Llama2-13b-nemo

Put a short model description here.

See the model architecture section and NeMo documentation for complete architecture details.

NVIDIA NeMo: Training

To train, fine-tune or play with the model you will need to install NVIDIA NeMo. We recommend you install it after you've installed latest Pytorch version.

pip install nemo_toolkit['all']

How to Use this Model

The model is available for use in the NeMo toolkit [1], and can be used as a pre-trained checkpoint for inference or for fine-tuning on another dataset.

Automatically instantiate the model

NOTE: Please update the model class below to match the class of the model being uploaded.

import nemo.core import ModelPT
model = ModelPT.from_pretrained("pe-nlp/llama2-13b-nemo")

NOTE

Add some information about how to use the model here. An example is provided for ASR inference below.

### Transcribing using Python
First, let's get a sample
```
wget https://dldata-public.s3.us-east-2.amazonaws.com/2086-149220-0033.wav
```
Then simply do:
```
asr_model.transcribe(['2086-149220-0033.wav'])
```

### Transcribing many audio files

```shell
python [NEMO_GIT_FOLDER]/examples/asr/transcribe_speech.py      pretrained_name="pe-nlp/llama2-13b-nemo"      audio_dir=""
```

Input Models input text only.

Output Models generate text only.

Model Architecture Llama 2 is an auto-regressive language model that uses an optimized transformer architecture. The tuned versions use supervised fine-tuning (SFT) and reinforcement learning with human feedback (RLHF) to align to human preferences for helpfulness and safety.

	Training Data	Params	Content Length	GQA	Tokens	LR
Llama 2	A new mix of publicly available online data	7B	4k	✗	2.0T	3.0 x 10-4
Llama 2	A new mix of publicly available online data	13B	4k	✗	2.0T	3.0 x 10-4
Llama 2	A new mix of publicly available online data	70B	4k	✔	2.0T	1.5 x 10-4

Llama 2 family of models. Token counts refer to pretraining data only. All models are trained with a global batch-size of 4M tokens. Bigger models - 70B -- use Grouped-Query Attention (GQA) for improved inference scalability.

Training

Training Factors We used custom training libraries, Meta's Research Super Cluster, and production clusters for pretraining. Fine-tuning, annotation, and evaluation were also performed on third-party cloud compute.

Carbon Footprint Pretraining utilized a cumulative 3.3M GPU hours of computation on hardware of type A100-80GB (TDP of 350-400W). Estimated total emissions were 539 tCO2eq, 100% of which were offset by Meta’s sustainability program.

	Time (GPU hours)	Power Consumption (W)	Carbon Emitted(tCO2eq)
Llama 2 7B	184320	400	31.22
Llama 2 13B	368640	400	62.44
Llama 2 70B	1720320	400	291.42
Total	3311616		539.00

CO₂ emissions during pretraining. Time: total GPU time required for training each model. Power Consumption: peak power capacity per GPU device for the GPUs used adjusted for power usage efficiency. 100% of the emissions are directly offset by Meta's sustainability program, and because we are openly releasing these models, the pretraining costs do not need to be incurred by others.

NOTE

An example is provided below for ASR

The NeMo toolkit [1] was used for training the models for over several hundred epochs. These model are trained with this [example script](https://github.com/NVIDIA/NeMo/blob/main/examples/asr/asr_transducer/speech_to_text_rnnt_bpe.py) and this [base config](https://github.com/NVIDIA/NeMo/blob/main/examples/asr/conf/fastconformer/fast-conformer_transducer_bpe.yaml).

The tokenizers for these models were built using the text transcripts of the train set with this [script](https://github.com/NVIDIA/NeMo/blob/main/scripts/tokenizers/process_asr_text_tokenizer.py).

Datasets

Overview Llama 2 was pretrained on 2 trillion tokens of data from publicly available sources. The fine-tuning data includes publicly available instruction datasets, as well as over one million new human-annotated examples. Neither the pretraining nor the fine-tuning datasets include Meta user data.

Data Freshness The pretraining data has a cutoff of September 2022, but some tuning data is more recent, up to July 2023.

NOTE

An example for the manifest section is provided below for ASR datasets

datasets:
- librispeech_asr
- fisher_corpus
- Switchboard-1
- WSJ-0
- WSJ-1
- National-Singapore-Corpus-Part-1
- National-Singapore-Corpus-Part-6
- vctk
- voxpopuli
- europarl
- multilingual_librispeech
- mozilla-foundation/common_voice_8_0
- MLCommons/peoples_speech

The corresponding text in this section for those datasets is stated below -

The model was trained on 64K hours of English speech collected and prepared by NVIDIA NeMo and Suno teams.

