CatalogWantedMirrorsLedgerOperating StatementDocsStudio statusCard system
alt.freeweights.text / tasktrove-dq-taco-step15-30b-a3b
From: laion <huggingface.co/laion>Newsgroups: alt.freeweights.textSubject:

tasktrove-dq-taco-step15-30b-a3b

Date: 2026-11-15Commit: 1442427Signed: 2DC28CC7B26A283B ✓

Reinforcement learning model with 30.5B parameters. Safetensors weights, 61.1 GB in all. Published by laion under Apache-2.0.

30.5B params61.1GBtasktrove-dq-taco-step15-30b-a3bsafetensorspermissive
FactsParams30.5BSize61.1 GBFormatssafetensorsTaskreinforcement learningLicenseApache-2.0From sourcepermissive
$ fw get laion/tasktrove-dq-taco-step15-30b-a3b
# manifest sha256 4241b0284b8c… · 61.1 GB · 33 files
$ fw verify ./tasktrove-dq-taco-step15-30b-a3b laion/tasktrove-dq-taco-step15-30b-a3b
From sourceHosted (signed)laion/tasktrove-dq-taco-step15-30b-a3bLast check: 2026-11-15 04:21 UTC
Run at homeGet help

Get help: custom fine-tune or private deploy. We review your setup before we quote.

Fine-tune of Qwen3-Coder-30B-A3B-Instruct.

Works with

Run8Fine-tune7Convert8Scan2

Runs with

Ranked by fit to this model (safetensors), then by GitHub stars.

TransformersLocalAItinygradTriton Inference ServerBentoMLCTranslate2ExLlamaV2LitServe
OverviewModel cardFilesUse itLicenseProvenanceModel tree

Overview

Size
61.1 GB
Weights
61.1 GB (safetensors)
Parameters
30.5B
Formats
safetensors
Task
reinforcement learning
License
Apache-2.0 (permissive)
Files
33, each with a sha256
Publisher
laion
Added
2026-11-15
Section
alt.freeweights.text
Lineage

Original release from the publisher.

Base model: Qwen/Qwen3-Coder-30B-A3B-Instruct

Where these bytes come from
Publisher at the pinned commit→You

Sources are tried in order. A file counts only when its size and sha256 match the manifest. Sources are checked when you ask for them, not on a timer.

Sources
BadgeHostLast checkResult
From sourcehuggingface.co2026-11-15 04:21 UTCpassed

Model card

From the publisher at commit 1442427. Shown as plain text with links; images open at the publisher.

tasktrove-dq-taco (step 15)#

RL checkpoint from the TaskTrove data-quality sweep, trained with SkyRL from Qwen/Qwen3-Coder-30B-A3B-Instruct using RLOO over agentic software-engineering tasks executed by OpenCode in sandboxed environments.

  • Base model: Qwen/Qwen3-Coder-30B-A3B-Instruct (Qwen3 MoE, 48 layers)
  • Checkpoint: global_step_15
  • Source run: rl-tasktrove-dq-sweep-30b-qwen3-coder-30-20260726-235656-574ba8
  • Weights: 16 safetensors shards, 61.1 GB

What this is for#

The sweep measures dataset quality, not model quality. Each arm trains the same base model on a different TaskTrove source so the sources can be compared. These checkpoints are research artifacts for that comparison. None has been evaluated as a general-purpose model, and no benchmark numbers are claimed here.

Training configuration#

RLOO (advantage_estimator: rloo_n) with megatron backend, tensor-parallel 4, pipeline-parallel 2, expert-parallel 4, across 32 H100s. The objective is deliberately unregularized: use_kl_loss: false, use_entropy_loss: false, and policy_update_steps: 1, which leaves the PPO clip ratio inert at 0.0. That choice makes entropy dynamics the primary failure mode across the sweep, and it is why several arms ended early.

Known issues with this checkpoint#

This checkpoint is salvaged from a run that failed. The arm ended early at step 38 with an entropy runaway: policy entropy rose to roughly 30x its step-1 value while reward collapsed to 0.000. The objective carries no anchor term — use_kl_loss: false, use_entropy_loss: false, and policy_update_steps: 1 leaves the PPO clip inert at ratio 0.0 — so nothing held the policy in place. Step 15 was selected by trailing-EMA reward as the last checkpoint before the collapse began. Treat it as a pre-collapse snapshot of a diverging run, not as a converged model.

Provenance#

Exported from a torch.distributed.checkpoint megatron checkpoint by re-running the trainer's own export path (bridge.save_hf_weights) at the checkpoint's own step, so no offline conversion was involved. Shard count, index total_size and weight_map completeness were verified against the object store before upload.

Training Traces#

Training-time OpenCode/Harbor rollouts for this run are published as a companion dataset: laion/terminal_bench_2_tasktrove_dq_taco_step15_30b_a3b_20260729_222705

The dataset contains the last episode of each trial (per make_and_upload_trace_dataset --episodes last) — the rollouts the policy was trained on.

