The TinyLlama project aims to pretrain a 1.1B Llama model on 3 trillion tokens. With some proper optimization, we can achieve this within a span of "just" 90 days using 16 A100-40G GPUs 🚀🚀. The training has started on 2023-09-01.
We adopted exactly the same architecture and tokenizer as Llama 2. This means TinyLlama can be plugged and played in many open-source projects built upon Llama. Besides, TinyLlama is compact with only 1.1B parameters. This compactness allows it to cater to a multitude of applications demanding a restricted computation and memory footprint.
Releases Schedule
We will be rolling out intermediate checkpoints following the below schedule. We also include some baseline models for comparison.
| Date | HF Checkpoint | Tokens | Step | HellaSwag Acc_norm |
|---|---|---|---|---|
| Baseline | StableLM-Alpha-3B | 800B | -- | 38.31 |
| Baseline | Pythia-1B-intermediate-step-50k-105b | 105B | 50k | 42.04 |
| Baseline | Pythia-1B | 300B | 143k | 47.16 |
| 2023-09-04 | TinyLlama-1.1B-intermediate-step-50k-105b | 105B | 50k | 43.50 |
| 2023-09-16 | -- | 500B | -- | -- |
| 2023-10-01 | -- | 1T | -- | -- |
| 2023-10-16 | -- | 1.5T | -- | -- |
| 2023-10-31 | -- | 2T | -- | -- |
| 2023-11-15 | -- | 2.5T | -- | -- |
| 2023-12-01 | -- | 3T | -- | -- |
It can be observed that TinyLlama has so far progressed well 🎉🎉.
Meanwhile, you can track the live cross entropy loss here.
Training Details
Below are some details of our training setup:
| Setting | Description |
|---|---|
| Parameters | 1.1B |
| Attention Variant | Grouped Query Attention |
| Model Size | Layers: 22, Heads: 32, Query Groups: 4, Embedding Size: 2048, Intermediate Size (Swiglu): 5632 |
| Sequence Length | 2048 |
| Batch Size | 2 million tokens (2048 * 1024) |
| Learning Rate | 4e-4 |
| Learning Rate Schedule | Cosine with 2000 warmup steps |
| Training Data | Slimpajama & Starcoderdata |
| Data Preprocessing | Excluded GitHub subset of Slimpajama; Sampled all code from Starcoderdata |
| Combined Dataset Size | 1 trillion tokens |
| Total Tokens During Training | 3 trillion (3 epochs/1430k steps) |
| Natural Language to Code Ratio | 7:3 |
| Hardware | 16 A100-40G GPUs |