Frogger

.gitignore

@@ -1,2 +1,8 @@
 __pycache__/
-scripts/notebooks/
+scripts/notebooks/
+matrix.egg-info/
+build/
+.pyre/
+.pyre_configuration
+.pyre_configuration.local
+.watchmanconfig

matrix/app_server/app_api.py

@@ -32,6 +32,7 @@
 from matrix.client.endpoint_cache import EndpointCache
 from matrix.common.cluster_info import ClusterInfo, get_head_http_host
 from matrix.utils.basics import convert_to_json_compatible, sanitize_app_name
+from matrix.utils.logging import get_logger
 from matrix.utils.os import download_s3_dir, lock_file, run_and_stream, run_async
 from matrix.utils.ray import (
     ACTOR_NAME_SPACE,
@@ -42,7 +43,7 @@
     kill_matrix_actors,
 )
 
-logger = logging.getLogger("ray.serve")
+logger = get_logger("ray.serve")
 
 DEPLOYMENT_YAML = "deployment.yaml"
 DEPLOYMENT_SGLANG_YAML = "deployment_sglang.yaml"

matrix/app_server/code/code_execution_app.py

@@ -15,9 +15,10 @@
 from starlette.responses import JSONResponse
 
 from matrix.app_server.code.sandbox_runner import SandboxRunner
+from matrix.utils.logging import get_logger
 
 CODE_EXEC_TIMEOUT = 10
-logger = logging.getLogger("ray.serve")
+logger = get_logger("ray.serve")
 
 
 @serve.deployment(ray_actor_options={"num_cpus": 1, "num_gpus": 0})

matrix/app_server/llm/azure_openai_proxy.py

@@ -23,7 +23,9 @@
     ErrorResponse,
 )
 
-logger = logging.getLogger("ray.serve")
+from matrix.utils.logging import get_logger
+
+logger = get_logger("ray.serve")
 
 app = FastAPI()
 
@@ -114,7 +116,7 @@ def build_app(cli_args: Dict[str, str]) -> serve.Application:
             arg_strings.extend([f"--{key}"])
         else:
             arg_strings.extend([f"--{key}", str(value)])
-    logger.info(arg_strings)
+    logger.info(",".join(arg_strings))
 
     args = argparse.parse_args(args=arg_strings)
 

matrix/app_server/llm/bedrock_proxy.py

@@ -19,7 +19,9 @@
 from starlette.requests import Request
 from vllm.entrypoints.openai.protocol import ChatCompletionRequest
 
-logger = logging.getLogger("ray.serve")
+from matrix.utils.logging import get_logger
+
+logger = get_logger("ray.serve")
 logging.getLogger("httpx").setLevel(logging.WARNING)
 logging.getLogger("openai._base_client").setLevel(logging.WARNING)
 
@@ -124,11 +126,11 @@ def build_app(cli_args: Dict[str, str]) -> serve.Application:
             arg_strings.extend([f"--{key}"])
         else:
             arg_strings.extend([f"--{key}", str(value)])
-    logger.info(arg_strings)
+    logger.info(",".join(arg_strings))
 
     args = argparse.parse_args(args=arg_strings)
 
-    logger.log(logging.INFO, f"args: {args}")
+    logger.info(f"args: {args}")
     assert "claude" in args.model_name.lower(), "Only Claude model is supported"
 
     return BedrockDeployment.options(  # type: ignore[attr-defined]

matrix/app_server/llm/deploy_sglang_app.py

@@ -27,10 +27,11 @@
 
 from matrix.common.cluster_info import ClusterInfo
 from matrix.utils.http import fetch_url, post_url
+from matrix.utils.logging import get_logger
 from matrix.utils.os import run_and_stream, stop_process
 from matrix.utils.ray import ACTOR_NAME_SPACE, get_matrix_actors, get_ray_head_node
 