Training Logs#

training_logs/ holds the parsed metric surface for this run — per-step training metrics, vLLM engine metrics, a summary report, and the reward-vs-steps plot — alongside the raw trainer log. Capability tokens have been redacted from the logs.

Files

33 files · 61.1 GBWeights 61.1 GB (safetensors). Sizes and sha256 from the signed manifest.
Show all 33 files with their sha256
FileSizesha256
.gitattributes1.71 KB704c8aef0c42eb41da6faeebd9940654d1a82eba4833624780e3560d30f306aa
README.md3.11 KB42785f51b214d074408cf327d2c61c6983a59320f5b901d3e3bf11f87302c996
chat_template.jinja6.21 KB5a38bfa05833266240066aedc497decc9b00cc0d3e3b8cceea98cf530196ab06
config.json1.02 KB3580b33db4d9c872e3d297884e3a07509a136f5399e878f53259cddb2c4f0eb6
generation_config.json216 Bf2a80a84a116bc7a296bbddfea4e48cf69a12f46e1925fe912b59c21a3651a90
model-00001-of-00016.safetensors4.00 GB7da5f6a8dcf175c4e78cbd6fcd1b03dcef3b7cfbd6924e6d01a5d0ba937e2e6d
model-00002-of-00016.safetensors4.00 GBfffda1a55e3bb5085933393189866d5bd9068a6b2766acf6c983a2f23b5b84ee
model-00003-of-00016.safetensors4.00 GBdb5e2ea3ac1959f3518b448d44e72760573f3df76034acd80a1df13178d37605
model-00004-of-00016.safetensors4.00 GB302188859c75736d110cf9e76640ae8ba7c90510833843d97d1af6d4e7b1218e
model-00005-of-00016.safetensors4.00 GB5225446469b13d6420ae1b656e84782f6dad5849255d37dc6ef5b1f3b480d79b
model-00006-of-00016.safetensors4.00 GB48a57e86ddc1b4cacd4e3297031d8471b06233c29f91bdb18b3463140fc1dfc5
model-00007-of-00016.safetensors4.00 GBaad0cef2e46245ab29640e8e963803a62f8603a746c2f213015001790178bb5c
model-00008-of-00016.safetensors4.00 GBb9837ac9e9e2570464551a98a6abd01bbc27b1c58514e9250627263a332888c1
model-00009-of-00016.safetensors4.00 GB9fd3ef512319bbb85bcc4c826393d79feae8f730482cb8018e6ea7b38a4abc89
model-00010-of-00016.safetensors4.00 GBc6a60f377538a8408816c45464f9df97a85e85b2a366b949323c58d4cac35590
model-00011-of-00016.safetensors4.00 GB64cf4e211c455d004a101dcef034688bae963a5c0555b53b187fa01d60dc59c6
model-00012-of-00016.safetensors3.99 GB7263ae7e19053fdc69333423bd0961517b00eab998e0fbe2e11e339639315a91
model-00013-of-00016.safetensors4.00 GBd89c3abd23237ce6329014816637afad83ddcbb1d99c0227c90b8bf938a2ebab
model-00014-of-00016.safetensors4.00 GB837e87da1a31b4cbc81bab9ce20093651da7dfcab79d701c9f9b37effe0885ed
model-00015-of-00016.safetensors4.00 GB308d26e6674d78a1f54ba7f73758bce7eec15b990bf5074ded2bfa6bccf343e8
model-00016-of-00016.safetensors1.09 GB39b593583a8cfc117a6c8a85b3e82f65c77bcfe1b88e766660c92d96c28031f5
model.safetensors.index.json1.78 MB7294b00543d1f59a1f55f93acd1a0d2c1ba80fbb158b8c26cc590deff12b3422
tokenizer.json11.4 MBbe75606093db2094d7cd20f3c2f385c212750648bd6ea4fb2bf507a6a4c55506
tokenizer_config.json695 Bf6f4c67578d26ccdf6535fa0bc88683b9a0e7acb6797aca1f4759ad3330c22e9
training_logs/20260730_095724_metrics_finelog.csv29.2 KB046492a88e5d05f640330b272e049d6609b7b4db54855bbd3609342a0ef1e27d
training_logs/20260730_095724_metrics_table.csv29.6 KB8ca259b3e738681c7de0db32cd586564a33135a8c92d3a2a458ae310d548925e
training_logs/20260730_095724_vllm_metrics_finelog.csv559 KB391da9c781f178fcac5531222ac6fee08bc5218b738ad848729421cdb54df5c6
training_logs/20260730_095724_vllm_metrics_table.csv610 KBe1839819481b0f74d8558de8f2ac4678c1d1567a1b682ea608c149cf8f3381b7
training_logs/574ba8_trainer.log54.6 MB8e59f99424581e24869f37c9d4360be8ddadee2cdf8abb89a3d846324b40f3e2
training_logs/metrics.csv29.6 KB8ca259b3e738681c7de0db32cd586564a33135a8c92d3a2a458ae310d548925e
training_logs/report.md13.3 KBdc7c1bcd71b4dcf0bd6e3326c65b5e537bed5cf57ec46bc62f60d28e3e48e80e
training_logs/reward_plot.png150 KB3791eb058185fdfa7642114916449e12194bdf2c414f04a9b305c4a66f93aec1
training_logs/vllm_metrics.csv610 KBe1839819481b0f74d8558de8f2ac4678c1d1567a1b682ea608c149cf8f3381b7
VersionsEach time the pinned upstream commit or file list changed in this repo. Newest first.
Upstream commitDateContents
144242737e pinned now2026-11-15 04:21 UTC33 files · 61.1 GB

