Merge branch 'debug-parallel'

Author: pskrunner14

python/scripts/batch_decode.py

@@ -16,6 +16,7 @@
 """
 
 import json
+import random
 import traceback
 
 from typing import List
@@ -67,7 +68,7 @@ def transcribe_audio(audio_stream, model: str, language_code: str, sample_rate=8
             data_bytes=len(audio_stream)
         )
 
-        response = client.recognize(config, audio, uuid="", timeout=1000)
+        response = client.recognize(config, audio, uuid=str(random.randint(1000, 100000)), timeout=1000)
     except Exception as e:
         print(f"error: {str(e)}")
         return []

python/scripts/example_client.py

@@ -82,7 +82,7 @@ def transcribe_chunks_streaming(client, audio_chunks, model: str, language_code:
                 data_bytes=chunk_len
             )
             audio_params = [(config(len(chunk)), RecognitionAudio(content=chunk)) for chunk in audio_chunks]
-            response = client.streaming_recognize_raw(audio_params, uuid="")
+            response = client.streaming_recognize_raw(audio_params, uuid=str(random.randint(1000, 100000)))
         else:
             audio = (RecognitionAudio(content=chunk) for chunk in audio_chunks)
             config = RecognitionConfig(
@@ -93,7 +93,7 @@ def transcribe_chunks_streaming(client, audio_chunks, model: str, language_code:
                 model=model,
                 word_level=word_level
             )
-            response = client.streaming_recognize(config, audio, uuid="")
+            response = client.streaming_recognize(config, audio, uuid=str(random.randint(1000, 100000)))
     except Exception as e:
         traceback.print_exc()
         print(f'error: {str(e)}')
@@ -125,7 +125,7 @@ def audio_params_gen(audio_chunks):
                 for chunk in audio_chunks:
                     yield config(len(chunk)), RecognitionAudio(content=chunk)
 
-            response_gen = client.bidi_streaming_recognize_raw(audio_params_gen(audio_chunks), uuid="")
+            response_gen = client.bidi_streaming_recognize_raw(audio_params_gen(audio_chunks), uuid=str(random.randint(1000, 100000)))
         else:
             config = RecognitionConfig(
                 sample_rate_hertz=sample_rate,
@@ -140,7 +140,7 @@ def audio_chunks_gen(audio_chunks):
                 for chunk in audio_chunks:
                     yield RecognitionAudio(content=chunk)
 
-            response_gen = client.bidi_streaming_recognize(config, audio_chunks_gen(audio_chunks), uuid="")
+            response_gen = client.bidi_streaming_recognize(config, audio_chunks_gen(audio_chunks), uuid=str(random.randint(1000, 100000)))
     except Exception as e:
         traceback.print_exc()
         print(f'error: {str(e)}')

python/scripts/parse_logs.py

@@ -0,0 +1,65 @@
+"""
+Script for parsing server debug logs.
+
+Usage:
+  parse_logs.py <logs-file> <metrics-csv>
+"""
+
+from docopt import docopt
+
+import pandas as pd
+
+
+def main(args):
+    df = parse_logs(args["<logs-file>"])
+    df.to_csv(args["<metrics-csv>"], index=False)
+
+
+def parse_logs(logs_file):
+    with open(logs_file, "r", encoding="utf-8") as f:
+        logs = f.read().split("\n")
+    
+    uuids = []
+
+    for log in logs:
+        if "decoder acquired" in log:
+            parts = log.split()
+            uuids.append(parts[parts.index("uuid:") + 1])
+
+    uuids = list(set(uuids))
+
+    df = pd.DataFrame(columns=["uuid", "chunk#", "read", "accepted", "decoded", "computed"])
+
+    for uuid in uuids:
+        l_logs = list(filter(lambda x: uuid in x, logs))
+        chunks = []
+        for log in l_logs:
+            if "chunk" in log and "received" in log:
+                parts = log.split()
+                chunks.append("chunk " + parts[parts.index("chunk") + 1])
+        chunks = sorted(list(set(chunks)))
+
+        for chunk in chunks:
+            chunk_logs = list(filter(lambda x: chunk in x, l_logs))
+            idx = l_logs.index(chunk_logs[0])
+
+            read = float(l_logs[idx + 1].split()[-1].replace("ms", ""))
+            accepted = float(l_logs[idx + 2].split()[-1].replace("ms", ""))
+            decoded = float(l_logs[idx + 3].split()[-1].replace("ms", ""))
+            computed = float(chunk_logs[1].split()[-1].replace("ms", ""))
+
+            df = df.append({
+                "uuid": uuid,
+                "chunk#": chunk,
+                "read": read,
+                "accepted": accepted,
+                "decoded": decoded,
+                "computed": computed
+            }, ignore_index=True)
+
+    return df
+
+
+if __name__ == "__main__":
+    args = docopt(__doc__)
+    main(args)