-logger = logging.getLogger("ray.sglang")
+logger = get_logger("ray.sglang")
 handler = logging.StreamHandler()
 formatter = logging.Formatter("%(asctime)s - %(levelname)s - %(name)s - %(message)s")
 handler.setFormatter(formatter)

matrix/app_server/llm/gemini_proxy.py

@@ -17,7 +17,9 @@
 from starlette.requests import Request
 from vllm.entrypoints.openai.protocol import ChatCompletionRequest
 
-logger = logging.getLogger("ray.serve")
+from matrix.utils.logging import get_logger
+
+logger = get_logger("ray.serve")
 
 app = FastAPI()
 
@@ -209,7 +211,7 @@ def build_app(cli_args: Dict[str, str]) -> serve.Application:
             arg_strings.extend([f"--{key}"])
         else:
             arg_strings.extend([f"--{key}", str(value)])
-    logger.info(arg_strings)
+    logger.info(",".join(arg_strings))
 
     args = argparse.parse_args(args=arg_strings)
 

matrix/app_server/llm/metagen_proxy.py

@@ -24,8 +24,9 @@
 )
 
 from matrix.utils.http import post_url
+from matrix.utils.logging import get_logger
 
-logger = logging.getLogger("ray.serve")
+logger = get_logger("ray.serve")
 
 app = FastAPI()
 
@@ -124,7 +125,7 @@ def build_app(cli_args: Dict[str, str]) -> serve.Application:
             arg_strings.extend([f"--{key}"])
         else:
             arg_strings.extend([f"--{key}", str(value)])
-    logger.info(arg_strings)
+    logger.info(",".join(arg_strings))
 
     args = argparse.parse_args(args=arg_strings)
 

matrix/app_server/llm/ray_serve_fastgen.py

@@ -31,7 +31,9 @@
 from ray.util.scheduling_strategies import PlacementGroupSchedulingStrategy
 from torch.distributed.device_mesh import DeviceMesh, init_device_mesh
 
-logger = logging.getLogger("ray.serve")
+from matrix.utils.logging import get_logger
+
+logger = get_logger("ray.serve")
 
 app = FastAPI()
 
@@ -350,7 +352,7 @@ def parse_args(cli_args: Dict[str, str]):
             arg_strings.extend([f"--{key}"])
         else:
             arg_strings.extend([f"--{key}", str(value)])
-    logger.info(arg_strings)
+    logger.info(",".join(arg_strings))
     parsed_args = parser.parse_args(args=arg_strings)
     return parsed_args
 

matrix/app_server/llm/ray_serve_vllm.py

@@ -41,6 +41,7 @@
 )
 
 from matrix.app_server.llm import openai_pb2
+from matrix.utils.logging import get_logger
 
 try:
     from vllm.entrypoints.openai.serving_engine import (  # type: ignore[attr-defined]
@@ -76,7 +77,7 @@
     "enable_tools",
 ]
 
-logger = logging.getLogger("ray.serve")
+logger = get_logger("ray.serve")
 
 app = FastAPI()
 
@@ -534,7 +535,7 @@ def parse_vllm_args(cli_args: Dict[str, str]):
                 arg_strings.extend([f"--{key}"])
             else:
                 arg_strings.extend([f"--{key}", str(value)])
-    logger.info(arg_strings)
+    logger.info(",".join(arg_strings))
     parsed_args = parser.parse_args(args=arg_strings)
     return parsed_args, deploy_args
 

matrix/app_server/llm/sagemaker_proxy.py

@@ -18,7 +18,9 @@
 from starlette.requests import Request
 from vllm.entrypoints.openai.protocol import ChatCompletionRequest
 
-logger = logging.getLogger("ray.serve")
+from matrix.utils.logging import get_logger
+
+logger = get_logger("ray.serve")
 logging.getLogger("httpx").setLevel(logging.WARNING)
 logging.getLogger("openai._base_client").setLevel(logging.WARNING)
 
