K-Dense-AI/scientific-agent-skills / skills/pathml
pathml
VerifiedSupports local computational pathology research with PathML: slide loading and tiling, preprocessing and QC, h5path storage, multiplex quantification, spatial graphs, and bounded model inference. Use for whole-slide H&E, CODEX, Vectra, Mesmer, HoVer-Net, and HACTNet workflows.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 12.8 KB |
| references/data_management.md | markdown | 13.2 KB |
| references/graphs.md | markdown | 11.3 KB |
| references/image_loading.md | markdown | 11.3 KB |
| references/machine_learning.md | markdown | 13.7 KB |
| references/multiparametric.md | markdown | 13.6 KB |
| references/preprocessing.md | markdown | 12.0 KB |
| scripts/_common.py | script | 12.4 KB |
| scripts/image_qc.py | script | 11.0 KB |
| scripts/plan_inference.py | script | 9.5 KB |
| scripts/plan_pipeline.py | script | 11.1 KB |
| scripts/slide_manifest.py | script | 14.2 KB |
| scripts/validate_spatial_schema.py | script | 16.4 KB |
scripts/plan_inference.py
View on GitHub#!/usr/bin/env python3
"""Plan bounded inference batches from numbers or a JSON model card only."""
from __future__ import annotations
import argparse
import math
import re
from pathlib import Path
from typing import Any
from _common import (
CliError,
MAX_JSON_BYTES,
emit_json,
finite_float,
load_json_object,
run_cli,
validate_keys,
)
DTYPE_BYTES = {
"uint8": 1,
"int8": 1,
"float16": 2,
"bfloat16": 2,
"float32": 4,
"float64": 8,
}
MODEL_SUFFIXES = {
".bin",
".ckpt",
".joblib",
".onnx",
".pickle",
".pkl",
".pt",
".pth",
".safetensors",
}
SHA256_PATTERN = re.compile(r"^[0-9a-fA-F]{64}$")
def _load_card(path: str | None, root: str) -> dict[str, Any]:
if path is None:
return {}
if Path(path).suffix.lower() in MODEL_SUFFIXES:
raise CliError(
"--model-card accepts JSON metadata only; model/checkpoint files "
"are intentionally never loaded"
)
card = load_json_object(path, root=root, max_bytes=MAX_JSON_BYTES)
validate_keys(
card,
allowed={
"schema_version",
"model_id",
"artifact_sha256",
"input_shape",
"dtype",
"output_elements_per_tile",
"activation_multiplier",
"parameter_bytes",
"runtime_workspace_mib",
},
required={"schema_version", "model_id", "input_shape", "dtype"},
context="model card",
)
if card["schema_version"] != "1.0":
raise CliError("model card schema_version must be '1.0'")
if not isinstance(card["model_id"], str) or not 1 <= len(card["model_id"]) <= 128:
raise CliError("model_id must be a nonempty string of at most 128 characters")
digest = card.get("artifact_sha256")
if digest is not None and (
not isinstance(digest, str) or not SHA256_PATTERN.fullmatch(digest)
):
raise CliError("artifact_sha256 must be exactly 64 hexadecimal characters")
shape = card["input_shape"]
if (
not isinstance(shape, list)
or len(shape) != 3
or any(isinstance(value, bool) or not isinstance(value, int) for value in shape)
):
raise CliError("input_shape must be [channels, height, width] integers")
return card
def _select(cli_value: Any, card: dict[str, Any], card_key: str, default: Any) -> Any:
return cli_value if cli_value is not None else card.get(card_key, default)
def make_plan(args: argparse.Namespace) -> dict[str, Any]:
card = _load_card(args.model_card, args.root)
shape = card.get("input_shape", [3, 256, 256])
channels = _select(args.channels, card, "channels", shape[0])
height = _select(args.height, card, "height", shape[1])
