code/README.md · changh95/hamer-p150 at main

Python API: warm-up so the first real call is fast (warmup_variants, model.warmup()), quiet logs, install extras

77e087f verified about 1 month ago

768 Bytes

# hamer (code/)

The `hamer` Python package: HaMeR hand mesh recovery on a Tenstorrent Blackhole chip, plus its HTTP server
(`hamer.server.app`).  Install into an environment that already has a built tt-metal / ttnn:

```bash
pip install -e code/            # or: pip install code/
```

```python
from hamer import HamerModel
with HamerModel.from_pretrained(device_id=0) as model:   # also warms up all call variants
    hand = model("media/sample1.jpg", bbox=[20, 170, 295, 260], is_right=True)   # path relative to the repo root
```

`from_pretrained` warms up the device and all call variants, so the first call is as fast as the later calls.
Install `code/[server,test]` for the HTTP server and the tests.

Full API reference: `PYTHON.md` at the repository root.

	# hamer (code/)

	The `hamer` Python package: HaMeR hand mesh recovery on a Tenstorrent Blackhole chip, plus its HTTP server
	(`hamer.server.app`). Install into an environment that already has a built tt-metal / ttnn:

	```bash
	pip install -e code/ # or: pip install code/
	```

	```python
	from hamer import HamerModel

	with HamerModel.from_pretrained(device_id=0) as model: # also warms up all call variants
	hand = model("media/sample1.jpg", bbox=[20, 170, 295, 260], is_right=True) # path relative to the repo root
	```

	`from_pretrained` warms up the device and all call variants, so the first call is as fast as the later calls.
	Install `code/[server,test]` for the HTTP server and the tests.

	Full API reference: `PYTHON.md` at the repository root.