Initial import: grid-bot — grid trading bot for BTC-USDT on Cifra Markets
This commit is contained in:
Kolp
@@ -0,0 +1,20 @@
|
||||
# Top level logging module for torch logging
|
||||
# Design doc: https://docs.google.com/document/d/1ZRfTWKa8eaPq1AxaiHrq4ASTPouzzlPiuquSBEJYwS8/edit#
|
||||
# Simple setup for onboarding (see above doc for more detail):
|
||||
# 1. register any top-level log qualified name for your module in torch._logging._registrations (see there for examples)
|
||||
# 2. register any artifacts (<artifact_name> below) in torch._logging._registrations
|
||||
# a. call getArtifactLogger(__name__, <artifact_name>) at your logging site instead of the standard logger to log your artifact
|
||||
import torch._logging._registrations
|
||||
|
||||
from ._internal import (
|
||||
_init_logs,
|
||||
DEFAULT_LOGGING,
|
||||
dtrace_structured,
|
||||
get_structured_logging_overhead,
|
||||
getArtifactLogger,
|
||||
hide_warnings,
|
||||
LazyString,
|
||||
set_logs,
|
||||
trace_structured,
|
||||
warning_once,
|
||||
)
|
||||
File diff suppressed because it is too large
Load Diff
@@ -0,0 +1,280 @@
|
||||
# flake8: noqa: B950
|
||||
from ._internal import register_artifact, register_log
|
||||
|
||||
|
||||
DYNAMIC = [
|
||||
"torch.fx.experimental.symbolic_shapes",
|
||||
"torch.fx.experimental.sym_node",
|
||||
"torch.fx.experimental.recording",
|
||||
]
|
||||
DISTRIBUTED = [
|
||||
"torch.distributed",
|
||||
"torch._dynamo.backends.distributed",
|
||||
"torch.nn.parallel.distributed",
|
||||
]
|
||||
|
||||
register_log(
|
||||
"async_compile",
|
||||
[
|
||||
"torch._inductor.async_compile",
|
||||
"torch._inductor.compile_worker.tracked_process_pool",
|
||||
],
|
||||
)
|
||||
register_log(
|
||||
"cache", ("torch._inductor.remote_cache", "torch._inductor.fb.remote_cache")
|
||||
)
|
||||
register_log("dynamo", ["torch._dynamo", *DYNAMIC])
|
||||
register_log("fake_tensor", ["torch._subclasses.fake_tensor"])
|
||||
register_log("aot", ["torch._functorch.aot_autograd", "torch._functorch._aot_autograd"])
|
||||
register_log("autograd", "torch.autograd")
|
||||
register_log("inductor", ["torch._inductor", "torch._inductor.cudagraph_trees"])
|
||||
|
||||
register_artifact(
|
||||
"cudagraphs",
|
||||
"Logs information from wrapping inductor generated code with cudagraphs.",
|
||||
)
|
||||
|
||||
register_log("dynamic", DYNAMIC)
|
||||
register_log("torch", "torch")
|
||||
register_log("distributed", DISTRIBUTED)
|
||||
register_log(
|
||||
"c10d", ["torch.distributed.distributed_c10d", "torch.distributed.rendezvous"]
|
||||
)
|
||||
register_log(
|
||||
"ddp", ["torch.nn.parallel.distributed", "torch._dynamo.backends.distributed"]
|
||||
)
|
||||
register_log("pp", ["torch.distributed.pipelining"])
|
||||
register_log("fsdp", ["torch.distributed.fsdp", "torch.distributed._composable.fsdp"])
|
||||
register_log("dtensor", ["torch.distributed._tensor", "torch.distributed.tensor"])
|
||||
register_log("onnx", "torch.onnx")
|
||||
register_log(
|
||||
"export",
|
||||
[
|
||||
"torch._dynamo",
|
||||
"torch.export",
|
||||
"torch.export.dynamic_shapes",
|
||||
*DYNAMIC,
|
||||
"torch._export.converter",
|
||||
"torch._export.non_strict_utils",
|
||||
"torch._export.serde.serialize",
|
||||
"torch.fx.experimental.proxy_tensor",
|
||||
],
|
||||
)
|
||||
|
||||
register_artifact(
|
||||
"guards",
|
||||
"This prints the guards for every compiled Dynamo frame. It does not tell you where the guards come from.",
|
||||
visible=True,
|
||||
)
|
||||
register_artifact("verbose_guards", "", off_by_default=True)
|
||||
register_artifact(
|
||||
"bytecode",
|
||||
"Prints the original and modified bytecode from Dynamo. Mostly useful if you're debugging our bytecode generation in Dynamo.",
|
||||
off_by_default=True,
|
||||
)
|
||||
register_artifact(
|
||||
"graph",
|
||||
"Prints the dynamo traced graph (prior to AOTDispatch) in a table. If you prefer python code use `graph_code` instead. ",
|
||||
)
|
||||
register_artifact("graph_code", "Like `graph`, but gives you the Python code instead.")