width = _select(args.width, card, "width", shape[2])
dtype = _select(args.dtype, card, "dtype", "float32")
output_elements = _select(
args.output_elements_per_tile,
card,
"output_elements_per_tile",
0,
)
activation_multiplier = _select(
args.activation_multiplier,
card,
"activation_multiplier",
8.0,
)
parameter_bytes = _select(
args.parameter_bytes,
card,
"parameter_bytes",
0,
)
runtime_workspace_mib = _select(
args.runtime_workspace_mib,
card,
"runtime_workspace_mib",
512.0,
)
integers = {
"tile_count": (args.tile_count, 1, 100_000_000),
"batch_size": (args.batch_size, 1, 1_000_000),
"channels": (channels, 1, 4096),
"height": (height, 1, 16384),
"width": (width, 1, 16384),
"output_elements_per_tile": (output_elements, 0, 10_000_000_000),
"parameter_bytes": (parameter_bytes, 0, 10_000_000_000_000),
}
checked: dict[str, int] = {}
for name, (value, minimum, maximum) in integers.items():
if isinstance(value, bool) or not isinstance(value, int):
raise CliError(f"{name} must be an integer")
if not minimum <= value <= maximum:
raise CliError(f"{name} must be between {minimum} and {maximum}")
checked[name] = value
if not isinstance(dtype, str) or dtype not in DTYPE_BYTES:
raise CliError(
"--dtype must be one of: " + ", ".join(sorted(DTYPE_BYTES))
)
activation_multiplier = finite_float(
activation_multiplier,
name="activation_multiplier",
minimum=0,
maximum=10_000,
)
runtime_workspace_mib = finite_float(
runtime_workspace_mib,
name="runtime_workspace_mib",
minimum=0,
maximum=1_000_000,
)
max_memory_mib = finite_float(
args.max_memory_mib,
name="max_memory_mib",
minimum=1,
maximum=10_000_000,
)
bytes_per_element = DTYPE_BYTES[dtype]
input_elements = checked["channels"] * checked["height"] * checked["width"]
input_bytes_per_tile = input_elements * bytes_per_element
output_bytes_per_tile = (
checked["output_elements_per_tile"] * bytes_per_element
)
activation_bytes_per_tile = math.ceil(
input_bytes_per_tile * activation_multiplier
)
variable_bytes_per_tile = (
input_bytes_per_tile + output_bytes_per_tile + activation_bytes_per_tile
)
persistent_bytes = checked["parameter_bytes"] + round(
runtime_workspace_mib * 1024**2
)
memory_budget_bytes = round(max_memory_mib * 1024**2)
available_variable_bytes = max(0, memory_budget_bytes - persistent_bytes)
recommended_max_batch = (
available_variable_bytes // variable_bytes_per_tile
if variable_bytes_per_tile
else checked["tile_count"]
)
recommended_max_batch = min(recommended_max_batch, checked["tile_count"])
planned_peak_bytes = persistent_bytes + (
variable_bytes_per_tile * checked["batch_size"]
)
within_memory = planned_peak_bytes <= memory_budget_bytes
number_of_batches = math.ceil(
checked["tile_count"] / checked["batch_size"]
)
final_batch_size = checked["tile_count"] % checked["batch_size"]
if final_batch_size == 0:
final_batch_size = min(checked["batch_size"], checked["tile_count"])
warnings = [
"Estimate excludes allocator fragmentation, framework caches, graph "
"workspace, postprocessing, stitching, and concurrent workers."
]
if not within_memory:
warnings.append(
"Requested batch exceeds the planning memory budget; reduce batch "
"size or change the explicit budget."
)
if checked["output_elements_per_tile"] == 0:
warnings.append(
"No output tensor size was supplied; output memory is omitted."
)
if checked["parameter_bytes"] == 0:
warnings.append("No parameter size was supplied; model weights are omitted.")
