* Switch most compile flows to use ireec.compile_file.
* re-add input type to compile_str path.
* Check if mlir_module exists before checking if it's a path or pyobject.
* Fix some save_dir cases
Print averaged results at the end of all iterations. Increase the
default number of iterations to 5.
Example:
```
Number of iterations: 5
Prefill: avg. 0.03 s, stddev 0.00
Decode: avg. 43.34 tokens/s, stdev 0.13
```
Also remove the -2 in the number of generated tokens -- I did not find
any evidence we need it.
-- This script adds a standalone utility for converting Llama IRs
to dynamic and combining them as well.
Signed-off-by: Abhishek Varma <abhishek@nod-labs.com>
Add flags to enable a non-internactive mode for microbenchmarking llama
models. In this mode, the system and user prompts are specified with CLI
flags, and the number of generated tokens and iterations is fixed.
Also move the stats below the response and trim any response blankspace.
Print a note ahead of a potentially long inactivity to set the right expectations.
Separately, we should add progress to the UI and make this loading faster.
Update vmfb naming for vulkan devices in order to resolve naming
conflicts in the presence of multiple vulkan devices.
Signed-Off-by: Gaurav Shukla <gaurav@nod-labs.com>
-- Currently SHARK suggests that vmfb has been saved, while
that is not the case and no vmfb is generated.
This creates a misdirection for IR/vmfbs which are of larger
size.
-- This commit therefore fixes that misdirection.
Signed-off-by: Abhishek Varma <abhishek@nod-labs.com>
-- This commit fixes the wrong Vulkan device being selected during
runtime.
-- It also adds couple of IREE compilation flags to target specific
Vulkan device.
-- It also changes the Vulkan device listing to be more in tune with
lowering control flow.
Signed-off-by: Abhishek Varma <abhishek@nod-labs.com>
The past key values are only used within the models themselves and can
be kept on device. For vulkan int4, this gives 44 tok/s (for the first
prompt) and settles at around 26 tok/s on 7900xtx.
This allows to pass more arguemnts to the IREE compiler
Example:
python my-app.py --additional_compile_args="--mlir-pretty-debuginfo --mlir-timing"
Co-authored-by: Boian Petkantchin <boian@nod-labs.com>
-- This commit adds Scaled Dot Product Flash Attention's decomposition
in shark_importer.
-- It also updates `iree-flow-enable-data-tiling` to `iree-opt-data-tiling`.
Signed-off-by: Abhishek Varma <abhishek@nod-labs.com>