|
||||
register_artifact(
|
||||
"graph_code_verbose",
|
||||
"Verbose FX pass logs, e.g. from tensorify_python_scalars and runtime_assert.",
|
||||
)
|
||||
register_artifact(
|
||||
"graph_sizes", "Prints the sizes of all FX nodes in the dynamo graph."
|
||||
)
|
||||
register_artifact(
|
||||
"trace_source",
|
||||
"As we execute bytecode, prints the file name / line number we are processing and the actual source code. Useful with `bytecode`",
|
||||
)
|
||||
register_artifact(
|
||||
"trace_call",
|
||||
"Like trace_source, but it will give you the per-expression blow-by-blow if your Python is recent enough.",
|
||||
)
|
||||
register_artifact(
|
||||
"trace_bytecode",
|
||||
"As we trace bytecode, prints the instruction and the current stack.",
|
||||
)
|
||||
register_artifact(
|
||||
"aot_graphs",
|
||||
"Prints the FX forward and backward graph generated by AOTDispatch, after partitioning. Useful to understand what's being given to Inductor",
|
||||
visible=True,
|
||||
)
|
||||
register_artifact(
|
||||
"aot_joint_graph",
|
||||
"Print FX joint graph from AOTAutograd, prior to partitioning. Useful for debugging partitioning",
|
||||
)
|
||||
register_artifact(
|
||||
"aot_graphs_effects",
|
||||
"Prints the FX forward and backward graph generated by AOTDispatch, useful for debugging effects processing.",
|
||||
visible=True,
|
||||
)
|
||||
register_artifact(
|
||||
"pre_grad_graphs",
|
||||
"Prints the FX graph before inductor pre grad passes. Useful to understand what's being given to Inductor before grad passes",
|
||||
)
|
||||
register_artifact(
|
||||
"post_grad_graphs",
|
||||
"Prints the FX graph generated by post grad passes. Useful to understand what's being given to Inductor after post grad passes",
|
||||
)
|
||||
register_artifact(
|
||||
"ir_pre_fusion",
|
||||
"Prints the IR before inductor fusion passes.",
|
||||
off_by_default=True,
|
||||
)
|
||||
register_artifact(
|
||||
"ir_post_fusion",
|
||||
"Prints the IR after inductor fusion passes.",
|
||||
off_by_default=True,
|
||||
)
|
||||
register_artifact(
|
||||
"compiled_autograd",
|
||||
"Prints various logs in compiled_autograd, including but not limited to the graphs. Useful for debugging compiled_autograd.",
|
||||
visible=True,
|
||||
)
|
||||
register_artifact(
|
||||
"compiled_autograd_verbose",
|
||||
"Will affect performance. Prints compiled_autograd logs with C++ info e.g. autograd node -> fx node mapping",
|
||||
off_by_default=True,
|
||||
)
|
||||
register_artifact(
|
||||
"ddp_graphs",
|
||||
"Only relevant for compiling DDP. DDP splits into multiple graphs to trigger comms early. This will print each individual graph here.",
|
||||
)
|
||||
register_artifact(
|
||||
"recompiles",
|
||||
"Prints the reason why we recompiled a graph. Very, very useful.",
|
||||
visible=True,
|
||||
)
|
||||
register_artifact(
|
||||
"recompiles_verbose",
|
||||
"Prints all guard checks that fail during a recompilation. "
|
||||
"At runtime, Dynamo will stop at the first failed check for each failing guard. "
|
||||
"So not all logged failing checks are actually ran by Dynamo.",
|
||||
visible=True,
|
||||
off_by_default=True,
|
||||
)
|
||||
register_artifact(
|
||||
"graph_breaks",
|
||||
"Prints whenever Dynamo decides that it needs to graph break (i.e. create a new graph). Useful for debugging why torch.compile has poor performance",
|
||||
visible=True,
|
||||
)
|
||||
register_artifact(
