README.md

---
license: apache-2.0
language:
- en
tags:
- sft
pipeline_tag: text-generation
widget:
  - text: <|startoftoken|>system\nYou are a helpful assistant<|endoftoken|><startoftoken>human\nWhat's the population of the earth?<|endoftoken|><startoftoken>assistant
---

# Pythia 1.4B SFT model revision 1

<!-- Provide a quick summary of what the model is/does. -->


# Model Details

## Model Description

Model was supervised fine tuned on only [Open Assistant](https://open-assistant.io/) crowd souce platform.

- **Developed by:** Open Assistant
- **Model type:** Pythia
- **Language(s) (NLP):** English
- **License:** Apache-2.0

## Model Sources [optional]

<!-- Provide the basic links for the model. -->

- **Repository:** [Open Assistant](https://github.com/LAION-AI/Open-Assistant)

# Uses

<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->


## Direct Use

<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
See the example on the right

# Bias, Risks, and Limitations

<!-- This section is meant to convey both technical and sociotechnical limitations. -->

[just read pythia](https://huggingface.co/EleutherAI/pythia-12b#out-of-scope-use)

## Recommendations

<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->

Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.

## How to Get Started with the Model

Use the code below to get started with the model.

```python
from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "theblackcat102/pythia-1.4b-deduped-sft-r2"

tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name).half().eval().cuda()

input_text = """
<|startoftoken|>system
You are a helpful assistant<|endoftoken|><|startoftoken|>human
What's the population of the earth?<|endoftoken|><|startoftoken|>assistant

"""
inputs = tokenizer(input_text, return_tensors="pt", padding=True).to(0)
outputs = model.generate(
    **inputs,
    early_stopping=True,
    max_new_tokens=args.max_new_tokens,
    do_sample=True,
    top_k=args.top_k,
    temperature=args.temperature,
    pad_token_id=tokenizer.eos_token_id,
    # dialogue_collator.py line 36
)
output = tokenizer.decode(outputs[0], truncate_before_pattern=[r"\n\n^#", "^'''", "\n\n\n"])
print(output)
```

# Training Details

## Training Data

<!-- This should link to a Data Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->

## Training Procedure 

```
deepspeed trainer_sft.py --configs defaults pythia-1-4b-ost --deepspeed
```

This model was trained for 200 iterations. After 200 iterations the accuracy started to drop and loss increasing which is a sign of overfitting.

### Training Hyperparameters

```
defaults:
  learning_rate: 1e-5
  gradient_checkpointing: false
  gradient_accumulation_steps: 32
  per_device_train_batch_size: 2
  per_device_eval_batch_size: 2
  weight_decay: 0.00
  warmup_steps: 600
  eval_steps: 250
  save_steps: 250
  max_length: 512
  num_train_epochs: 2
  logging_steps: 10
  max_grad_norm: 2.0
  save_total_limit: 4
  fp16: true
  eval_accumulation_steps:
  freeze_layer:
  datasets:
    - oa_private:
        data_path: .cache
        split: sft
        val_split: 0.01
        fraction: 1
        file: 2023-02-26_oasst_default.jsonl
  cache_dir: .cache
  loss_fn: CrossEntropyLoss
  eval_size:
  log_dir: "base"
  quantization: false
  seq2seqmodel: false
  poly_eps: 1.0
  fuse_gelu: false
  log_wandb: true
  samples_mixing: true # uses collator that mixes samples in the batch to create a single sample with possible multiple tasks within
  verbose: false


pythia-1-4b-ost:
  learning_rate: 1e-6
  model_name: EleutherAI/pythia-1.4b-deduped
  weight_decay: 0.01
  max_length: 1024
  warmup_steps: 100
  gradient_checkpointing: false
  gradient_accumulation_steps: 12
  per_device_train_batch_size: 5
  per_device_eval_batch_size: 6
  eval_steps: 100
  save_steps: 100
  num_train_epochs: 50
  save_total_limit: 4
```


# Evaluation

<!-- This section describes the evaluation protocols and provides the results. -->

## Testing Data, Factors & Metrics

### Testing Data

<!-- This should link to a Data Card if possible. -->

[More Information Needed]

### Factors

<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->

[More Information Needed]

### Metrics

<!-- These are the evaluation metrics being used, ideally with a description of why. -->

[More Information Needed]

## Results



### Summary



# Model Examination [optional]

<!-- Relevant interpretability work for the model goes here -->

[More Information Needed]

# Environmental Impact

<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->

Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).

- **Hardware Type:** [More Information Needed]
- **Hours used:** [More Information Needed]
- **Cloud Provider:** [More Information Needed]
- **Compute Region:** [More Information Needed]
- **Carbon Emitted:** [More Information Needed]

# Technical Specifications [optional]

## Model Architecture and Objective

[More Information Needed]

## Compute Infrastructure

[More Information Needed]

### Hardware

[More Information Needed]

### Software

[More Information Needed]

# Citation [optional]

<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->

**BibTeX:**

[More Information Needed]

**APA:**

[More Information Needed]

# Glossary [optional]

<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->

[More Information Needed]

# Acknowledgements

- [LAION](https://laion.ai/) & EleutherAI
- [Stability.ai](https://stability.ai/) : this project wouldn't be possible without their compute resource
- [Teams and contributors at Open Assistant](https://github.com/LAION-AI/Open-Assistant/graphs/contributors) : who put their time after their day job or whatever into this project
- [Huggingface](https://huggingface.co/) : For the storage and spaces here

# Model Card Authors [optional]

[More Information Needed]

# Model Card Contact

[More Information Needed]