Commit
·
6896250
1
Parent(s):
2bd198e
realtime flag added to websockets route
Browse files
app.py
CHANGED
@@ -23,6 +23,7 @@ from typing import Optional
|
|
23 |
|
24 |
|
25 |
import json, asyncio, base64
|
|
|
26 |
|
27 |
# --- Patch T5X mesh helpers for GPUs on JAX >= 0.7 (coords present, no core_on_chip) ---
|
28 |
def _patch_t5x_for_gpu_coords():
|
@@ -902,6 +903,27 @@ def read_root():
|
|
902 |
"""
|
903 |
return Response(content=html_content, media_type="text/html")
|
904 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
905 |
@app.websocket("/ws/jam")
|
906 |
async def ws_jam(websocket: WebSocket):
|
907 |
await websocket.accept()
|
@@ -1004,44 +1026,61 @@ async def ws_jam(websocket: WebSocket):
|
|
1004 |
# seed a fresh state with a silent context like warmup
|
1005 |
mrt = get_mrt()
|
1006 |
state = mrt.init_state()
|
1007 |
-
# build exact-length silent context
|
1008 |
codec_fps = float(mrt.codec.frame_rate)
|
1009 |
ctx_seconds = float(mrt.config.context_length_frames) / codec_fps
|
1010 |
sr = int(mrt.sample_rate)
|
1011 |
samples = int(max(1, round(ctx_seconds * sr)))
|
1012 |
silent = au.Waveform(np.zeros((samples,2), np.float32), sr)
|
1013 |
-
|
1014 |
-
tokens = tokens_full[:, :mrt.config.decoder_codec_rvq_depth]
|
1015 |
state.context_tokens = tokens
|
1016 |
-
|
1017 |
websocket._mrt = mrt
|
1018 |
websocket._state = state
|
1019 |
-
websocket._style = mrt
|
|
|
|
|
1020 |
websocket._rt_running = True
|
1021 |
websocket._rt_sr = sr
|
1022 |
websocket._rt_topk = int(params.get("topk", 40))
|
1023 |
websocket._rt_temp = float(params.get("temperature", 1.1))
|
1024 |
websocket._rt_guid = float(params.get("guidance_weight", 1.1))
|
|
|
1025 |
await send_json({"type":"started","mode":"rt"})
|
1026 |
# kick off a background task to stream ~2s chunks
|
1027 |
async def _rt_loop():
|
1028 |
try:
|
|
|
|
|
|
|
1029 |
while websocket._rt_running:
|
|
|
1030 |
mrt.guidance_weight = websocket._rt_guid
|
1031 |
-
mrt.temperature
|
1032 |
-
mrt.topk
|
|
|
|
|
1033 |
wav, new_state = mrt.generate_chunk(state=websocket._state, style=websocket._style)
|
1034 |
websocket._state = new_state
|
|
|
1035 |
x = wav.samples.astype(np.float32, copy=False)
|
1036 |
buf = io.BytesIO()
|
1037 |
sf.write(buf, x, mrt.sample_rate, subtype="FLOAT", format="WAV")
|
|
|
1038 |
if binary_audio:
|
1039 |
await websocket.send_bytes(buf.getvalue())
|
1040 |
await send_json({"type":"chunk_meta","metadata":{"sample_rate":mrt.sample_rate}})
|
1041 |
else:
|
1042 |
b64 = base64.b64encode(buf.getvalue()).decode("utf-8")
|
1043 |
await send_json({"type":"chunk","audio_base64":b64,
|
1044 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1045 |
except Exception as e:
|
1046 |
await send_json({"type":"error","error":str(e)})
|
1047 |
asyncio.create_task(_rt_loop())
|
@@ -1088,8 +1127,13 @@ async def ws_jam(websocket: WebSocket):
|
|
1088 |
websocket._rt_temp = float(msg.get("temperature", websocket._rt_temp))
|
1089 |
websocket._rt_topk = int(msg.get("topk", websocket._rt_topk))
|
1090 |
websocket._rt_guid = float(msg.get("guidance_weight", websocket._rt_guid))
|
1091 |
-
|
1092 |
-
|
|
|
|
|
|
|
|
|
|
|
1093 |
await send_json({"type":"status","updated":"rt-knobs"})
|
1094 |
|
1095 |
elif mtype == "consume" and mode == "bar":
|
|
|
23 |
|
24 |
|
25 |
import json, asyncio, base64
|
26 |
+
import time
|
27 |
|
28 |
# --- Patch T5X mesh helpers for GPUs on JAX >= 0.7 (coords present, no core_on_chip) ---
|
29 |
def _patch_t5x_for_gpu_coords():
|
|
|
903 |
"""
|
904 |
return Response(content=html_content, media_type="text/html")
|
905 |
|
906 |
+
|
907 |
+
# ----------------------------
|
908 |
+
# websockets route
|
909 |
+
# ----------------------------
|
910 |
+
|
911 |
+
|
912 |
+
|
913 |
+
def _combine_styles(mrt, styles_str: str = "", weights_str: str = ""):
|
914 |
+
extra = [s.strip() for s in (styles_str or "").split(",") if s.strip()]
|
915 |
+
if not extra:
|
916 |
+
return mrt.embed_style("warmup")
|
917 |
+