return {
"model_card_used": bool(card),
"model_id": card.get("model_id"),
"artifact_sha256": card.get("artifact_sha256"),
"checkpoint_or_model_loaded": False,
"input_shape_chw": [
checked["channels"],
checked["height"],
checked["width"],
],
"dtype": dtype,
"bytes_per_element": bytes_per_element,
"tile_count": checked["tile_count"],
"requested_batch_size": checked["batch_size"],
"number_of_batches": number_of_batches,
"final_batch_size": final_batch_size,
"input_bytes_per_tile": input_bytes_per_tile,
"output_bytes_per_tile": output_bytes_per_tile,
"activation_bytes_per_tile_estimate": activation_bytes_per_tile,
"persistent_bytes_estimate": persistent_bytes,
"planned_peak_memory_mib": round(planned_peak_bytes / 1024**2, 6),
"memory_budget_mib": max_memory_mib,
"within_memory_budget": within_memory,
"recommended_max_batch_under_estimate": int(recommended_max_batch),
"estimated_total_output_gib": round(
checked["tile_count"] * output_bytes_per_tile / 1024**3,
6,
),
"warnings": warnings,
"note": (
"This planner reads numbers or strict JSON metadata only. It never "
"opens model/checkpoint artifacts or imports PathML/Torch/ONNX."
),
}
def build_parser() -> argparse.ArgumentParser:
parser = argparse.ArgumentParser(
description=(
"Plan bounded inference batches from dimensions or a strict JSON "
"model card. Model/checkpoint files are never loaded."
)
)
parser.add_argument("--tile-count", type=int, required=True)
parser.add_argument("--batch-size", type=int, required=True)
parser.add_argument("--model-card")
parser.add_argument("--root", default=".")
parser.add_argument("--channels", type=int)
parser.add_argument("--height", type=int)
parser.add_argument("--width", type=int)
parser.add_argument("--dtype", choices=sorted(DTYPE_BYTES))
parser.add_argument("--output-elements-per-tile", type=int)
parser.add_argument("--activation-multiplier", type=float)
parser.add_argument("--parameter-bytes", type=int)
parser.add_argument("--runtime-workspace-mib", type=float)
parser.add_argument("--max-memory-mib", type=float, default=4096.0)
parser.add_argument("--output")
parser.add_argument("--force", action="store_true")
return parser
def main() -> None:
args = build_parser().parse_args()
report = make_plan(args)
emit_json(report, output=args.output, root=args.root, force=args.force)
if __name__ == "__main__":
raise SystemExit(run_cli(main))
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| canvas-designanthropics/skillsCreate beautiful visual art in .png and .pdf documents using design philosophy. You should use this skill when the user asks to create a poster, piece of art, design, or other static piece. Create original visual designs, never copying existing artists' work to avoid copyright violations. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| docxanthropics/skillsUse this skill whenever the user wants to create, read, edit, or manipulate Word documents (.docx files) or Word templates (.dotx files). Triggers include: any mention of 'Word doc', 'word document', '.docx', '.dotx', or requests to produce professional documents with formatting like tables of contents, headings, page numbers, or letterheads. Also use when extracting or reorganizing content from .docx or .dotx files, inserting or replacing images in documents, performing find-and-replace in Word files, working with tracked changes or comments, or converting content into a polished Word document. If the user asks for a 'report', 'memo', 'letter', 'template', or similar deliverable as a Word or .docx file, use this skill. Do NOT use for PDFs, spreadsheets, Google Docs, or general coding tasks unrelated to document generation. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| pdfanthropics/skillsUse this skill whenever the user wants to do anything with PDF files. This includes reading or extracting text/tables from PDFs, combining or merging multiple PDFs into one, splitting PDFs apart, rotating pages, adding watermarks, creating new PDFs, filling PDF forms, encrypting/decrypting PDFs, extracting images, and OCR on scanned PDFs to make them searchable. If the user mentions a .pdf file or asks to produce one, use this skill. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| pptxanthropics/skillsUse this skill any time a .pptx or .potx file is involved in any way — as input, output, or both. This includes: creating slide decks, pitch decks, or presentations; reading, parsing, or extracting text from any .pptx or .potx file (even if the extracted content will be used elsewhere, like in an email or summary); editing, modifying, or updating existing presentations; combining or splitting slide files; working with templates (.potx), layouts, speaker notes, or comments. Trigger whenever the user mentions "deck," "slides," "presentation," or references a .pptx or .potx filename, regardless of what they plan to do with the content afterward. If a .pptx or .potx file needs to be opened, created, or touched, use this skill. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |