|
--- |
|
language: |
|
- en |
|
license: llama3 |
|
library_name: transformers |
|
datasets: |
|
- prince-canuma/fineweb-CC-MAIN-2024-10-1B-en |
|
- HuggingFaceFW/fineweb |
|
tags: |
|
- Llama-3-6B |
|
- 6B |
|
base_model: |
|
- prince-canuma/Llama-3-6B-v0 |
|
--- |
|
**Exllamav2** quant (**exl2** / **6.0 bpw**) made with ExLlamaV2 v0.0.21 |
|
|
|
Other EXL2 quants: |
|
| **Quant** | **Model Size** | **lm_head** | |
|
| ----- | ---------- | ------- | |
|
|<center>**[2.2](https://huggingface.co/Zoyd/prince-canuma_Llama-3-6B-v0.1-2_2bpw_exl2)**</center> | <center>2787 MB</center> | <center>6</center> | |
|
|<center>**[2.5](https://huggingface.co/Zoyd/prince-canuma_Llama-3-6B-v0.1-2_5bpw_exl2)**</center> | <center>2959 MB</center> | <center>6</center> | |
|
|<center>**[3.0](https://huggingface.co/Zoyd/prince-canuma_Llama-3-6B-v0.1-3_0bpw_exl2)**</center> | <center>3259 MB</center> | <center>6</center> | |
|
|<center>**[3.5](https://huggingface.co/Zoyd/prince-canuma_Llama-3-6B-v0.1-3_5bpw_exl2)**</center> | <center>3583 MB</center> | <center>6</center> | |
|
|<center>**[3.75](https://huggingface.co/Zoyd/prince-canuma_Llama-3-6B-v0.1-3_75bpw_exl2)**</center> | <center>3739 MB</center> | <center>6</center> | |
|
|<center>**[4.0](https://huggingface.co/Zoyd/prince-canuma_Llama-3-6B-v0.1-4_0bpw_exl2)**</center> | <center>3895 MB</center> | <center>6</center> | |
|
|<center>**[4.25](https://huggingface.co/Zoyd/prince-canuma_Llama-3-6B-v0.1-4_25bpw_exl2)**</center> | <center>4051 MB</center> | <center>6</center> | |
|
|<center>**[5.0](https://huggingface.co/Zoyd/prince-canuma_Llama-3-6B-v0.1-5_0bpw_exl2)**</center> | <center>4519 MB</center> | <center>6</center> | |
|
|<center>**[6.0](https://huggingface.co/Zoyd/prince-canuma_Llama-3-6B-v0.1-6_0bpw_exl2)**</center> | <center>5247 MB</center> | <center>8</center> | |
|
|<center>**[6.5](https://huggingface.co/Zoyd/prince-canuma_Llama-3-6B-v0.1-6_5bpw_exl2)**</center> | <center>5548 MB</center> | <center>8</center> | |
|
|<center>**[8.0](https://huggingface.co/Zoyd/prince-canuma_Llama-3-6B-v0.1-8_0bpw_exl2)**</center> | <center>6436 MB</center> | <center>8</center> | |
|
|
|
|
|
# Model Summary |
|
<img src="images/llama-3-6B icon.jpeg" width="500" alt="Llama-3-6B"/> |
|
|
|
Introducing the world's first Llama-3 base model with 6B parameters. This model is a pretrained version of [prince-canuma/Llama-3-6B-v0](https://huggingface.co/prince-canuma/Llama-3-6B-v0), which was created from Meta-Llama-3-8B using a technique called [downcycling](https://youtube.com/playlist?list=PLDn_JsyofyfTH5_5V1MNb8UYKxMl6IMNy&si=9hcOol4KHIgWThgt) . |
|
The model was continually pretrained on 1 billion tokens of English-only text from fineweb, achieving impressive results on the evaluation set: |
|
- Loss: 2.4942 |
|
|
|
<!-- Provide a longer summary of what this model is. --> |
|
|
|
## Model Description |
|
|
|
<!-- Provide a longer summary of what this model is. --> |
|
This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated. |
|
|
|
- **Developed by:** [Prince Canuma](https://huggingface.co/prince-canuma) |
|
- **Sponsored by:** General |
|
- **Model type:** Llama |
|
- **License:** [Llama-3](https://llama.meta.com/llama3/license) |
|
- **Pretrained from model:** prince-canuma/Llama-3-6B-v0 |
|
|
|
### Model Sources |
|
|
|
<!-- Provide the basic links for the model. --> |
|
|
|
- **Repository:** https://github.com/Blaizzy/Coding-LLMs-from-scratch/tree/main/Llama-3 |
|
- **Video:** https://youtube.com/playlist?list=PLDn_JsyofyfTH5_5V1MNb8UYKxMl6IMNy&si=5Y4cm-6wrMOD1Abr |
|
|
|
## 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. --> |
|
