File size: 8,648 Bytes
4689c2b | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 | import time
import threading
import torch
gen_lock = threading.Lock()
_MAIN_PROCESS_RUNNING_KEY = "main_process_running"
def get_gen_info(state):
cache = state.get("gen", None)
if cache == None:
cache = dict()
state["gen"] = cache
return cache
def _main_generation_active_locked(gen):
return bool(gen.get(_MAIN_PROCESS_RUNNING_KEY, False))
def set_main_generation_running(state, running):
gen = get_gen_info(state)
with gen_lock:
if running:
gen[_MAIN_PROCESS_RUNNING_KEY] = True
else:
gen.pop(_MAIN_PROCESS_RUNNING_KEY, None)
def any_GPU_process_running(state, process_id, ignore_main = False):
gen = get_gen_info(state)
#"process:" + process_id
with gen_lock:
process_status = gen.get("process_status", None)
if process_status == "process:main" and not _main_generation_active_locked(gen):
return False
return process_status is not None and not (process_status =="process:main" and ignore_main)
def _get_gpu_residents(gen):
residents = gen.get("gpu_residents", None)
if residents is None:
residents = {}
gen["gpu_residents"] = residents
return residents
def _drop_gpu_resident_locked(gen, process_id):
_get_gpu_residents(gen).pop(process_id, None)
def _collect_resident_release_actions_locked(gen, requester_id = None):
release_actions = []
residents = _get_gpu_residents(gen)
for resident_id, resident_info in list(residents.items()):
if resident_id == requester_id:
residents.pop(resident_id, None)
continue
if not bool(resident_info.get("force_release_on_acquire", False)):
continue
release_callback = resident_info.get("release_vram_callback", None)
if not callable(release_callback):
residents.pop(resident_id, None)
continue
release_actions.append((resident_id, resident_info.get("process_name", resident_id), release_callback))
residents.pop(resident_id, None)
return release_actions
def _run_release_actions(release_actions):
for resident_id, process_name, release_callback in release_actions:
try:
release_callback()
except Exception as exc:
print(f"[GPU] Unable to release resident VRAM for {process_name} ({resident_id}): {exc}")
if len(release_actions) > 0 and torch.cuda.is_available():
torch.cuda.synchronize()
def register_GPU_resident(state, process_id, process_name, release_vram_callback = None, force_release_on_acquire = True):
gen = get_gen_info(state)
with gen_lock:
_get_gpu_residents(gen)[process_id] = {
"process_name": process_name,
"release_vram_callback": release_vram_callback,
"force_release_on_acquire": bool(force_release_on_acquire),
}
def unregister_GPU_resident(state, process_id):
gen = get_gen_info(state)
with gen_lock:
_drop_gpu_resident_locked(gen, process_id)
def force_release_GPU_resident(state, process_id):
gen = get_gen_info(state)
release_callback = None
with gen_lock:
resident_info = _get_gpu_residents(gen).pop(process_id, None)
if resident_info is not None:
release_callback = resident_info.get("release_vram_callback", None)
if callable(release_callback):
release_callback()
if torch.cuda.is_available():
torch.cuda.synchronize()
def acquire_main_GPU_ressources(state):
gen = get_gen_info(state)
release_actions = []
while True:
with gen_lock:
process_status = gen.get("process_status", None)
if process_status is None or process_status == "process:main":
release_actions = _collect_resident_release_actions_locked(gen, requester_id="main")
gen["process_status"] = "process:main"
break
time.sleep(0.1)
_run_release_actions(release_actions)
if torch.cuda.is_available():
torch.cuda.synchronize()
def acquire_GPU_ressources(state, process_id, process_name, gr = None, custom_pause_msg = None, custom_wait_msg = None):
gen = get_gen_info(state)
original_process_status = None
release_actions = []
while True:
with gen_lock:
process_hierarchy = gen.get("process_hierarchy", None)
if process_hierarchy is None:
process_hierarchy = dict()
gen["process_hierarchy"]= process_hierarchy
process_status = gen.get("process_status", None)
if process_status is None:
_drop_gpu_resident_locked(gen, process_id)
original_process_status = None
release_actions = _collect_resident_release_actions_locked(gen, requester_id=process_id)
gen["process_status"] = "process:" + process_id
break
elif process_status == "request:" + process_id and not _main_generation_active_locked(gen):
_drop_gpu_resident_locked(gen, process_id)
original_process_status = None
release_actions = _collect_resident_release_actions_locked(gen, requester_id=process_id)
gen["process_status"] = "process:" + process_id
break
elif process_status == "process:main":
if not _main_generation_active_locked(gen):
_drop_gpu_resident_locked(gen, process_id)
original_process_status = None
release_actions = _collect_resident_release_actions_locked(gen, requester_id=process_id)
gen["process_status"] = "process:" + process_id
break
original_process_status = process_status
gen["process_status"] = "request:" + process_id
gen["pause_msg"] = custom_pause_msg if custom_pause_msg is not None else f"Generation Suspended while using {process_name}"
break
elif process_status == "process:" + process_id:
_drop_gpu_resident_locked(gen, process_id)
break
time.sleep(0.1)
_run_release_actions(release_actions)
if original_process_status is not None:
total_wait = 0
wait_time = 0.1
wait_msg_displayed = False
while True:
with gen_lock:
process_status = gen.get("process_status", None)
if process_status == "process:" + process_id:
break
if process_status is None or (process_status == "request:" + process_id and not _main_generation_active_locked(gen)):
# handle case when main process has finished at some point in between the last check and now
gen["process_status"] = "process:" + process_id
break
total_wait += wait_time
if round(total_wait,2) >= 5 and gr is not None and not wait_msg_displayed:
wait_msg_displayed = True
if custom_wait_msg is None:
gr.Info(f"Process {process_name} is Suspended while waiting that GPU Ressources become available")
else:
gr.Info(custom_wait_msg)
time.sleep(wait_time)
with gen_lock:
process_hierarchy[process_id] = original_process_status
if torch.cuda.is_available():
torch.cuda.synchronize()
def release_GPU_ressources(state, process_id, keep_resident = False, process_name = None, release_vram_callback = None, force_release_on_acquire = True):
gen = get_gen_info(state)
if torch.cuda.is_available():
torch.cuda.synchronize()
with gen_lock:
if keep_resident:
_get_gpu_residents(gen)[process_id] = {
"process_name": process_name or process_id,
"release_vram_callback": release_vram_callback,
"force_release_on_acquire": bool(force_release_on_acquire),
}
else:
_drop_gpu_resident_locked(gen, process_id)
process_hierarchy = gen.get("process_hierarchy", {})
restore_status = process_hierarchy.pop(process_id, None)
if restore_status == "process:main" and not _main_generation_active_locked(gen):
restore_status = None
gen["process_status"] = restore_status
|