[Autotuner] WIP: CLI refactor

docs/user/CLIGuideAutotuner.md

@@ -0,0 +1,70 @@
+# Autotuner CLI Guide
+
+AutoTuner may be triggered by specifying a configuration path instead of
+typing every single argument one-by-one.
+
+## Motivation of Configuration files
+
+Why configuration files?
+- Improve CLI user experience - less typing, less mistakes.
+- Greatly simplify the reproducibility of experiments - share your configuration
+file for easier experiment tracking!
+
+Previously for a command like tune, you might have had to do:
+
+```shell
+openroad_autotuner --design gcd --platform sky130hd --experiment abcdef \
+                   --verbose \
+                   --jobs 4 --openroad_threads 16 \
+                   tune \
+                   --config ../../flow/designs/sky130hd/gcd/autotuner.json \
+                   --samples 5 --iterations 1 --algorithm hyperopt \
+                   --resources_per_trial 1.0 --seed 42
+```
+
+With our new approach all you have to do is specify a YAML document `cli.yaml`
+
+```yaml
+---
+# Common
+design: gcd
+platform: sky130hd
+experiment: test
+verbose: 0
+
+# Workload
+jobs: 4 
+openroad_threads: 16
+
+# Tune-specific (set these if mode is tune)
+tune:
+  algorithm: hyperopt
+  eval: default
+  samples: 10
+  iterations: 1
+  resources_per_trial: 1.0
+  reference: null
+  perturbation: 25
+  seed: 42
+  config: ../../flow/designs/sky130hd/gcd/autotuner.json
+```
+
+and run:
+
+```bash
+# Make sure you run this in `./tools/AutoTuner` directory.
+python3 -m autotuner.distributed --yaml cli.yaml tune
+```
+
+## How to generate new config files
+
+```bash
+python3 -m autotuner.distributed --design gcd --platform sky130hd --experiment abcdef \
+                   --verbose \
+                   --jobs 4 --openroad_threads 16 \
+                   --print_config \
+                   tune \
+                   --config ../../flow/designs/sky130hd/gcd/autotuner.json \
+                   --samples 5 --iterations 1 --algorithm hyperopt \
+                   --resources_per_trial 1.0 --seed 42 > new_test.yaml
+```

docs/user/InstructionsForAutoTuner.md

@@ -104,6 +104,9 @@ The `autotuner.distributed` module uses [Ray's](https://docs.ray.io/en/latest/in
 fully utilize available hardware resources from a single server 
 configuration, on-premise or over the cloud with multiple CPUs.
 
+For advanced users, we have provided a configuration file, please
+refer to [this guide](./CLIGuideAutotuner.md).
+
 The two modes of operation:
 - `sweep`, where every possible parameter combination in the search space is tested
 - `tune`, where we use Ray's Tune feature to intelligently search the space and optimize hyperparameters using one of the algorithms listed above.
@@ -124,25 +127,25 @@ The following commands should be run from `./tools/AutoTuner`.
 
 #### Tune only 
 
-* AutoTuner: `openroad_autotuner tune -h`
+* AutoTuner: `python -m autotuner.distributed tune -h`
 
 Example:
 
 ```shell
-openroad_autotuner --design gcd --platform sky130hd \
+python -m autotuner.distributed --design gcd --platform sky130hd \
+                       tune --samples 5 \
                        --config ../../flow/designs/sky130hd/gcd/autotuner.json \
-                       tune --samples 5
 ```
 #### Sweep only 
 
-* Parameter sweeping: `openroad_autotuner sweep -h`
+* Parameter sweeping: `python -m autotuner.distributed sweep -h`
 
 Example:
 
 ```shell
-openroad_autotuner --design gcd --platform sky130hd \
-                       --config src/autotuner/distributed-sweep-example.json \
-                       sweep
+python -m autotuner.distributed --design gcd --platform sky130hd \
+                       sweep \
+                       --config src/autotuner/distributed-sweep-example.json
 ```
 
 #### Plot images

tools/AutoTuner/cli.yaml

@@ -0,0 +1,29 @@
+---
+# Common
+design: gcd
+platform: sky130hd
+experiment: test
+timeout: null
+verbose: 0
+
+# Workload
+jobs: 4 
+openroad_threads: 16
+server: null
+port: 10001
+
+# Tune-specific
+tune:
+  config: ../../flow/designs/sky130hd/gcd/autotuner.json
+  algorithm: hyperopt
+  eval: default
+  samples: 10
+  iterations: 1
+  resources_per_trial: 1.0
+  reference: null
+  perturbation: 25
+  seed: 42
+
+# Sweep-specific
+sweep:
+  config: ./src/autotuner/distributed-sweep-example.json

tools/AutoTuner/requirements.txt

@@ -11,3 +11,4 @@ SQLAlchemy==1.4.17
 urllib3>=1.26.17
 matplotlib==3.10.0
 pyyaml==6.0.1
+jsonargparse==4.38.0