You can use this model to create instruct and chat versions for various use cases such as: Coding assistant, RAG, Function Calling and more. |
|
|
|
### Limitations |
|
|
|
This model inherits some of the base model's limitations and some additional ones from it's creation process, such as: |
|
- Limited scope for coding and math: According to benchmarks, this model needs more pretraining/finetuning on code and math data to excel at reasoning tasks. |
|
- Language Limitations: This model was continually pretrained on english only data. If you are planning to use it for multilingual use cases I recommend fine-tuning or continued pretraining. |
|
|
|
## How to Get Started with the Model |
|
|
|
Use the code below to get started with the model. |
|
|
|
```python |
|
from transformers import AutoModelForCausalLM, AutoConfig, AutoTokenizer |
|
|
|
# Load model, config and tokenizer |
|
model_name = "prince-canuma/Llama-3-6B-v0.1" |
|
model = AutoModelForCausalLM.from_pretrained(model_name) |
|
tokenizer = AutoTokenizer.from_pretrained(model_name) |
|
|
|
inputs = tokenizer( |
|
[ |
|
"Who created Python?" |
|
], return_tensors = "pt") |
|
|
|
from transformers import TextStreamer |
|
text_streamer = TextStreamer(tokenizer) |
|
_ = model.generate(**inputs, streamer = text_streamer, max_new_tokens = 200) |
|
|
|
``` |
|
|
|
Output: |
|
```shell |
|
<|begin_of_text|>Who created Python? What is Python used for? What is the difference between Python 2 and Python 3? What is the difference between Python and Python 3? |
|
Python is a programming language that was created by Guido van Rossum in 1991. It is a widely used language for web development, data science, and machine learning. Python is also used for creating software applications and games. |
|
Python is a powerful language that is easy to learn and use. It has a large library of built-in functions and packages that make it easy to write code. Python is also a very popular language for web development, with many popular web frameworks such as Django and Flask being written in Python. |
|
Python is also used for data science and machine learning. It has a large library of packages for data analysis, machine learning, and artificial intelligence. Python is also used for creating software applications and games. |
|
Python 2 and Python 3 are two different versions of the Python language. Python 2 was the original version of the |
|
``` |
|
|
|
|
|
## Training Details |
|
|
|
### Downcycling |
|
|
|
<img src="images/downcycling.jpeg" width="500" alt="Llama-3-8B-vs-6B-v0"/> |
|
Fig 1. Downcycling workflow as also described in [arxiv.org/abs/2404.08634](https://arxiv.org/abs/2404.08634). |
|
|
|
A technique that allows you to create new LLMs of diversa sizes from checkpoints of large pretrained models. |
|
You take a reference model (i.e., Llama-3-8B) and copy the weights of 24 layers out of 32 layers alongside embedding and prediction heads. |
|
Then you initialize a smaller target model with 24 layers and load those pretrained weights. |
|
|
|
This new model will most likely still output legible outputs, but for it to perform well you need continue the pretraining. |
|
|
|
<img src="images/Llama-3-8B-vs-6B-v0.png" width="500" alt="Llama-3-8B-vs-6B-v0"/> |
|
Fig 2. Downcycled model vs Reference model, without continued pretraining. |
|
|
|
### Training Data |
|
|
|
For continued pretrained, I extracted 1B tokens from [Huggingface's FineWeb CC-Main-2024-10](https://huggingface.co/datasets/HuggingFaceFW/fineweb#breakdown-by-dumpcrawl) slice. |
|
|
|
|
|
#### Training hyperparameters |
|
|
|
The following hyperparameters were used during training: |