|
||||
"side_effects",
|
||||
"Prints all side effects that Dynamo codegenerates, including mutations to variables, attributes, cells, and globals. Useful for debugging side effect handling",
|
||||
visible=True,
|
||||
)
|
||||
register_artifact(
|
||||
"not_implemented",
|
||||
"Prints log messages whenever we return NotImplemented in a multi-dispatch, letting you trace through each object we attempted to dispatch to",
|
||||
)
|
||||
register_artifact(
|
||||
"output_code",
|
||||
"Prints the code that Inductor generates (either Triton or C++)",
|
||||
off_by_default=True,
|
||||
visible=True,
|
||||
)
|
||||
register_artifact(
|
||||
"kernel_code",
|
||||
"Prints the code that Inductor generates (on a per-kernel basis)",
|
||||
off_by_default=True,
|
||||
visible=True,
|
||||
)
|
||||
register_artifact(
|
||||
"schedule",
|
||||
"Inductor scheduler information. Useful if working on Inductor fusion algo",
|
||||
off_by_default=True,
|
||||
)
|
||||
register_artifact("perf_hints", "", off_by_default=True)
|
||||
register_artifact("onnx_diagnostics", "", off_by_default=True)
|
||||
register_artifact("compute_dependencies", "", off_by_default=True)
|
||||
register_artifact(
|
||||
"fusion",
|
||||
"Detailed Inductor fusion decisions. More detailed than 'schedule'",
|
||||
off_by_default=True,
|
||||
)
|
||||
register_artifact(
|
||||
"loop_ordering",
|
||||
"Logs related to loop ordering",
|
||||
off_by_default=True,
|
||||
)
|
||||
register_artifact(
|
||||
"loop_tiling",
|
||||
"Logs related to loop ordering",
|
||||
off_by_default=True,
|
||||
)
|
||||
|
||||
register_artifact(
|
||||
"auto_chunker",
|
||||
"Logs related to the auto chunker",
|
||||
off_by_default=True,
|
||||
)
|
||||
|
||||
register_artifact(
|
||||
"overlap",
|
||||
"Detailed Inductor compute/comm overlap decisions",
|
||||
off_by_default=True,
|
||||
)
|
||||
register_artifact(
|
||||
"overlap_scheduling",
|
||||
"Detailed Inductor overlap scheduling pass information",
|
||||
off_by_default=True,
|
||||
)
|
||||
register_artifact(
|
||||
"sym_node",
|
||||
"Logs extra info for various SymNode operations",
|
||||
off_by_default=True,
|
||||
)
|
||||
register_artifact(
|
||||
"trace_shape_events",
|
||||
"Logs traces for every ShapeEnv operation that we record for replay",
|
||||
off_by_default=True,
|
||||
)
|
||||
register_artifact(
|
||||
"cudagraph_static_inputs",
|
||||
"Logs static inputs handling in dynamo, AOT, and cudagraphs",
|
||||
off_by_default=True,
|
||||
)
|
||||
register_artifact(
|
||||
"benchmarking",
|
||||
"Detailed Inductor benchmarking information.",
|
||||
off_by_default=True,
|
||||
)
|
||||
register_artifact(
|
||||
"node_runtime_estimation",
|
||||
"Node runtime estimation for compile-time optimization decisions.",
|
||||
off_by_default=True,
|
||||
)
|
||||
register_artifact(
|
||||
"autotuning",
|
||||
"Autotuning choice logs, such as kernel source, perf, and tuning parameters.",
|
||||
off_by_default=True,
|
||||
)
|
||||
register_artifact(
|
||||
"graph_region_expansion",
|
||||
"Logs detailed steps of the duplicate graph region tracker expansion algorithm",
|
||||
off_by_default=True,
|
||||
)
|
||||
|
||||
register_artifact(
|
||||
"inductor_metrics",
|
||||
"Logs Inductor metrics, such as num_bytes, nodes_num_elem, node_runtimes",
|
||||
off_by_default=True,
|
||||
)
|
||||
register_artifact(
|
||||
"hierarchical_compile",
|
||||
"Logs debug info for hierarchical compilation",
