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

Update perf and latest features for llm models (Mar 10) #18821

Merged
merged 3 commits into from
Mar 10, 2025
+24 −20
Merged

Update perf and latest features for llm models (Mar 10) #18821

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
ee0d56b
Update perf and latest features for llm models (Mar 10)
skhorasganiTT Mar 7, 2025
35d4f68
Update MODEL_UPDATES.md
skhorasganiTT Mar 7, 2025
7739d3d
Re-arrange LLMs in model table to group by family, except for top models
skhorasganiTT Mar 7, 2025
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
33 changes: 16 additions & 17 deletions README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -26,25 +26,24 @@

| Model | Batch | Hardware | ttft (ms) | t/s/u | Target<br>t/s/u | t/s | TT-Metalium Release | vLLM Tenstorrent Repo Release |
|---------------------------------------------------------------|-------|----------------------------------------------------------|-----------|-------|-----------------|--------|---------------------------------------------------|---------------------------------------------------------------------------------------------------|
| [QwQ 32B (TP=8)](./models/demos/llama3) | 32 | [QuietBox](https://tenstorrent.com/hardware/tt-quietbox) | 133 | 25.2 | | 464.0 | [main](https://github.com/tenstorrent/tt-metal/) | [9ac3783](https://github.com/tenstorrent/vllm/tree/9ac3783d5e3a4547f879f2cdadaab8571047a0a8) |
| [DeepSeek R1 Distill Llama 3.3 70B (TP=8)](./models/demos/llama3) | 32 | [QuietBox](https://tenstorrent.com/hardware/tt-quietbox) | 180 | 15.2 | 20 | 486.4 | [v0.56.0-rc33](https://github.com/tenstorrent/tt-metal/tree/v0.56.0-rc33) | [9ac3783](https://github.com/tenstorrent/vllm/tree/9ac3783d5e3a4547f879f2cdadaab8571047a0a8) |
| [Llama 3.1 70B (TP=8)](./models/demos/t3000/llama3_70b) | 32 | [QuietBox](https://tenstorrent.com/hardware/tt-quietbox) | 190 | 15.1 | 20 | 483.2 | [v0.54.0-rc2](https://github.com/tenstorrent/tt-metal/tree/v0.54.0-rc2) | [9531611](https://github.com/tenstorrent/vllm/tree/953161188c50f10da95a88ab305e23977ebd3750) |
| [Llama 3.2 11B Vision (TP=2)](./models/demos/llama3) | 16 | [n300](https://tenstorrent.com/hardware/wormhole) | 2550 | 15.8 | 17 | 252.8 | [v0.56.0-rc6](https://github.com/tenstorrent/tt-metal/tree/v0.56.0-rc6) | [b9564bf](https://github.com/tenstorrent/vllm/tree/b9564bf364e95a3850619fc7b2ed968cc71e30b7) |
| [Qwen 2.5 7B (TP=2)](./models/demos/llama3) | 32 | [n300](https://tenstorrent.com/hardware/wormhole) | 126 | 32.5 | 38 | 1040.0 | [v0.56.0-rc33](https://github.com/tenstorrent/tt-metal/tree/v0.56.0-rc33) | [9ac3783](https://github.com/tenstorrent/vllm/tree/9ac3783d5e3a4547f879f2cdadaab8571047a0a8) |
| [Falcon 7B](./models/demos/wormhole/falcon7b) | 32 | [n150](https://tenstorrent.com/hardware/wormhole) | 71 | 18.1 | 26 | 579.2 | [v0.56.0-rc33](https://github.com/tenstorrent/tt-metal/tree/v0.56.0-rc33) | |
| [Mistral 7B](./models/demos/wormhole/mistral7b) | 32 | [n150](https://tenstorrent.com/hardware/wormhole) | | 9.9 | 25 | 316.8 | [v0.51.0-rc28](https://github.com/tenstorrent/tt-metal/tree/v0.51.0-rc28) | |
