Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Llama3 8b perf numbers on A100 #166

Merged
merged 3 commits into from
Jun 17, 2024
Merged

Llama3 8b perf numbers on A100 #166

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
6e00f49
Add Llama3-8B perf numbers
yanboliang Apr 29, 2024
0fb6914
Update
yanboliang Jun 16, 2024
744c927
Update
yanboliang Jun 17, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 7 additions & 0 deletions README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -70,6 +70,7 @@ codellama/CodeLlama-34b-Python-hf
mistralai/Mistral-7B-v0.1
mistralai/Mistral-7B-Instruct-v0.1
mistralai/Mistral-7B-Instruct-v0.2
meta-llama/Meta-Llama-3-8B
```

For example, to convert Llama-2-7b-chat-hf
Expand All @@ -89,6 +90,8 @@ Benchmarks run on an 8xA100-80GB, power limited to 330W with a hybrid cube mesh
| Llama-2-70B | Base | OOM ||
| | 8-bit | 19.13 | 1322.58 |
| | 4-bit (G=32) | 25.25 | 1097.66 |
| Llama-3-8B | Base | 94.25 | 1411.95 |
| | 8-bit | 139.55 | 1047.23 |

### Speculative Sampling
[Verifier: Llama-70B (int4), Draft: Llama-7B (int4)](./scripts/speculate_70B_int4.sh): 48.4 tok/s
Expand All @@ -104,6 +107,10 @@ Benchmarks run on an 8xA100-80GB, power limited to 330W with a hybrid cube mesh
| | 2 | 21.32 | 1481.87 |
| | 4 | 38.01 | 1340.76 |
| | 8 | 62.50 | 1135.29 |
| Llama-3-8B | 1 | 94.19 | 1411.76 |
| | 2 | 150.48 | 1208.80 |
| | 4 | 219.77 | 991.63 |
| | 8 | 274.65 | 768.55 |

### Tensor Parallelism + Quantization
| Model | Technique | Tokens/Second | Memory Bandwidth (GB/s) |
Expand Down