|
||||
off_by_default=True,
|
||||
)
|
||||
register_artifact(
|
||||
"annotation",
|
||||
"Logs detailed steps of the creating annotation on graph nodes",
|
||||
off_by_default=True,
|
||||
)
|
||||
register_artifact("custom_format_test_artifact", "Testing only", log_format="")
|
||||
register_artifact(
|
||||
"caching",
|
||||
"Detailed Inductor caching information.",
|
||||
off_by_default=True,
|
||||
)
|
||||
@@ -0,0 +1,63 @@
|
||||
from collections.abc import Callable
|
||||
from typing import TypeAlias
|
||||
|
||||
|
||||
try:
|
||||
from fbscribelogger import ( # type: ignore[import-untyped, import-not-found]
|
||||
make_scribe_logger,
|
||||
)
|
||||
except ImportError:
|
||||
TAtom: TypeAlias = int | float | bool | str
|
||||
TField: TypeAlias = TAtom | list[TAtom]
|
||||
TLazyField: TypeAlias = TField | Callable[[], TField]
|
||||
|
||||
def make_scribe_logger(name: str, thrift_src: str) -> Callable[..., None]:
|
||||
def inner(**kwargs: TLazyField) -> None:
|
||||
pass
|
||||
|
||||
return inner
|
||||
|
||||
|
||||
open_source_signpost = make_scribe_logger(
|
||||
"TorchOpenSourceSignpost",
|
||||
"""
|
||||
struct TorchOpenSourceSignpostLogEntry {
|
||||
|
||||
# The commit SHA that triggered the workflow, e.g., 02a6b1d30f338206a71d0b75bfa09d85fac0028a. Derived from GITHUB_SHA.
|
||||
4: optional string commit_sha;
|
||||
|
||||
# Commit date (not author date) of the commit in commit_sha as timestamp, e.g., 1724208105. Increasing if merge bot is used, though not monotonic; duplicates occur when stack is landed.
|
||||
5: optional i64 commit_date;
|
||||
|
||||
# The fully-formed ref of the branch or tag that triggered the workflow run, e.g., refs/pull/133891/merge or refs/heads/main. Derived from GITHUB_REF.
|
||||
6: optional string github_ref;
|
||||
|
||||
# Indicates if branch protections or rulesets are configured for the ref that triggered the workflow run. Derived from GITHUB_REF_PROTECTED.
|
||||
7: optional bool github_ref_protected;
|
||||
|
||||
# A unique number for each attempt of a particular workflow run in a repository, e.g., 1. Derived from GITHUB_RUN_ATTEMPT.
|
||||
8: optional string github_run_attempt;
|
||||
|
||||
# A unique number for each workflow run within a repository, e.g., 19471190684. Derived from GITHUB_RUN_ID.
|
||||
9: optional string github_run_id;
|
||||
|
||||
# A unique number for each run of a particular workflow in a repository, e.g., 238742. Derived from GITHUB_RUN_NUMBER.
|
||||
10: optional string github_run_number_str;
|
||||
|
||||
# The name of the current job. Derived from JOB_NAME, e.g., linux-jammy-py3.8-gcc11 / test (default, 3, 4, linux.2xlarge).
|
||||
11: optional string job_name;
|
||||
|
||||
# The GitHub user who triggered the job. Derived from GITHUB_TRIGGERING_ACTOR.
|
||||
12: optional string github_triggering_actor;
|
||||
13: optional string name; # Event name
|
||||
14: optional string parameters; # Parameters (JSON data)
|
||||
16: optional string subsystem; # Subsystem the event is associated with
|
||||
|
||||
# The unit timestamp in second for the Scuba Time Column override
|
||||
17: optional i64 time;
|
||||
|
||||
# The weight of the record according to current sampling rate
|
||||
18: optional i64 weight;
|
||||
}
|
||||
""", # noqa: B950
|
||||
)
|
||||
@@ -0,0 +1,109 @@
|
||||
"""
|
||||
Utilities for converting data types into structured JSON for dumping.
