metadata

license: mit
base_model: HuggingFaceH4/zephyr-7b-gemma-sft
tags:
  - alignment-handbook
  - trl
  - dpo
  - generated_from_trainer
datasets:
  - argilla/dpo-mix-7k
pipeline_tag: text-generation
model-index:
  - name: zephyr-7b-gemma
    results:
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: MT-Bench
          type: unknown
        metrics:
          - type: unknown
            name: score
            value: 7.81
        source:
          url: https://huggingface.co/spaces/lmsys/mt-bench

Model Card for Zephyr 7B Gemma

Zephyr is a series of language models that are trained to act as helpful assistants. Zephyr 7B Gemma is the third model in the series, and is a fine-tuned version of google/gemma-7b that was trained on on a mix of publicly available, synthetic datasets using Direct Preference Optimization (DPO). You can reproduce the training of this model via the recipe provided in the Alignment Handbook.

Model description

Model type: A 7B parameter GPT-like model fine-tuned on a mix of publicly available, synthetic datasets.
Language(s) (NLP): Primarily English
License: MIT
Finetuned from model: google/gemma-7b

Model Sources

Repository: https://github.com/huggingface/alignment-handbook
Demo: https://huggingface.co/spaces/HuggingFaceH4/zephyr-chat

Performance

Model	MT Bench	IFEval
zephyr-7b-gemma	7.81	28.76
zephyr-7b-beta	7.34	43.81
gemma-7b-it	6.38	38.01

Model	AGIEval	GPT4All	TruthfulQA	BigBench	Average
zephyr-7b-gemma	34.22	66.37	52.19	37.10	47.47
zephyr-7b-beta	37.52	71.77	55.26	39.77	51.08
gemma-7b-it	21.33	40.84	41.70	30.25	33.53

Intended uses & limitations

The model was initially fine-tuned on the DEITA 10K dataset, which contains a diverse range of synthetic dialogues generated by ChatGPT. We then further aligned the model with 🤗 TRL's DPOTrainer on the argilla/dpo-mix-7k dataset, which contains 7k prompts and model completions that are ranked by GPT-4. As a result, the model can be used for chat and you can check out our demo to test its capabilities.

Here's how you can run the model using the pipeline() function from 🤗 Transformers:

# Install transformers from source - only needed for versions <= v4.38.1
# pip install git+https://github.com/huggingface/transformers.git
# pip install accelerate

import torch
from transformers import pipeline

pipe = pipeline(
    "text-generation",
    model="HuggingFaceH4/zephyr-7b-gemma",
    device_map="auto",
    torch_dtype=torch.bfloat16,
)
messages = [
    {
        "role": "system",
        "content": "",  # Model not yet trained for follow this
    },
    {"role": "user", "content": "How many helicopters can a human eat in one sitting?"},
]
outputs = pipe(
    messages,
    max_new_tokens=128,
    do_sample=True,
    temperature=0.7,
    top_k=50,
    top_p=0.95,
    stop_sequence="<|im_end|>",
)
print(outputs[0]["generated_text"][-1]["content"])
# It is not possible for a human to eat a helicopter in one sitting, as a
# helicopter is a large and inedible machine. Helicopters are made of metal,
# plastic, and other materials that are not meant to be consumed by humans.
# Eating a helicopter would be extremely dangerous and would likely cause
# serious health problems, including choking, suffocation, and poisoning. It is
# important to only eat food that is safe and intended for human consumption.

Bias, Risks, and Limitations

Zephyr 7B Gemma has not been aligned to human preferences for safety within the RLHF phase or deployed with in-the-loop filtering of responses like ChatGPT, so the model can produce problematic outputs (especially when prompted to do so). It is also unknown what the size and composition of the corpus was used to train the base model (google/gemma-7b), however it is likely to have included a mix of Web data and technical sources like books and code. See the StarCoder2 model card for an example of this.

Training and evaluation data

This model is a fine-tuned version of HuggingFaceH4/zephyr-7b-gemma-sft on the argilla/dpo-mix-7k dataset.

It achieves the following results on the evaluation set:

Loss: 0.4695
Rewards/chosen: -3.3746
Rewards/rejected: -4.9715
Rewards/accuracies: 0.7188
Rewards/margins: 1.5970
Logps/rejected: -459.4853
Logps/chosen: -429.9115
Logits/rejected: 86.4684
Logits/chosen: 92.8200

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 5e-07
train_batch_size: 2
eval_batch_size: 4
seed: 42
distributed_type: multi-GPU
num_devices: 8
gradient_accumulation_steps: 8
total_train_batch_size: 128
total_eval_batch_size: 32
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: cosine
lr_scheduler_warmup_ratio: 0.1
num_epochs: 2

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.1923	1.9	100	0.4736	-3.4575	-4.9556	0.75	1.4980	-459.1662	-431.5707	86.3863	92.7360

Framework versions

Transformers 4.39.0.dev0
Pytorch 2.1.2+cu121
Datasets 2.14.6
Tokenizers 0.15.1