Use it

$ fw get laion/tasktrove-dq-taco-step15-30b-a3b
# manifest sha256 4241b0284b8c… · 61.1 GB · 33 files
$ fw verify ./tasktrove-dq-taco-step15-30b-a3b laion/tasktrove-dq-taco-step15-30b-a3b
fw.pyChecks the signature and every sha256, tries each source in order, resumes.
$ python3 fw.py get https://freeweights.allternit.com/manifests/laion/tasktrove-dq-taco-step15-30b-a3b.json --dir ./tasktrove-dq-taco-step15-30b-a3b $ python3 fw.py verify ./tasktrove-dq-taco-step15-30b-a3b ./laion__tasktrove-dq-taco-step15-30b-a3b.json
hfThe Hugging Face CLI with one setting. Same repo id.
$ HF_ENDPOINT=https://freeweights.allternit.com hf download laion/tasktrove-dq-taco-step15-30b-a3b --revision 144242737e27fe0f45766b6baabd5c187a0881ae --local-dir ./tasktrove-dq-taco-step15-30b-a3b
Run at home

License

Name
Apache License 2.0
Id
Apache-2.0
Class
Permissive
Redistribution
yes
Listing
Hosted (signed)
Decided
License as declared by the publisher. No person read it. 2026-11-15
Text in
README.md
Text sha256
42785f51b214d074408cf327d2c61c6983a59320f5b901d3e3bf11f87302c996
Terms gate
No extra terms beyond the license.

Read the license text

Licenses are checked against the license table. A custom license passes only after a person reads it.

Provenance

From: laion <huggingface.co/laion>Newsgroups: alt.freeweights.textSubject: tasktrove-dq-taco-step15-30b-a3bDate: 2026-11-15Commit: 1442427Signed: 2DC28CC7B26A283B ✓
Upstream
laion/tasktrove-dq-taco-step15-30b-a3b
Pinned commit
144242737e27fe0f45766b6baabd5c187a0881ae
Manifest
tasktrove-dq-taco-step15-30b-a3b.json · .minisig
Manifest sha256
4241b0284b8cb47f3d7816615c4c61d1af1a991eee0f2ca3a3483eba63ae9a8b
Signature
minisign, Ed25519, key 2DC28CC7B26A283B
Created
2026-11-15 04:21 UTC
Generator
freeweights-manifest/0.1.0
Takedowns
None.

How to verify

$ curl -O https://freeweights.allternit.com/manifests/laion/tasktrove-dq-taco-step15-30b-a3b.json $ curl -O https://freeweights.allternit.com/manifests/laion/tasktrove-dq-taco-step15-30b-a3b.json.minisig $ minisign -Vm tasktrove-dq-taco-step15-30b-a3b.json -P RWQ7KGqyx4zCLXtUFOqvfv8L3g+8fpKJ1FsuxPyOYMF+4KufbUxKq1Oq $ shasum -a 256 <file> # compare with the Files tab

A manifest whose signature fails must be rejected. A file whose size or sha256 does not match must be discarded. Verify a download

Model tree

Parent

Qwen/Qwen3-Coder-30B-A3B-Instruct · this model is a finetune of it

No model in the catalog names this one as its parent.

Get it
Download
$ HF_ENDPOINT=https://freeweights.allternit.com hf download laion/tasktrove-dq-taco-step15-30b-a3b
Open in Freeweights Studio

Pull, verify and run it in the Allternit app on Desktop, web or phone.

Signed manifest
minisign, key 2DC28CC7B26A283B. tasktrove-dq-taco-step15-30b-a3b.json. Provenance. Takedown history: none.
Source
huggingface.co/laion/tasktrove-dq-taco-step15-30b-a3b

Related models

All of alt.freeweights.text →
From: laion <huggingface.co/laion>Subject: a2-rl-e2egit_large-50-32B
From: laion <huggingface.co/laion>Subject: a3-rl-DCAgent_exp_rpt_e2egit-large-15-8B
From: laion <huggingface.co/laion>Subject: a3-rl-DCAgent_exp_rpt_e2egit-v2-10-8B
From: laion <huggingface.co/laion>Subject: a3-rl-DCAgent_exp_rpt_pymethods2test-large-80-8B
From: laion <huggingface.co/laion>Subject: a3-rl-DCAgent_exp_rpt_unitsyn-python-large-10-8B
From: laion <huggingface.co/laion>Subject: a3-rl-DCAgent_mix_h2_language_proportional-65-8B