---
type: "post"
id: "checkpoints-available-in-both-jax-and-pytorch-via-aac216"
slug: "checkpoints-available-in-both-jax-and-pytorch-via-aac216"
title: "Checkpoints available in both JAX and PyTorch via Hugging Face ."
description: "Google’s Hugging Face card makes the TAP family’s checkpoint catalog easier to inspect and use across existing video workflows."
retrieval_nugget: "Google’s Hugging Face card makes the TAP family’s checkpoint catalog easier to inspect and use across existing video workflows."
published_at: "2026-10-10"
updated_at: "2026-10-10"
record_date: "2026-10-10"
date_kind: "published_at"
topics: ["google"]
source_urls: ["https://huggingface.co/google/tapnet"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:checkpoints-available-in-both-jax-and-pytorch-via-aac216"
status: "published"
visuals: []
editorial_provenance: {"schema_version":"newruntime-editorial-copy-v1","content_status":"source_grounded_final","final_copy_sha256":"sha256:4c1e63cd76b0beaa1fcbb72b2c688a43cb5c5d0cbab2047ec110bb2f9f857003","reviewed_at":"2026-10-10T23:01:48.801Z","source_evidence_count":1,"verified_claim_count":3,"site_analysis_schema_version":"newruntime-site-analysis-v1","site_object_kind":"field_note","observed_fact_count":3,"implication_count":2,"watch_condition_count":1,"related_record_count":0}
analysis: {"schema_version":"newruntime-site-analysis-v1","object_kind":"field_note","thesis":"Google’s Hugging Face card makes the TAP family’s checkpoint catalog easier to inspect and use across existing video workflows.","observed_facts":[{"text":"The card identifies the checkpoints as official pretrained models from Google DeepMind and Google Research.","source_urls":["https://huggingface.co/google/tapnet"]},{"text":"The catalog lists TAPIR checkpoints in both JAX and PyTorch and includes online and offline variants for some models.","source_urls":["https://huggingface.co/google/tapnet"]},{"text":"The card states that the checkpoints and software are licensed under Apache 2.0.","source_urls":["https://huggingface.co/google/tapnet"]}],"mechanism":"The mechanism is a model catalog that exposes checkpoint files, framework choices, operating modes, setup examples, and licensing in one place.","why_now":"This is an implementation and distribution update, not evidence of a new model release or a comparative performance advance.","implications":["That lowers friction for builders who need to test point tracking in an existing stack rather than porting around a single framework-specific artifact.","The practical implication is that teams can compare integration paths for point tracking without treating framework support as a proxy for model quality."],"evidence_boundary":"The source provides checkpoint metadata and reported benchmark figures, but does not establish independent reproducibility, current comparative standing, or suitability for a particular production video pipeline.","watch_conditions":["Watch whether maintained examples and community integrations make these checkpoints straightforward to run on real workloads, and whether users report consistent behavior across the listed frameworks."],"related_records":[],"new_branch_reason":"No prior New Runtime record is supplied for this checkpoint catalog and framework-availability update, so it starts a distinct analytical record."}
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/checkpoints-available-in-both-jax-and-pytorch-via-aac216/","markdown":"https://newruntime.com/posts/checkpoints-available-in-both-jax-and-pytorch-via-aac216.md","json":"https://newruntime.com/posts/checkpoints-available-in-both-jax-and-pytorch-via-aac216.json"}
---

# Checkpoints available in both JAX and PyTorch via Hugging Face .

## Retrieval answer

Google’s Hugging Face card makes the TAP family’s checkpoint catalog easier to inspect and use across existing video workflows.

Google’s Hugging Face card makes the TAP family’s checkpoint catalog easier to inspect and use across existing video workflows. The card lists official pretrained checkpoints from Google DeepMind and Google Research, and says the family tracks a queried point on a physical surface across video frames while estimating visibility. This is an implementation and distribution update, not evidence of a new model release or a comparative performance advance.

The mechanism is a model catalog that exposes checkpoint files, framework choices, operating modes, setup examples, and licensing in one place. The listed TAPIR family includes JAX and PyTorch checkpoints, and the page also shows online and offline variants for some models. That lowers friction for builders who need to test point tracking in an existing stack rather than porting around a single framework-specific artifact. The card states Apache 2.0 licensing for checkpoints and software.

The practical implication is that teams can compare integration paths for point tracking without treating framework support as a proxy for model quality. The source provides checkpoint metadata and reported benchmark figures, but does not establish independent reproducibility, current comparative standing, or suitability for a particular production video pipeline. Watch whether maintained examples and community integrations make these checkpoints straightforward to run on real workloads, and whether users report consistent behavior across the listed frameworks.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Google](https://newruntime.com/topics/google.md): Explore the google topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Source Ledger](https://newruntime.com/source-ledger.json): Inspect the public source evidence graph.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)