| [Mamba 2.8B](./models/demos/wormhole/mamba) | 32 | [n150](https://tenstorrent.com/hardware/wormhole) | 48 | 12.3 | 41 | 393.6 | [v0.51.0-rc26](https://github.com/tenstorrent/tt-metal/tree/v0.51.0-rc26) | |
| [Llama 3.1 8B](./models/demos/llama3) | 32 | [n150](https://tenstorrent.com/hardware/wormhole) | 168 | 24.0 | 23 | 768.0 | [v0.56.0-rc6](https://github.com/tenstorrent/tt-metal/tree/v0.56.0-rc6) | [b9564bf](https://github.com/tenstorrent/vllm/tree/b9564bf364e95a3850619fc7b2ed968cc71e30b7) |
| [Llama 3.2 1B](./models/demos/llama3) | 32 | [n150](https://tenstorrent.com/hardware/wormhole) | 56 | 59.4 | 160 | 1900.8 | [v0.56.0-rc6](https://github.com/tenstorrent/tt-metal/tree/v0.56.0-rc6) | [b9564bf](https://github.com/tenstorrent/vllm/tree/b9564bf364e95a3850619fc7b2ed968cc71e30b7) |
| [Llama 3.2 3B](./models/demos/llama3) | 32 | [n150](https://tenstorrent.com/hardware/wormhole) | 97 | 36.5 | 60 | 1168.0 | [v0.56.0-rc6](https://github.com/tenstorrent/tt-metal/tree/v0.56.0-rc6) | [b9564bf](https://github.com/tenstorrent/vllm/tree/b9564bf364e95a3850619fc7b2ed968cc71e30b7) |
| [Falcon 7B (DP=8)](./models/demos/t3000/falcon7b) | 256 | [QuietBox](https://tenstorrent.com/hardware/tt-quietbox) | 88 | 15.5 | 26 | 3968.0 | [v0.56.0-rc33](https://github.com/tenstorrent/tt-metal/tree/v0.56.0-rc33) | |
| [Falcon 40B (TP=8)](./models/demos/t3000/falcon40b) | 32 | [QuietBox](https://tenstorrent.com/hardware/tt-quietbox) | | 5.3 | 36 | 169.6 | [v0.55.0-rc20](https://github.com/tenstorrent/tt-metal/tree/v0.55.0-rc20) | |
| [Mixtral 8x7B (TP=8)](./models/demos/t3000/mixtral8x7b) | 32 | [QuietBox](https://tenstorrent.com/hardware/tt-quietbox) | 227 | 14.9 | 33 | 476.8 | [v0.56.0-rc6](https://github.com/tenstorrent/tt-metal/tree/v0.56.0-rc6) | |
| [Qwen 2.5 72B (TP=8)](./models/demos/llama3) | 32 | [QuietBox](https://tenstorrent.com/hardware/tt-quietbox) | 333 | 14.5 | 20 | 464.0 | [v0.56.0-rc33](https://github.com/tenstorrent/tt-metal/tree/v0.56.0-rc33) | [9ac3783](https://github.com/tenstorrent/vllm/tree/9ac3783d5e3a4547f879f2cdadaab8571047a0a8) |
| [QwQ 32B (TP=8)](./models/demos/llama3) | 32 | [QuietBox](https://tenstorrent.com/hardware/tt-quietbox) | 133 | 25.2 | | 464.0 | [v0.56.0-rc51](https://github.com/tenstorrent/tt-metal/tree/v0.56.0-rc51) | [e2e0002](https://github.com/tenstorrent/vllm/tree/e2e0002ac7dcbc5793983c0f967474d4dcab21f8) |
| [DeepSeek R1 Distill Llama 3.3 70B (TP=8)](./models/demos/llama3) | 32 | [QuietBox](https://tenstorrent.com/hardware/tt-quietbox) | 180 | 15.2 | 20 | 486.4 | [v0.56.0-rc47](https://github.com/tenstorrent/tt-metal/tree/v0.56.0-rc47) | [e2e0002](https://github.com/tenstorrent/vllm/tree/e2e0002ac7dcbc5793983c0f967474d4dcab21f8) |
| [Llama 3.1 70B (TP=8)](./models/demos/llama3) | 32 | [QuietBox](https://tenstorrent.com/hardware/tt-quietbox) | 180 | 15.2 | 20 | 486.4 | [v0.56.0-rc47](https://github.com/tenstorrent/tt-metal/tree/v0.56.0-rc47) | [e2e0002](https://github.com/tenstorrent/vllm/tree/e2e0002ac7dcbc5793983c0f967474d4dcab21f8) |
| [Llama 3.2 11B Vision (TP=2)](./models/demos/llama3) | 16 | [n300](https://tenstorrent.com/hardware/wormhole) | 2550 | 15.8 | 17 | 252.8 | [v0.56.0-rc6](https://github.com/tenstorrent/tt-metal/tree/v0.56.0-rc6) | [e2e0002](https://github.com/tenstorrent/vllm/tree/e2e0002ac7dcbc5793983c0f967474d4dcab21f8) |