|
- learning_rate: 0.0002 |
|
- train_batch_size: 2 |
|
- eval_batch_size: 2 |
|
- seed: 42 |
|
- distributed_type: multi-GPU |
|
- num_devices: 4 |
|
- gradient_accumulation_steps: 8 |
|
- total_train_batch_size: 64 |
|
- total_eval_batch_size: 8 |
|
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 |
|
- lr_scheduler_type: cosine |
|
- lr_scheduler_warmup_steps: 100 |
|
- num_epochs: 2 |
|
|
|
[<img src="https://raw.githubusercontent.com/OpenAccess-AI-Collective/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>](https://github.com/OpenAccess-AI-Collective/axolotl) |
|
<details><summary>See axolotl config</summary> |
|
|
|
axolotl version: `0.4.0` |
|
```yaml |
|
base_model: prince-canuma/Llama-3-6B-v0.1 |
|
model_type: AutoModelForCausalLM |
|
tokenizer_type: AutoTokenizer |
|
|
|
load_in_8bit: false |
|
load_in_4bit: true |
|
strict: false |
|
|
|
datasets: |
|
- path: prince-canuma/fineweb-CC-MAIN-2024-10-1B-en |
|
type: completion |
|
split: train |
|
dataset_prepared_path: last_run_prepared |
|
val_set_size: 0.001 |
|
output_dir: ./llama-3-6b |
|
save_safetensors: true |
|
adapter: qlora |
|
lora_model_dir: |
|
|
|
sequence_len: 8192 |
|
sample_packing: false |
|
pad_to_sequence_len: false |
|
|
|
lora_r: 128 |
|
lora_alpha: 128 |
|
lora_dropout: 0.05 |
|
lora_target_modules: |
|
lora_target_linear: true |
|
lora_fan_in_fan_out: |
|
|
|
|
|
wandb_project: llama-3-6b |
|
wandb_entity: |
|
wandb_watch: |
|
wandb_name: |
|
wandb_log_model: |
|
|
|
gradient_accumulation_steps: 8 |
|
micro_batch_size: 2 |
|
num_epochs: 2 |
|
optimizer: paged_adamw_32bit |
|
lr_scheduler: cosine |
|
learning_rate: 2e-4 |
|
|
|
train_on_inputs: false |
|
group_by_length: false |
|
bf16: auto |
|
fp16: |
|
tf32: false |
|
|
|
gradient_checkpointing: true |
|
early_stopping_patience: |
|
resume_from_checkpoint: |
|
local_rank: |
|
logging_steps: 1 |
|
xformers_attention: |
|
flash_attention: true |
|
|
|
warmup_steps: 100 |
|
evals_per_epoch: 4 |
|
eval_table_size: |
|
save_steps: 4000 |
|
debug: |
|
deepspeed: |
|
weight_decay: 0.0 |
|
fsdp: |
|
fsdp_config: |
|
special_tokens: |
|
pad_token: "<|reserved_special_token_0|>" |
|
|
|
|
|
``` |
|
|
|
</details><br> |
|
|
|
### Training results |
|
|
|
There were 3 distinct experiments. In these experiments, QLoRA was used instead of Full Fine-tuning due to budget constraints. |
|
- v0: This was a test ran for 1K steps to check if the model would improve with QLoRA params. |
|
- v1: Here the QLoRA parameters where tweaked (Rank and Alpha). |
|
- v2: This was the main experiment, ran for 2 epochs on 1B tokens from FineWeb. |
|
|
|
All details can be found on my Wandb dashboard: https://wandb.ai/prince-canuma/llama-3-6b?nw=nwuserprincecanuma |
|
|
|
<img src="images/Training Loss.png" width="500" alt="Llama-3-8B-vs-6B-v0"/> |
|
Fig 3. Experiment training loss charts on wandb. |
|
|
|
Overal metrics: |
|
|
|
| Training Loss | Epoch | Step | Validation Loss | |
|
|:-------------:|:-----:|:-----:|:---------------:| |
|
| 7.1562 | 0.0 | 1 | 7.1806 | |
|
| 2.7339 | 0.25 | 5867 | 2.6266 | |
|
| 2.6905 | 0.5 | 11734 | 2.5872 | |
|
| 2.6134 | 0.75 | 17601 | 2.5549 | |
|
| 2.532 | 1.0 | 23468 | 2.5235 | |
|
| 2.5319 | 1.25 | 29335 | 2.5067 | |
|
| 2.3336 | 1.5 | 35202 | 2.4968 | |
|
| 2.3486 | 1.75 | 41069 | 2.4942 | |
|
|
|
|
|
|
|
|
|
### Framework versions |
|
|
|
- PEFT 0.10.0 |
|
- Transformers 4.40.0.dev0 |
|
- Pytorch 2.2.0+cu121 |
|
- Datasets 2.15.0 |
|
- Tokenizers 0.15.0 |
|
|
|
### Hardware: |
|
|
|
- 4xRTX6000 using JarvisLabs (Sponsored by [General Catalyst](https://www.generalcatalyst.com/) thanks to Viet) |
