File size: 14,985 Bytes
714cf46 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 | import base64
import json
import os
import subprocess
import sys
import time
from pathlib import Path
def _is_json_scalar(value):
return value is None or isinstance(value, (bool, int, float, str))
def _to_json_safe(value):
if _is_json_scalar(value):
return value
if isinstance(value, list):
return [_to_json_safe(item) for item in value]
if isinstance(value, tuple):
return [_to_json_safe(item) for item in value]
if isinstance(value, dict):
converted = {}
for key in value:
converted[key] = _to_json_safe(value[key])
return converted
if isinstance(value, Path):
return str(value)
return str(value)
def _should_auto_run_modal(args):
if "PROTIFY_JOB_ID" in os.environ and os.environ["PROTIFY_JOB_ID"] != "":
return False
if args.replay_path is not None:
return False
if not args.modal_cli_credentials_provided:
return False
return args.modal_token_id is not None and args.modal_token_secret is not None
def _modal_subprocess_env(args):
env = os.environ.copy()
env["MODAL_TOKEN_ID"] = args.modal_token_id
env["MODAL_TOKEN_SECRET"] = args.modal_token_secret
env["PYTHONIOENCODING"] = "utf-8"
env["PYTHONUTF8"] = "1"
return env
def _repo_root():
return Path(__file__).resolve().parents[2]
def _deploy_modal_backend(args):
repo_root = _repo_root()
backend_path = repo_root / "src" / "protify" / "modal_backend.py"
assert backend_path.exists(), f"Modal backend not found at {backend_path}"
app_name = "protify-backend"
env = _modal_subprocess_env(args)
primary_command = [sys.executable, "-m", "modal", "deploy", str(backend_path), "--name", app_name]
try:
process = subprocess.run(
primary_command,
cwd=str(repo_root),
env=env,
capture_output=True,
text=True,
encoding="utf-8",
errors="replace",
)
except FileNotFoundError:
fallback_command = ["modal", "deploy", str(backend_path), "--name", app_name]
process = subprocess.run(
fallback_command,
cwd=str(repo_root),
env=env,
capture_output=True,
text=True,
encoding="utf-8",
errors="replace",
)
if process.returncode != 0:
stderr_text = process.stderr if process.stderr is not None else ""
stdout_text = process.stdout if process.stdout is not None else ""
combined_output = f"{stdout_text}\n{stderr_text}".strip()
if "No module named modal" in combined_output:
raise RuntimeError("Modal is not installed in this Python environment. Install it with: py -m pip install modal")
raise RuntimeError(f"Modal deploy failed:\n{combined_output}")
stdout_text = process.stdout if process.stdout is not None else ""
if stdout_text:
print(stdout_text[-4000:])
def _build_modal_config_from_args(args):
config = {}
excluded_keys = {
"modal_token_id",
"modal_token_secret",
"modal_api_key",
"modal_cli_credentials_provided",
"rebuild_modal",
"delete_modal_embeddings",
}
for key in args.__dict__:
if key in excluded_keys:
continue
config[key] = _to_json_safe(args.__dict__[key])
config["replay_path"] = None
return config
def _save_modal_artifacts(result_payload, output_root, job_id):
output_root_path = Path(output_root)
job_dir = output_root_path / job_id
job_dir.mkdir(parents=True, exist_ok=True)
files_payload = result_payload["files"] if "files" in result_payload else {}
for rel_path in files_payload:
local_path = job_dir / Path(rel_path)
local_path.parent.mkdir(parents=True, exist_ok=True)
with open(local_path, "w", encoding="utf-8") as file:
file.write(files_payload[rel_path])
images_payload = result_payload["images"] if "images" in result_payload else {}
for rel_path in images_payload:
image_info = images_payload[rel_path]
if "data" not in image_info:
continue
local_path = job_dir / Path(rel_path)
local_path.parent.mkdir(parents=True, exist_ok=True)
image_bytes = base64.b64decode(image_info["data"])
with open(local_path, "wb") as file:
file.write(image_bytes)
summary_path = job_dir / "modal_fetch_summary.json"
with open(summary_path, "w", encoding="utf-8") as file:
json.dump(result_payload, file, indent=2)
return str(job_dir)
def _coerce_modal_terminal_payload(remote_result):