| [Qwen 2.5 7B (TP=2)](./models/demos/llama3) | 32 | [n300](https://tenstorrent.com/hardware/wormhole) | 126 | 32.5 | 38 | 1040.0 | [v0.56.0-rc33](https://github.com/tenstorrent/tt-metal/tree/v0.56.0-rc33) | [e2e0002](https://github.com/tenstorrent/vllm/tree/e2e0002ac7dcbc5793983c0f967474d4dcab21f8) |
| [Qwen 2.5 72B (TP=8)](./models/demos/llama3) | 32 | [QuietBox](https://tenstorrent.com/hardware/tt-quietbox) | 333 | 14.5 | 20 | 464.0 | [v0.56.0-rc33](https://github.com/tenstorrent/tt-metal/tree/v0.56.0-rc33) | [e2e0002](https://github.com/tenstorrent/vllm/tree/e2e0002ac7dcbc5793983c0f967474d4dcab21f8) |
| [Falcon 7B](./models/demos/wormhole/falcon7b) | 32 | [n150](https://tenstorrent.com/hardware/wormhole) | 70 | 18.3 | 26 | 585.6 | [v0.56.0-rc47](https://github.com/tenstorrent/tt-metal/tree/v0.56.0-rc47) | |
| [Falcon 7B (DP=8)](./models/demos/t3000/falcon7b) | 256 | [QuietBox](https://tenstorrent.com/hardware/tt-quietbox) | 88 | 15.5 | 26 | 3968.0 | [v0.56.0-rc47](https://github.com/tenstorrent/tt-metal/tree/v0.56.0-rc47) | |
| [Falcon 7B (DP=32)](./models/demos/tg/falcon7b) | 1024 | [Galaxy](https://tenstorrent.com/hardware/galaxy) | 223 | 4.8 | 26 | 4915.2 | [v0.56.0-rc6](https://github.com/tenstorrent/tt-metal/tree/v0.56.0-rc6) | |
| [Falcon 40B (TP=8)](./models/demos/t3000/falcon40b) | 32 | [QuietBox](https://tenstorrent.com/hardware/tt-quietbox) | | 5.3 | 36 | 169.6 | [v0.56.0-rc45](https://github.com/tenstorrent/tt-metal/tree/v0.56.0-rc45) | |
| [Llama 3.1 8B](./models/demos/llama3) | 32 | [n150](https://tenstorrent.com/hardware/wormhole) | 141 | 24.6 | 23 | 787.2 | [v0.56.0-rc47](https://github.com/tenstorrent/tt-metal/tree/v0.56.0-rc47) | [e2e0002](https://github.com/tenstorrent/vllm/tree/e2e0002ac7dcbc5793983c0f967474d4dcab21f8) |
| [Llama 3.2 1B](./models/demos/llama3) | 32 | [n150](https://tenstorrent.com/hardware/wormhole) | 50 | 67.6 | 160 | 2163.2 | [v0.56.0-rc47](https://github.com/tenstorrent/tt-metal/tree/v0.56.0-rc47) | [e2e0002](https://github.com/tenstorrent/vllm/tree/e2e0002ac7dcbc5793983c0f967474d4dcab21f8) |
| [Llama 3.2 3B](./models/demos/llama3) | 32 | [n150](https://tenstorrent.com/hardware/wormhole) | 78 | 43.5 | 60 | 1392.0 | [v0.56.0-rc47](https://github.com/tenstorrent/tt-metal/tree/v0.56.0-rc47) | [e2e0002](https://github.com/tenstorrent/vllm/tree/e2e0002ac7dcbc5793983c0f967474d4dcab21f8) |
| [Mamba 2.8B](./models/demos/wormhole/mamba) | 32 | [n150](https://tenstorrent.com/hardware/wormhole) | 48 | 12.3 | 41 | 393.6 | [v0.51.0-rc26](https://github.com/tenstorrent/tt-metal/tree/v0.51.0-rc26) | |
| [Mistral 7B](./models/demos/wormhole/mistral7b) | 32 | [n150](https://tenstorrent.com/hardware/wormhole) | | 9.9 | 25 | 316.8 | [v0.51.0-rc28](https://github.com/tenstorrent/tt-metal/tree/v0.51.0-rc28) | |
| [Mixtral 8x7B (TP=8)](./models/demos/t3000/mixtral8x7b) | 32 | [QuietBox](https://tenstorrent.com/hardware/tt-quietbox) | 227 | 15.2 | 33 | 486.4 | [v0.56.0-rc47](https://github.com/tenstorrent/tt-metal/tree/v0.56.0-rc47) | |