|
|
|
|
|
## Evaluation |
|
|
|
<!-- This section describes the evaluation protocols and provides the results. --> |
|
|
|
#### Benchmarks |
|
|
|
<!-- These are the evaluation metrics being used, ideally with a description of why. --> |
|
|
|
- **Hellaswag**: a dataset for studying grounded commonsense inference. |
|
- **ARC**: a multiple-choice question-answering dataset. |
|
from science exams from grade 3 to grade 9. |
|
- **MMLU**: a test with 57 tasks to measure a text model's multitask accuracy. |
|
- **TruthfulQA**: a test to measure a model's propensity to reproduce falsehoods commonly found online. |
|
- **Winogrande**: for commonsense reasoning. |
|
- **GSM8k**: diverse grade school math word problems to measure a model's |
|
ability to solve multi-step mathematical reasoning problems. |
|
|
|
### Results |
|
|
|
<img src="images/comparison_model_scores_histogram.png" width="500" alt="Llama-3-8B-vs-6B-v0"/> |
|
Fig 4. Performance comparision of Llama-3-8B, Llama-3-6B and Llama-3-6B (w/ continued pretraining) |
|
|
|
Pretraining for 2 epochs on 1B tokens had a positive effect across the board. The new base model now performs competitively with its reference model (Llama-3-8B) whilst being 1.3x smaller. |
|
|
|
<img src="images/Comparision_of_Model_Scores.png" width="500" alt="All-vs-Llama-3-6B-v0"/> |
|
Fig 5. Performance comparision of Llama-3-8B, Llama-2-13B, Yi-1.5-6B and Llama-3-6B. |
|
|
|
Llama-3-6B is competive with model within it's category and upto 2x larger than it self across 6 diverse benchmarks. |
|
|
|
#### Summary and future directions: |
|
|
|
This experiment was a success! Using this technique, I'll be able to build many variants. This is the first of many new base models I intend to create. |
|
|
|
Next, I plan to explore different data mixtures and perform full fine-tuning, all of which will contribute to developing other small model as well as larger and more robust models. |
|
|
|
|
|
## Citation |
|
|
|
<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. --> |
|
|
|
### **BibTeX:** |
|
|
|
```bibtex |
|
@misc{prince2024downcycling, |
|
title={Efficient LLM Downcycling: Generating Diverse Model Sizes from Pretrained Giants}, |
|
author={Prince Canuma}, |
|
year={2024}, |
|
} |
|
``` |
|
|
|
# **Thank You!** |
|
|
|
I want to extend my heartfelt thanks to the community for the invaluable expertise and unwavering support. |
|
|
|
Additionally, I would like to thank Viet from General Catalyst (GC) for providing me with the much needed compute. |
|
|
|
This is my most ambitious project yet, and it wouldn't have been possible without the incredible open-source ML community! |
|
|
|
Developers, I am eager to see and hear about the innovative fine-tunes and applications you create. |
|
|
|
Users, I am excited to learn about your experiences and use cases. |
|
|
|
Thank you for your interest and support! |
|
|
|
## References: |
|
|
|
```bibtex |
|
@misc{komatsuzaki2023sparse, |
|
title={Sparse Upcycling: Training Mixture-of-Experts from Dense Checkpoints}, |
|
author={Aran Komatsuzaki and Joan Puigcerver and James Lee-Thorp and Carlos Riquelme Ruiz and Basil Mustafa and Joshua Ainslie and Yi Tay and Mostafa Dehghani and Neil Houlsby}, |
|
year={2023}, |
|
eprint={2212.05055}, |
|
archivePrefix={arXiv}, |
|
primaryClass={cs.LG} |
|
} |
|
``` |
|
|
|
```bibtex |
|
@misc{sanyal2024pretraining, |
|
title={Pre-training Small Base LMs with Fewer Tokens}, |
|
author={Sunny Sanyal and Sujay Sanghavi and Alexandros G. Dimakis}, |
|
year={2024}, |
|
eprint={2404.08634}, |
|
archivePrefix={arXiv}, |
|
primaryClass={cs.CL} |
|
} |
|
``` |