if isinstance(remote_result, dict):
payload = dict(remote_result)
if "status" not in payload:
if "success" in payload and payload["success"]:
payload["status"] = "SUCCESS"
elif "success" in payload and not payload["success"]:
payload["status"] = "FAILED"
else:
payload["status"] = "SUCCESS"
return payload
return {"status": "SUCCESS"}
def _run_on_modal_cli(args):
try:
import modal
except Exception as error:
raise RuntimeError("Modal SDK is required for CLI remote execution. Install with: py -m pip install modal") from error
app_name = "protify-backend"
gpu_type = "A10"
if "modal_gpu_type" in args.__dict__ and args.modal_gpu_type is not None:
gpu_type = args.modal_gpu_type
timeout_seconds = 86400
if "modal_timeout_seconds" in args.__dict__ and args.modal_timeout_seconds is not None:
timeout_seconds = args.modal_timeout_seconds
poll_interval_seconds = 5
if "modal_poll_interval_seconds" in args.__dict__ and args.modal_poll_interval_seconds is not None:
poll_interval_seconds = args.modal_poll_interval_seconds
log_tail_chars = 5000
if "modal_log_tail_chars" in args.__dict__ and args.modal_log_tail_chars is not None:
log_tail_chars = args.modal_log_tail_chars
max_stale_heartbeat_seconds = 600
if "modal_max_stale_heartbeat_seconds" in args.__dict__ and args.modal_max_stale_heartbeat_seconds is not None:
max_stale_heartbeat_seconds = args.modal_max_stale_heartbeat_seconds
artifacts_root = "modal_artifacts"
if "modal_artifacts_dir" in args.__dict__ and args.modal_artifacts_dir is not None:
artifacts_root = args.modal_artifacts_dir
if args.rebuild_modal:
print("Rebuilding Modal backend due to --rebuild_modal ...")
_deploy_modal_backend(args)
config = _build_modal_config_from_args(args)
submit_fn = modal.Function.from_name(app_name, "submit_protify_job")
status_fn = modal.Function.from_name(app_name, "get_job_status")
log_delta_fn = modal.Function.from_name(app_name, "get_job_log_delta")
results_fn = modal.Function.from_name(app_name, "get_results")
delete_embeddings_fn = modal.Function.from_name(app_name, "delete_modal_embeddings")
if args.delete_modal_embeddings:
print("Deleting Modal embedding cache due to --delete_modal_embeddings ...")
try:
delete_embeddings_payload = delete_embeddings_fn.remote()
except Exception:
print("Modal embedding delete failed before app/function lookup succeeded; attempting deploy then retry...")
_deploy_modal_backend(args)
submit_fn = modal.Function.from_name(app_name, "submit_protify_job")
status_fn = modal.Function.from_name(app_name, "get_job_status")
log_delta_fn = modal.Function.from_name(app_name, "get_job_log_delta")
results_fn = modal.Function.from_name(app_name, "get_results")
delete_embeddings_fn = modal.Function.from_name(app_name, "delete_modal_embeddings")
delete_embeddings_payload = delete_embeddings_fn.remote()
if isinstance(delete_embeddings_payload, dict) and "message" in delete_embeddings_payload:
print(delete_embeddings_payload["message"])
has_dataset_run = len(args.data_names) > 0 or len(args.data_dirs) > 0
if not has_dataset_run and not args.proteingym:
return 0
try:
submit_result = submit_fn.remote(
config=config,
gpu_type=gpu_type,
hf_token=args.hf_token,
wandb_api_key=args.wandb_api_key,
synthyra_api_key=args.synthyra_api_key,
timeout_seconds=timeout_seconds,
)
except Exception:
print("Modal submit failed before app/function lookup succeeded; attempting deploy then retry...")
_deploy_modal_backend(args)
submit_fn = modal.Function.from_name(app_name, "submit_protify_job")
status_fn = modal.Function.from_name(app_name, "get_job_status")
log_delta_fn = modal.Function.from_name(app_name, "get_job_log_delta")
results_fn = modal.Function.from_name(app_name, "get_results")
submit_result = submit_fn.remote(
config=config,
gpu_type=gpu_type,
hf_token=args.hf_token,
wandb_api_key=args.wandb_api_key,
synthyra_api_key=args.synthyra_api_key,
timeout_seconds=timeout_seconds,
)
assert isinstance(submit_result, dict), "Modal submit response is not a dictionary."
assert "job_id" in submit_result, "Modal submit response missing job_id."
