pra-qwen3-0.6b

PRA Runtime Bundle for Qwen/Qwen3-0.6B · HF

What this PRA Runtime Bundle is

This repository packages the model-specific Progressive Retrieval Attention (PRA) structural mapping, runtime profiles, optional learned components, compatibility metadata, and measured qualification evidence. It does not contain the base-model weights and is not an ordinary LoRA quality fine-tune.

• Base model: Qwen/Qwen3-0.6B
• Immutable revision: c1899de289a04d12100db370d81485cdf75e47ca
• Architecture: Qwen3ForCausalLM
• Parameters: 596049920
• Tokenizer revision: c1899de289a04d12100db370d81485cdf75e47ca
• Serving precision: UNSPECIFIED / UNSPECIFIED

Recommended configuration

• Engine: hf
• Recommended PRA mode: Selected Context
• Recommended profile: BALANCED
• Bundle evidence tier: RESEARCH
• Native Memory status: controlled validation

Availability, qualification, and recommendation are separate. A mode may be implemented without being qualified or recommended for this identity.

Precision qualification

Precision evidence is scoped to the exact model conversion, engine, mode, and profile. Qualification does not transfer automatically between BF16, INT8, INT4, or encoding-specific formats.

Headline results

No paired end-task headline is available for this exact model, revision, quantization, engine, profile, and execution mode. Routing diagnostics below must not be interpreted as application quality.

Evidence by engine, mode, and profile

Each row identifies the exact runtime surface for which metrics are available. MEASURED counts scalar metrics with real observations; missing profile/mode combinations are not inferred from another row.

Canonical staged evidence

A complete staged cohort is not packaged for this exact identity.

Existing selector-frozen Selected Context versus Native Memory measurements remain reported below as transport evidence; they are not silently relabeled as adaptor evidence.

Installation

pip install 'pra-hf[hf-hub,hf-runtime]'
pra doctor

Quickstart

pra inspect Qwen/Qwen3-0.6B -e hf -a EInnovator/pra-qwen3-0.6b
pra evaluate Qwen/Qwen3-0.6B -e hf -D qasper -a EInnovator/pra-qwen3-0.6b
pra recommend .pra/runs/latest
pra serve Qwen/Qwen3-0.6B -e hf -a EInnovator/pra-qwen3-0.6b -p balanced

Profiles

Engine compatibility

End-to-end qualification

What remains to be measured: paired end-task quality for this exact bundle identity.

Native Memory qualification

What remains to be measured: paired Selected Context versus Native Memory quality and serving economics.

Research diagnostics

These are qualification measurements, not guaranteed production performance. Run pra evaluate on your hardware and workload. Engine version, profile, cohort, evidence tier, date, and artifact provenance remain recorded in qualification/ and bundle.yaml.

How to evaluate locally

pra evaluate Qwen/Qwen3-0.6B -e hf -a EInnovator/pra-qwen3-0.6b -D qasper -o .pra/runs/qasper
pra recommend .pra/runs/qasper
pra report .pra/runs/qasper --format html

Known limitations

• This is a research/reference bundle for mechanism reproduction, not the flagship production-economics demonstration.
• The routing cohort is small and QASPER-specific; transfer routing is not a production claim.
• Native Memory has controlled HF evidence only for this exact model revision and must be requalified per engine, quantization, and hardware.
• Native Serving is not qualified by this bundle.

Training/creation

• Datasets: QASPER
• Seed: 11
• Method: multi-positive softmax
• Parameter Count: 262144
• Base Revision: c1899de289a04d12100db370d81485cdf75e47ca

Reproducibility

• PRA commit: 81f42d69936bf50eb6fe11a0f7477b415bbf250d
• Bundle build commit: 81f42d69936bf50eb6fe11a0f7477b415bbf250d
• Bundle schema: 2
• PRA package: 0.2.0rc1
• Component fingerprints and file checksums are recorded in bundle.yaml.

Community/support

• PRA documentation
• Source repository
• Issues
• Contribution guide
• Canonical PRA Bundles Collection
• EInnovator on Hugging Face