The training dataset consists of private subset with 40K hours of English speech plus 24K hours from the following public datasets:

- Librispeech 960 hours of English speech
- Fisher Corpus
- Switchboard-1 Dataset
- WSJ-0 and WSJ-1
- National Speech Corpus (Part 1, Part 6)
- VCTK
- VoxPopuli (EN)
- Europarl-ASR (EN)
- Multilingual Librispeech (MLS EN) - 2,000 hour subset
- Mozilla Common Voice (v7.0)
- People's Speech  - 12,000 hour subset

Performance

In this section, we report the results for the Llama 1 and Llama 2 models on standard academic benchmarks.For all the evaluations, we use our internal evaluations library.

Model	Size	Code	Commonsense Reasoning	World Knowledge	Reading Comprehension	Math	MMLU	BBH	AGI Eval
Llama 1	7B	14.1	60.8	46.2	58.5	6.95	35.1	30.3	23.9
Llama 1	13B	18.9	66.1	52.6	62.3	10.9	46.9	37.0	33.9
Llama 1	33B	26.0	70.0	58.4	67.6	21.4	57.8	39.8	41.7
Llama 1	65B	30.7	70.7	60.5	68.6	30.8	63.4	43.5	47.6
Llama 2	7B	16.8	63.9	48.9	61.3	14.6	45.3	32.6	29.3
Llama 2	13B	24.5	66.9	55.4	65.8	28.7	54.8	39.4	39.1
Llama 2	70B	37.5	71.9	63.6	69.4	35.2	68.9	51.2	54.2

Overall performance on grouped academic benchmarks. Code: We report the average pass@1 scores of our models on HumanEval and MBPP. Commonsense Reasoning: We report the average of PIQA, SIQA, HellaSwag, WinoGrande, ARC easy and challenge, OpenBookQA, and CommonsenseQA. We report 7-shot results for CommonSenseQA and 0-shot results for all other benchmarks. World Knowledge: We evaluate the 5-shot performance on NaturalQuestions and TriviaQA and report the average. Reading Comprehension: For reading comprehension, we report the 0-shot average on SQuAD, QuAC, and BoolQ. MATH: We report the average of the GSM8K (8 shot) and MATH (4 shot) benchmarks at top 1.

		TruthfulQA	Toxigen
Llama 1	7B	27.42	23.00
Llama 1	13B	41.74	23.08
Llama 1	33B	44.19	22.57
Llama 1	65B	48.71	21.77
Llama 2	7B	33.29	21.25
Llama 2	13B	41.86	26.10
Llama 2	70B	50.18	24.60

Evaluation of pretrained LLMs on automatic safety benchmarks. For TruthfulQA, we present the percentage of generations that are both truthful and informative (the higher the better). For ToxiGen, we present the percentage of toxic generations (the smaller the better).

		TruthfulQA	Toxigen
Llama-2-Chat	7B	57.04	0.00
Llama-2-Chat	13B	62.18	0.00
Llama-2-Chat	70B	64.14	0.01

Evaluation of fine-tuned LLMs on different safety datasets. Same metric definitions as above.

NOTE

An example is provided below for ASR metrics list that can be added to the top of the README

model-index:
- name: PUT_MODEL_NAME
  results:
  - task:
      name: Automatic Speech Recognition
      type: automatic-speech-recognition
    dataset:
      name: AMI (Meetings test)
      type: edinburghcstr/ami
      config: ihm
      split: test
      args:
        language: en
    metrics:
    - name: Test WER
      type: wer
      value: 17.10
  - task:
      name: Automatic Speech Recognition
      type: automatic-speech-recognition
    dataset:
      name: Earnings-22
      type: revdotcom/earnings22
      split: test
      args:
        language: en
    metrics:
    - name: Test WER
      type: wer
      value: 14.11

Provide any caveats about the results presented in the top of the discussion so that nuance is not lost.

It should ideally be in a tabular format (you can use the following website to make your tables in markdown format - https://www.tablesgenerator.com/markdown_tables)**

Limitations

Llama 2 is a new technology that carries risks with use. Testing conducted to date has been in English, and has not covered, nor could it cover all scenarios. For these reasons, as with all LLMs, Llama 2’s potential outputs cannot be predicted in advance, and the model may in some instances produce inaccurate, biased or other objectionable responses to user prompts. Therefore, before deploying any applications of Llama 2, developers should perform safety testing and tuning tailored to their specific applications of the model.

Please see the Responsible Use Guide available at https://ai.meta.com/llama/responsible-use-guide/

Note

An example is provided below 

Since this model was trained on publicly available speech datasets, the performance of this model might degrade for speech which includes technical terms, or vernacular that the model has not been trained on. The model might also perform worse for accented speech.

License

License to use this model is covered by the CC-BY-4.0. By downloading the public and release version of the model, you accept the terms and conditions of the CC-BY-4.0 license.

References

Provide appropriate references in the markdown link format below. Please order them numerically.

[1] NVIDIA NeMo Toolkit