job_id = submit_result["job_id"]
function_call_id = submit_result["function_call_id"] if "function_call_id" in submit_result else None
print(f"Modal job submitted: {job_id}")
if function_call_id is not None:
print(f"Modal function call id: {function_call_id}")
terminal_states = {"SUCCESS", "FAILED", "TERMINATED", "TIMEOUT"}
final_status_payload = None
poll_start_time = time.time()
max_poll_seconds = int(timeout_seconds) + 900
status_print_interval_seconds = 15
last_status_print_time = 0.0
last_status_line = ""
missing_status_count = 0
log_offset = 0
function_call = None
if function_call_id is not None:
function_call = modal.FunctionCall.from_id(function_call_id)
def _emit_remote_logs():
nonlocal log_offset
delta_payload = log_delta_fn.remote(job_id=job_id, offset=log_offset, max_chars=log_tail_chars)
if isinstance(delta_payload, dict):
if "next_offset" in delta_payload and isinstance(delta_payload["next_offset"], int):
log_offset = delta_payload["next_offset"]
if "chunk" in delta_payload and delta_payload["chunk"]:
sys.stdout.write(delta_payload["chunk"])
sys.stdout.flush()
while True:
_emit_remote_logs()
status_payload = status_fn.remote(job_id=job_id)
assert isinstance(status_payload, dict), "Modal status response is not a dictionary."
if "success" in status_payload and status_payload["success"]:
missing_status_count = 0
status_value = status_payload["status"] if "status" in status_payload else "UNKNOWN"
phase_value = status_payload["phase"] if "phase" in status_payload else "N/A"
heartbeat_age = status_payload["heartbeat_age_seconds"] if "heartbeat_age_seconds" in status_payload else None
heartbeat_text = "N/A" if heartbeat_age is None else f"{heartbeat_age:.1f}s"
status_line = f"[Modal] status={status_value} phase={phase_value} heartbeat_age={heartbeat_text}"
if status_value in terminal_states:
final_status_payload = dict(status_payload)
break
else:
missing_status_count += 1
status_line = "[Modal] state=queued_or_initializing"
if missing_status_count % 6 == 0 and "error" in status_payload and status_payload["error"]:
status_line = f"[Modal] state=queued_or_initializing detail={status_payload['error']}"
now = time.time()
if status_line != last_status_line or (now - last_status_print_time) >= status_print_interval_seconds:
print(status_line)
last_status_line = status_line
last_status_print_time = now
if function_call is not None:
try:
remote_result = function_call.get(timeout=0)
final_status_payload = _coerce_modal_terminal_payload(remote_result)
if "phase" not in final_status_payload and "phase" in status_payload:
final_status_payload["phase"] = status_payload["phase"]
break
except TimeoutError:
pass
except Exception as error:
final_status_payload = {"status": "FAILED", "error": f"Function call failed: {error}"}
break
elapsed_seconds = now - poll_start_time
if elapsed_seconds > max_poll_seconds:
final_status_payload = {
"status": "TIMEOUT",
"phase": "poll_timeout",
"error": f"Polling exceeded timeout window ({max_poll_seconds} seconds).",
}
break
if "success" in status_payload and status_payload["success"] and "heartbeat_age_seconds" in status_payload:
heartbeat_age = status_payload["heartbeat_age_seconds"]
if heartbeat_age is not None and heartbeat_age > max_stale_heartbeat_seconds and function_call is None:
final_status_payload = {
"status": "FAILED",
"phase": "stale_heartbeat",
"error": f"Heartbeat stale for {heartbeat_age:.1f}s with no function_call_id available.",
}
break
time.sleep(max(1, int(poll_interval_seconds)))
final_delta_payload = log_delta_fn.remote(job_id=job_id, offset=log_offset, max_chars=log_tail_chars * 8)
if isinstance(final_delta_payload, dict):
if "chunk" in final_delta_payload and final_delta_payload["chunk"]:
sys.stdout.write(final_delta_payload["chunk"])
sys.stdout.flush()
try:
results_payload = results_fn.remote(job_id=job_id)
except Exception as error:
results_payload = {"success": False, "error": str(error)}
if isinstance(results_payload, dict) and "success" in results_payload and results_payload["success"]:
artifacts_dir = _save_modal_artifacts(results_payload, artifacts_root, job_id)
print(f"Modal artifacts saved to {artifacts_dir}")
if final_status_payload is None:
final_status_payload = {"status": "FAILED", "error": "No terminal status was resolved."}
final_status = final_status_payload["status"] if "status" in final_status_payload else "FAILED"
if final_status != "SUCCESS":
if "error" in final_status_payload and final_status_payload["error"]:
print(f"Modal job failed: {final_status_payload['error']}")
return 1
return 0
|