@@ -122,11 +124,11 @@ def build_app(cli_args: Dict[str, str]) -> serve.Application:
             arg_strings.extend([f"--{key}"])
         else:
             arg_strings.extend([f"--{key}", str(value)])
-    logger.info(arg_strings)
+    logger.info(",".join(arg_strings))
 
     args = argparse.parse_args(args=arg_strings)
 
-    logger.log(logging.INFO, f"args: {args}")
+    logger.info(f"args: {args}")
 
     return SageMakerDeployment.options(  # type: ignore[attr-defined]
         placement_group_bundles=pg_resources,

matrix/app_server/vision/optical_flow.py

@@ -21,8 +21,9 @@
 from torch.utils.data import DataLoader, Dataset
 
 from matrix.app_server.vision.utils import SamplingMode, TorchCodecVideoDataset
+from matrix.utils.logging import get_logger
 
-logger = logging.getLogger("ray.serve")
+logger = get_logger("ray.serve")
 logging.getLogger("httpx").setLevel(logging.WARNING)
 logging.getLogger("openai._base_client").setLevel(logging.WARNING)
 
@@ -207,7 +208,7 @@ def build_app(cli_args: Dict[str, str]) -> serve.Application:
     for key, value in cli_args.items():
         if value is not None:
             arg_strings.extend([f"--{key}", str(value)])
-    logger.info(arg_strings)
+    logger.info(",".join(arg_strings))
 
     args = argparse.parse_args(arg_strings)
 

matrix/app_server/vision/perception_encoder.py

@@ -29,8 +29,9 @@
     execute_with_retry,
     get_image_transform,
 )
+from matrix.utils.logging import get_logger
 
-logger = logging.getLogger("ray.serve")
+logger = get_logger("ray.serve")
 logging.getLogger("httpx").setLevel(logging.WARNING)
 logging.getLogger("openai._base_client").setLevel(logging.WARNING)
 
@@ -289,7 +290,7 @@ def build_app(cli_args: Dict[str, str]) -> serve.Application:
     for key, value in cli_args.items():
         if value is not None:
             arg_strings.extend([f"--{key}", str(value)])
-    logger.info(arg_strings)
+    logger.info(",".join(arg_strings))
 
     args = argparse.parse_args(arg_strings)
 

matrix/app_server/vision/utils.py

@@ -20,7 +20,9 @@
 from torch.utils.data import Dataset
 from torchcodec.decoders import VideoDecoder
 
-logger = logging.getLogger("matrix.app_server.vision.utils")
+from matrix.utils.logging import get_logger
+
+logger = get_logger("matrix.app_server.vision.utils")
 logging.getLogger("httpx").setLevel(logging.WARNING)
 
 

matrix/client/client_utils.py

@@ -13,8 +13,9 @@
 import typing as tp
 
 from matrix.client.endpoint_cache import EndpointCache
+from matrix.utils.logging import get_logger
 
-logger = logging.getLogger("matrix.client.utils")
+logger = get_logger("matrix.client.utils")
 logging.getLogger("httpx").setLevel(logging.WARNING)
 
 

matrix/client/container_client.py

@@ -11,8 +11,9 @@
 import aiohttp
 
 from matrix.utils.http import post_url
+from matrix.utils.logging import get_logger
 
-logger = logging.getLogger(__name__)
+logger = get_logger(__name__)
 
 
 class ContainerClient:

matrix/client/endpoint_cache.py

@@ -11,9 +11,10 @@
 
 from matrix.common.cluster_info import ClusterInfo, get_head_http_host
 from matrix.utils.http import fetch_url
+from matrix.utils.logging import get_logger
 from matrix.utils.ray import get_ray_dashboard_address
 
