[release/0.5] Fix fasttrackml python client (#907)

Backport of #897

.gitignore

@@ -4,7 +4,6 @@
 # Binaries
 dist/
 fml
-fasttrackml
 *.exe
 __debug_bin*
 
@@ -29,5 +28,4 @@ python/*.egg-info/
 wheelhouse/
 
 # Python package
-!python/fasttrackml/
 __pycache__/

Makefile

@@ -115,7 +115,7 @@ python-env: ## create python virtual environment.
 .PHONY: python-dist
 python-dist: go-build python-env ## build python wheels.
 	@echo '>>> Building Python Wheels.'
-	@VERSION=$(VERSION) pipenv run python3 -m pip wheel ./python/fasttrackml --wheel-dir=wheelhouse --no-deps
+	@VERSION=$(VERSION) pipenv run python3 -m pip wheel ./python --wheel-dir=wheelhouse --no-deps
 
 .PHONY: python-format
 python-format: python-env ## format python code.

python/LICENSE

@@ -0,0 +1 @@
+../LICENSE

python/README.md

@@ -0,0 +1 @@
+../README.md

python/fasttrackml/__init__.py

@@ -1,15 +1,17 @@
 import contextlib
 
 import mlflow
-from fasttrackml.client import FasttrackmlClient
 from mlflow import *
 
+from .client import FasttrackmlClient
+
 del log_metric, log_metrics
-from fasttrackml.fluent import log_metric, log_metrics
+from .fluent import log_metric, log_metrics
 
 __all__ = [name for name in dir() if name in dir(mlflow)]
 __all__.append("FasttrackmlClient")
 
 with contextlib.suppress(Exception):
-    from mlflow import gateway 
+    from mlflow import gateway
+
     __all__.append("gateway")

python/fasttrackml/_tracking_service/client.py

@@ -1,7 +1,5 @@
 from typing import Dict, Optional, Sequence
 
-from fasttrackml.entities.metric import Metric
-from fasttrackml.store.custom_rest_store import CustomRestStore
 from mlflow.entities import Param, RunTag
 from mlflow.store.tracking import GET_METRIC_HISTORY_MAX_RESULTS
 from mlflow.tracking._tracking_service.client import TrackingServiceClient
@@ -13,22 +11,35 @@
 from mlflow.utils.time import get_current_time_millis
 from mlflow.utils.validation import MAX_METRICS_PER_BATCH
 
+from ..entities.metric import Metric
+from ..store.custom_rest_store import CustomRestStore
+
 
 class FasttrackmlTrackingServiceClient(TrackingServiceClient):
 
     def __init__(self, tracking_uri):
         super().__init__(tracking_uri)
         self.custom_store = CustomRestStore(lambda: MlflowHostCreds(self.tracking_uri))
 
-    def log_metric(self, run_id: str, key:str, value:float, timestamp: Optional[int] = None, step:Optional[int] = None, context: Optional[dict] = None):
+    def log_metric(
+        self,
+        run_id: str,
+        key: str,
+        value: float,
+        timestamp: Optional[int] = None,
+        step: Optional[int] = None,
+        context: Optional[dict] = None,
+    ):
         timestamp = timestamp if timestamp is not None else get_current_time_millis()
         step = step if step is not None else 0
         context = context if context else {}
         metric_value = convert_metric_value_to_float_if_possible(value)
         metric = Metric(key, metric_value, timestamp, step, context)
         self.custom_store.log_metric(run_id, metric)
-
-    def log_batch(self, run_id: str, metrics: Sequence[Metric]=(), params: Sequence[Param]=(), tags: Sequence[RunTag]=()):
+
+    def log_batch(
+        self, run_id: str, metrics: Sequence[Metric] = (), params: Sequence[Param] = (), tags: Sequence[RunTag] = ()
+    ):
         for metrics_batch in chunk_list(metrics, chunk_size=MAX_METRICS_PER_BATCH):
             self.custom_store.log_batch(run_id=run_id, metrics=metrics_batch)
 
@@ -56,7 +67,7 @@ def get_metric_history(self, run_id, key):
             history.extend(paged_history)
             token = paged_history.token
         return history
-    
+
     def get_metric_histories(
         self,
         experiment_ids: Optional[Sequence[str]] = None,
@@ -66,7 +77,7 @@ def get_metric_histories(
         run_view_type: int = None,
         max_results: int = 10000000,
         search_all_experiments: bool = False,
-        experiment_names: Optional[Sequence[str]] = None,    
+        experiment_names: Optional[Sequence[str]] = None,
         context: Optional[Dict[str, object]] = None,
     ):
         return self.custom_store.get_metric_histories(