|
||||
"""
|
||||
|
||||
import inspect
|
||||
import os
|
||||
import traceback
|
||||
from collections.abc import Sequence
|
||||
from typing import Any
|
||||
|
||||
import torch._logging._internal
|
||||
|
||||
|
||||
INTERN_TABLE: dict[str, int] = {}
|
||||
|
||||
|
||||
DUMPED_FILES: set[str] = set()
|
||||
|
||||
|
||||
def intern_string(s: str | None) -> int:
|
||||
if s is None:
|
||||
return -1
|
||||
|
||||
r = INTERN_TABLE.get(s)
|
||||
if r is None:
|
||||
r = len(INTERN_TABLE)
|
||||
INTERN_TABLE[s] = r
|
||||
torch._logging._internal.trace_structured(
|
||||
"str", lambda: (s, r), suppress_context=True
|
||||
)
|
||||
return r
|
||||
|
||||
|
||||
def dump_file(filename: str) -> None:
|
||||
if "eval_with_key" not in filename:
|
||||
return
|
||||
if filename in DUMPED_FILES:
|
||||
return
|
||||
DUMPED_FILES.add(filename)
|
||||
from torch.fx.graph_module import _loader
|
||||
|
||||
torch._logging._internal.trace_structured(
|
||||
"dump_file",
|
||||
metadata_fn=lambda: {
|
||||
"name": filename,
|
||||
},
|
||||
payload_fn=lambda: _loader.get_source(filename),
|
||||
)
|
||||
|
||||
|
||||
def from_traceback(tb: Sequence[traceback.FrameSummary]) -> list[dict[str, Any]]:
|
||||
# dict naming convention here coincides with
|
||||
# python/combined_traceback.cpp
|
||||
r = [
|
||||
{
|
||||
"line": frame.lineno,
|
||||
"name": frame.name,
|
||||
"filename": intern_string(frame.filename),
|
||||
"loc": frame.line,
|
||||
}
|
||||
for frame in tb
|
||||
]
|
||||
return r
|
||||
|
||||
|
||||
def get_user_stack(num_frames: int) -> list[dict[str, Any]]:
|
||||
from torch._guards import TracingContext
|
||||
from torch.utils._traceback import CapturedTraceback
|
||||
|
||||
user_tb = TracingContext.extract_stack()
|
||||
if user_tb:
|
||||
return from_traceback(user_tb[-1 * num_frames :])
|
||||
|
||||
tb = CapturedTraceback.extract().summary()
|
||||
|
||||
# Filter out frames that are within the torch/ codebase
|
||||
torch_filepath = os.path.dirname(inspect.getfile(torch)) + os.path.sep
|
||||
for i, frame in enumerate(reversed(tb)):
|
||||
if torch_filepath not in frame.filename:
|
||||
# Only display `num_frames` frames in the traceback
|
||||
filtered_tb = tb[len(tb) - i - num_frames : len(tb) - i]
|
||||
return from_traceback(filtered_tb)
|
||||
|
||||
return from_traceback(tb[-1 * num_frames :])
|
||||
|
||||
|
||||
def get_framework_stack(
|
||||
num_frames: int = 25, cpp: bool = False
|
||||
) -> list[dict[str, Any]]:
|
||||
"""
|
||||
Returns the traceback for the user stack and the framework stack
|
||||
"""
|
||||
from torch.fx.experimental.symbolic_shapes import uninteresting_files
|
||||
from torch.utils._traceback import CapturedTraceback
|
||||
|
||||
tb = CapturedTraceback.extract(cpp=cpp).summary()
|
||||
tb = [
|
||||
frame
|
||||
for frame in tb
|
||||
if (
|
||||
(
|
||||
frame.filename.endswith(".py")
|
||||
and frame.filename not in uninteresting_files()
|
||||
)
|
||||
or ("at::" in frame.name or "torch::" in frame.name)
|
||||
)
|
||||
]
|
||||
|
||||
return from_traceback(tb[-1 * num_frames :])
|
||||
Reference in New Issue
Block a user