Skip to content

Commit

Permalink
Added descriptions for environment variable overrides to README
Browse files Browse the repository at this point in the history
  • Loading branch information
AleksKnezevic committed Jan 3, 2025
1 parent 2566b58 commit 4014010
Showing 1 changed file with 14 additions and 1 deletion.
15 changes: 14 additions & 1 deletion README.md
Original file line number Diff line number Diff line change
Expand Up @@ -3,7 +3,7 @@
The tt-torch project requires environment setup from tt-mlir project. https://github.com/tenstorrent/tt-mlir/
See documentation of tt-mlir and follow Environment setup. https://docs.tenstorrent.com/tt-mlir/build.html

## Steps:
## Compile Steps:
```
source env/activate
cmake -G Ninja -B build
Expand All @@ -26,3 +26,16 @@ pre-commit run --all-files
```

For more information visit [pre-commit](https://pre-commit.com/)


## Controlling Behaviour

You can use the following environment variables to override default behaviour:

| Environment Variable | Behaviour | Default |
| -------------------- | --------- | --------
| TT_TORCH_COMPILE_DEPTH | Sets the maximum compile depth, see `tt_torch/tools/utils.py` for options. | `EXECUTE` |
| TT_TORCH_VERIFY_INTERMEDIATES | Sets whether to verify intermediate tensors agains pytorch when running with compile depth `EXECUTE_OP_BY_OP`. | False |
| TT_TORCH_CONSTEVAL | Sets whether to enable consteval on the torch fx graph before compiling. | False |
| TT_TORCH_CONSTEVAL_PARAMETERS | Sets whether to also consteval the parameters, not only the embedded constants. | False |
| TT_TORCH_ENABLE_IR_PRINTING | Sets whether to enable printing MLIR for all conversion steps from StableHLO to TTNN. Be warned, this forces single core compile, so is much slower. | False |

0 comments on commit 4014010

Please sign in to comment.