sw = [float(x) for x in (weights_str or "").split(",") if x.strip()]
|
918 |
+
embeds, weights = [], []
|
919 |
+
for i, s in enumerate(extra):
|
920 |
+
embeds.append(mrt.embed_style(s))
|
921 |
+
weights.append(sw[i] if i < len(sw) else 1.0)
|
922 |
+
wsum = sum(weights) or 1.0
|
923 |
+
weights = [w/wsum for w in weights]
|
924 |
+
import numpy as np
|
925 |
+
return np.sum([w*e for w, e in zip(weights, embeds)], axis=0).astype(np.float32)
|
926 |
+
|
927 |
@app.websocket("/ws/jam")
|
928 |
async def ws_jam(websocket: WebSocket):
|
929 |
await websocket.accept()
|
|
|
1026 |
# seed a fresh state with a silent context like warmup
|
1027 |
mrt = get_mrt()
|
1028 |
state = mrt.init_state()
|
|
|
1029 |
codec_fps = float(mrt.codec.frame_rate)
|
1030 |
ctx_seconds = float(mrt.config.context_length_frames) / codec_fps
|
1031 |
sr = int(mrt.sample_rate)
|
1032 |
samples = int(max(1, round(ctx_seconds * sr)))
|
1033 |
silent = au.Waveform(np.zeros((samples,2), np.float32), sr)
|
1034 |
+
tokens = mrt.codec.encode(silent).astype(np.int32)[:, :mrt.config.decoder_codec_rvq_depth]
|
|
|
1035 |
state.context_tokens = tokens
|
1036 |
+
|
1037 |
websocket._mrt = mrt
|
1038 |
websocket._state = state
|
1039 |
+
websocket._style = _combine_styles(mrt,
|
1040 |
+
params.get("styles","warmup"),
|
1041 |
+
params.get("style_weights",""))
|
1042 |
websocket._rt_running = True
|
1043 |
websocket._rt_sr = sr
|
1044 |
websocket._rt_topk = int(params.get("topk", 40))
|
1045 |
websocket._rt_temp = float(params.get("temperature", 1.1))
|
1046 |
websocket._rt_guid = float(params.get("guidance_weight", 1.1))
|
1047 |
+
websocket._pace = params.get("pace", "asap") # "realtime" | "asap"
|
1048 |
await send_json({"type":"started","mode":"rt"})
|
1049 |
# kick off a background task to stream ~2s chunks
|
1050 |
async def _rt_loop():
|
1051 |
try:
|
1052 |
+
mrt = websocket._mrt
|
1053 |
+
chunk_secs = (mrt.config.chunk_length_frames * mrt.config.frame_length_samples) / float(mrt.sample_rate)
|
1054 |
+
target_next = time.perf_counter()
|
1055 |
while websocket._rt_running:
|
1056 |
+
t0 = time.perf_counter()
|
1057 |
mrt.guidance_weight = websocket._rt_guid
|
1058 |
+
mrt.temperature = websocket._rt_temp
|
1059 |
+
mrt.topk = websocket._rt_topk
|
1060 |
+
|
1061 |
+
# style already in websocket._style
|
1062 |
wav, new_state = mrt.generate_chunk(state=websocket._state, style=websocket._style)
|
1063 |
websocket._state = new_state
|
1064 |
+
|
1065 |
x = wav.samples.astype(np.float32, copy=False)
|
1066 |
buf = io.BytesIO()
|
1067 |
sf.write(buf, x, mrt.sample_rate, subtype="FLOAT", format="WAV")
|
1068 |
+
|
1069 |
if binary_audio:
|
1070 |
await websocket.send_bytes(buf.getvalue())
|
1071 |
await send_json({"type":"chunk_meta","metadata":{"sample_rate":mrt.sample_rate}})
|
1072 |
else:
|
1073 |
b64 = base64.b64encode(buf.getvalue()).decode("utf-8")
|
1074 |
await send_json({"type":"chunk","audio_base64":b64,
|
1075 |
+
"metadata":{"sample_rate":mrt.sample_rate}})
|
1076 |
+
|
1077 |
+
# --- pacing ---
|
1078 |
+
if getattr(websocket, "_pace", "asap") == "realtime":
|
1079 |
+
t1 = time.perf_counter()
|
1080 |
+
target_next += chunk_secs
|
1081 |
+
sleep_s = max(0.0, target_next - t1 - 0.02) # tiny safety margin
|
1082 |
+
if sleep_s > 0:
|
1083 |
+
await asyncio.sleep(sleep_s)
|
1084 |
except Exception as e:
|
1085 |
await send_json({"type":"error","error":str(e)})
|
1086 |
asyncio.create_task(_rt_loop())
|
|
|
1127 |
websocket._rt_temp = float(msg.get("temperature", websocket._rt_temp))
|
1128 |
websocket._rt_topk = int(msg.get("topk", websocket._rt_topk))
|
1129 |
websocket._rt_guid = float(msg.get("guidance_weight", websocket._rt_guid))
|
1130 |
+
|
1131 |
+
if ("styles" in msg) or ("style_weights" in msg):
|
1132 |
+
websocket._style = _combine_styles(
|
1133 |
+
websocket._mrt,
|
1134 |
+
msg.get("styles", ""),
|
1135 |
+
msg.get("style_weights", "")
|
1136 |
+
)
|
1137 |
await send_json({"type":"status","updated":"rt-knobs"})
|
1138 |
|
1139 |
elif mtype == "consume" and mode == "bar":
|