-logger = logging.getLogger("endpoint_cache")
+logger = get_logger("endpoint_cache")
 
 
 class EndpointCache:

matrix/client/execute_code.py

@@ -19,12 +19,13 @@
 
 from matrix.client.client_utils import get_an_endpoint_url, save_to_jsonl
 from matrix.client.endpoint_cache import EndpointCache
+from matrix.utils.logging import get_logger
 
 # Configure logging for execute_code.py
 logging.basicConfig(
     level=logging.INFO, format="%(asctime)s - %(name)s - %(levelname)s - %(message)s"
 )
-logger = logging.getLogger("execute_code")
+logger = get_logger("execute_code")
 # Optionally suppress noisy logs from imported modules if not already done
 logging.getLogger("httpx").setLevel(logging.WARNING)
 

matrix/client/process_vision_data.py

@@ -23,11 +23,12 @@
     save_to_jsonl,
 )
 from matrix.client.endpoint_cache import EndpointCache
+from matrix.utils.logging import get_logger
 
 logging.basicConfig(
     level=logging.INFO, format="%(asctime)s - %(name)s - %(levelname)s - %(message)s"
 )
-logger = logging.getLogger("process_vision_data")
+logger = get_logger("process_vision_data")
 logging.getLogger("httpx").setLevel(logging.WARNING)
 
 

matrix/client/query_llm.py

@@ -32,13 +32,14 @@
 from matrix.app_server.llm import openai_pb2, openai_pb2_grpc
 from matrix.client.client_utils import get_an_endpoint_url, save_to_jsonl
 from matrix.client.endpoint_cache import EndpointCache
+from matrix.utils.logging import get_logger
 from matrix.utils.os import batch_requests_async, run_async
 
 CHAR_PER_TOKEN = 3.61
 logging.basicConfig(
     level=logging.INFO, format="%(asctime)s - %(name)s - %(levelname)s - %(message)s"
 )
-logger = logging.getLogger("query_llm")
+logger = get_logger("query_llm")
 logging.getLogger("httpx").setLevel(logging.WARNING)
 logging.getLogger("openai._base_client").setLevel(logging.WARNING)
 

matrix/cluster/ray_dashboard_job.py

@@ -15,11 +15,13 @@
 
 import ray
 
+from matrix.utils.logging import get_logger
+
 # Configure logging
 logging.basicConfig(
     level=logging.INFO, format="%(asctime)s - %(name)s - %(levelname)s - %(message)s"
 )
-logger = logging.getLogger("RayDashboardJob")
+logger = get_logger("RayDashboardJob")
 
 
 @ray.remote(max_restarts=10)

matrix/data_pipeline/clustering/sample.py

@@ -131,7 +131,7 @@ def select_sample_ids_from_cluster(cluster_df):
         else:
             embeddings = np.vstack(cluster_df["embedding"])
             centroid = np.mean(embeddings, axis=0)
-            logger.info(embeddings.shape, centroid.shape)
+            logger.info(f"{embeddings.shape}, {centroid.shape}")
             distances_to_centroid = np.linalg.norm(embeddings - centroid, axis=1)
 
             # First pick the closest point to centroid

matrix/data_pipeline/clustering/utils.py

@@ -20,12 +20,13 @@
 from sentence_transformers import SentenceTransformer
 
 from matrix.utils.basics import get_nested_value, get_user_message_from_llama3_prompt
+from matrix.utils.logging import get_logger
 
 # Basic Logging Setup
 logging.basicConfig(
     level=logging.INFO, format="%(asctime)s - %(levelname)s - %(message)s"
 )
-logger = logging.getLogger(__name__)
+logger = get_logger(__name__)
 
 # --- Model Saving/Loading ---
 

matrix/data_pipeline/generate/vllm_generate.py

@@ -15,7 +15,9 @@
 from fire import Fire
 from vllm import LLM, SamplingParams
 
-logger = logging.getLogger(__name__)
+from matrix.utils.logging import get_logger
+
+logger = get_logger(__name__)
 
 USER_MESSAGE = "<user_message>"