> **Last Update:** March 6, 2025
> **Last Update:** March 10, 2025
>
> **Notes:**
>
Expand Down
11 changes: 8 additions & 3 deletions models/MODEL_UPDATES.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,13 +4,18 @@
>
> Please refer to the front-page [README](../README.md) for the latest verified release for each model.

## March 10, 2025

### [QwQ-32B](demos/llama3)
- Added support for QwQ-32B on QuietBox.

## February 24, 2025

### [DeepSeek R1 Distill Llama 3.3 70B](demos/llama3)
- Added support for DeepSeek R1 Distill Llama 3.3 70B on T3000.
- Added support for DeepSeek R1 Distill Llama 3.3 70B on QuietBox.

### [Qwen 2.5](demos/llama3)
- Added support for Qwen2.5-7B on N300 and Qwen2.5-72B on T3000.
- Added support for Qwen2.5-7B on N300 and Qwen2.5-72B on QuietBox.

### [Llama 3.1/3.2](demos/llama3)
> **Note:** This feature is available as of release [v0.56.0-rc37](https://github.com/tenstorrent/tt-metal/tree/v0.56.0-rc37)
Expand All @@ -29,7 +34,7 @@
## January 13, 2025

### [Llama 3.1/3.2](demos/llama3)
- Integrated Llama3 models (1B/3B/8B/11B/70B) into [vLLM fork](https://github.com/tenstorrent/vllm/tree/dev/tt_metal) for all compatible Tenstorrent devices (N150/N300/T3000/Galaxy).
- Integrated Llama3 models (1B/3B/8B/11B/70B) into [vLLM fork](https://github.com/tenstorrent/vllm/tree/dev/tt_metal) for all compatible Tenstorrent devices (N150/N300/QuietBox/Galaxy).
- Enabled prefill with the maximum context length (131072) when running the Llama3 text models on smaller devices (N150/N300) via chunked prefill.

## December 16, 2024
Expand Down
Loading