fix(config): drop redundant qwen3-8b-dta config and example

[curnane-lab]

Motivation

As reported in #679, VP-Drafter support was removed in PR #660: 832263f dropped the training_mode/prefix_weight_base wiring in scripts/train_dflash.py, and fe0ad18 removed the remaining vp_drafter loss implementation from specforge/core/dflash.py. The two fields left in dflash_config of configs/qwen3-8b-dta.json are no longer read anywhere and are silently ignored, so training silently falls back to regular DFlash.

Without those dead fields, qwen3-8b-dta.json is byte-identical to configs/qwen3-8b-dflash.json. The config and examples/run_qwen3_8b_dta_online.sh were added together in 0b948e7 solely for the removed VP-Drafter training mode, so both are now redundant — keeping a duplicate config only risks future drift.

Modifications

• Deleted configs/qwen3-8b-dta.json (a byte-identical duplicate of qwen3-8b-dflash.json once the dead fields are excluded).
• Deleted examples/run_qwen3_8b_dta_online.sh (its only distinctive reference was the deleted config; the same training can be launched with --draft-config-path configs/qwen3-8b-dflash.json).

Related Issues

Fixes #679.

Accuracy Test

N/A — config/example cleanup, no model-side changes.

Benchmark & Profiling

N/A — no performance impact expected.

Checklist

• Format your code according to the https://docs.sglang.ai/references/contribution_guide.html#code-formatting-with-pre-commit .
• Add unit tests as outlined in the https://docs.sglang.ai/references/contribution_guide.html#running-unit-tests-adding-to-ci .
• Update documentation / docstrings / example tutorials as needed, according to https://docs.sglang.ai/references/contribution_guide.html#writing-documentation-running-docs-ci .
• Provide throughput / latency benchmark results and accuracy evaluation results as needed, according to https://docs.sglang.ai/references/benchmark_and_profiling.html and https://docs.sglang.ai/references/accuracy_evaluation.html .
• For reviewers: If you haven't made any contributions to this PR and are only assisting with merging the main branch, please remove yourself as a co-author when merging the PR.
• Please feel free to join our Slack channel at https://sgl-fru7574.slack.com/archives/C09784E3EN6 to discuss your PR.

[gemini-code-assist]

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!