Compilation And Export

What it does

When your agent uses it

• Tasks that involve LLM inference and serving

Example prompts

• “/compilation-and-export”

Requirements

• Python 3

Workflow steps

6 steps, taken from the first numbered list in SKILL.md.

1. 1Verify install/features with the root environment probe if the environment is unknown.
2. 2Put the model in eval mode and move model plus representative inputs to CUDA.
3. 3Decide static or dynamic inputs. For dynamic shapes, create torch_tensorrt.Input(min_shape=..., opt_shape=..., max_shape=..., dtype=...)…
4. 4Start with FP32 or FP16 only after checking numerical tolerance expectations. Avoid INT8/FP8/FP4 until ModelOpt or calibration…
5. 5Compile a small representative shape, run the compiled output and PyTorch output, and compare with torch.testing.assert_close at…
6. 6Save only after execution works. Choose .ep, .ts, .pt2, .engine, or .pte from the artifact matrix.

What it can do on your machine

Read from SKILL.md and the folder at commit ac3fe1a. It shows what the files ask for, not the result of running them.

• Tool permissions

  Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

  From allowed-tools in the SKILL.md frontmatter.

• Runs code

  Ships 1 file in scripts/ (Python), which the agent can run.

  From the folder's file list and the shell code blocks in SKILL.md.

• Network

  No URLs in SKILL.md.

  From URLs in SKILL.md, links to its own repository left out.

• Credentials

  Names no API keys, tokens, secrets or passwords.

  From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Compilation And Export loads about 1.3k tokens when it runs, and up to ~7.4k if it reads all its reference files. Until then it costs about 51 tokens; SKILL.md has 448 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it

~51

When it runs · the whole SKILL.md, loaded when a task matches

~1.3k

With references · SKILL.md plus every file in references/, read only if the agent opens them

~7.4k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); the scripts in this folder are not scanned.

SKILL.md

The full file from VectorSpaceLab/AREX-Skill at commit ac3fe1a, republished under its BSD-3-Clause licence (© VectorSpaceLab). 448 words, ~1,291 tokens.

name

compilation-and-export

description

Use this sub-skill for Torch-TensorRT model compilation, dynamic input planning, torch.export workflows, save/load formats, raw TensorRT engines, and compile-time troubleshooting.

metadata.disco-role

operating

disable-model-invocation

true

license

BSD 3-Clause

import torch
import torch_tensorrt

model = MyModel().eval().cuda()
optimized = torch.compile(
    model,
    backend="torch_tensorrt",
    options={"enabled_precisions": {torch.float16}, "min_block_size": 3},
)
out = optimized(torch.randn(1, 3, 224, 224, device="cuda"))  # first call compiles

import torch
import torch_tensorrt

model = MyModel().eval().cuda()
inputs = [torch_tensorrt.Input((1, 3, 224, 224), dtype=torch.float32)]
compiled = torch_tensorrt.compile(model, ir="dynamo", inputs=inputs)

import torch
import torch_tensorrt

model = MyModel().eval().cuda()
example = (torch.randn(4, 128, device="cuda"),)
batch = torch.export.Dim("batch", min=1, max=16)
exported = torch.export.export(model, example, dynamic_shapes={"x": {0: batch}})
compiled = torch_tensorrt.dynamo.compile(
    exported,
    inputs=[torch_tensorrt.Input(min_shape=(1, 128), opt_shape=(4, 128), max_shape=(16, 128))],
)

© VectorSpaceLab, BSD-3-Clause. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

SKILL.md and 6 other files (scripts, references) in skills/repositories/repo-skills/torch-tensorrt/sub-skills/compilation-and-export of VectorSpaceLab/AREX-Skill.

• SKILL.md
• references/api-reference.md
• references/dynamic-shapes-and-inputs.md
• references/serialization-and-engines.md
• references/troubleshooting.md
• references/workflows.md
• scripts/compile_probe.py

Open the folder on GitHubat commit ac3fe1a