Update README.md

9b711ed about 1 year ago

No virus

6.8 kB

	---
	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]