2024}
}