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

Minor update to note about llama70b implemetations in tt-metal readme #52

Merged
merged 1 commit into from
Jan 10, 2025
Merged

Minor update to note about llama70b implemetations in tt-metal readme #52

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
2 changes: 1 addition & 1 deletion tt_metal/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -65,7 +65,7 @@ MESH_DEVICE=T3K_LINE WH_ARCH_YAML=wormhole_b0_80_arch_eth_dispatch.yaml python e

**Note 1 (Llama70B)**: To run Llama70B on Galaxy, set `MESH_DEVICE=TG` and do not set `WH_ARCH_YAML=...`.

**Note 2 (Llama70B)**: By default, this will run the newer tt-metal implementation of Llama70B from the [Llama3 demo](https://github.com/tenstorrent/tt-metal/tree/main/models/demos/llama3) which is currently ~10% slower than the [old Llama70b implemenentation](https://github.com/tenstorrent/tt-metal/tree/main/models/demos/t3000/llama3_70b). To run with the old implementation, set `MESH_DEVICE=T3K_RING` and modify the `TtLlamaForCausalLM` model import in [offline_inference_tt.py](https://github.com/tenstorrent/vllm/blob/dev/examples/offline_inference_tt.py) to `from models.demos.t3000.llama2_70b.tt.generator_vllm import TtLlamaForCausalLM`.
**Note 2 (Llama70B)**: By default, this will run the newer tt-metal implementation of Llama70B from the [Llama3 demo](https://github.com/tenstorrent/tt-metal/tree/main/models/demos/llama3). To run with the [old Llama70B implemenentation](https://github.com/tenstorrent/tt-metal/tree/main/models/demos/t3000/llama3_70b), set `MESH_DEVICE=T3K_RING` and modify the `TtLlamaForCausalLM` model import in [offline_inference_tt.py](https://github.com/tenstorrent/vllm/blob/dev/examples/offline_inference_tt.py) to `from models.demos.t3000.llama2_70b.tt.generator_vllm import TtLlamaForCausalLM`.

**Note 3 (Other Llama Versions)**: By default, the inference example will run with Llama-3.1-70B. To run with Llama-3.1-8B, Llama-3.2-1B, or Llama-3.2-3B, ensure that the apprioriate environment variables are set as per the [demo instructions](https://github.com/tenstorrent/tt-metal/tree/main/models/demos/llama3), then set `MESH_DEVICE=<device>` (valid options for `<device>` are `N150`, `N300`, `T3K_LINE`, or `TG`) and one of the following:
- Llama-3.1-8B: `--model "meta-llama/Meta-Llama-3.1-8B"`
Expand Down
Loading