SKILLEMALL.ai
GitHub ↗ RU Sign in

CD finetune-speech-llm

Fine-tune a SALM speech-LLM checkpoint (speech encoder + projection + LLM decoder) with NeMo speechlm2 on a new domain or task, and evaluate it with vLLM. Use to adapt a speech-to-text LLM to new audio, improve WER on a domain corpus, train speech translation, spoken QA or audio classification, or debug a fine-tune that shows no improvement. Covers planning and cost, manifests, config derivation, preflight, training, LoRA export and merge, evaluation, diagnosis, checkpoint averaging, and ensembling.

NVIDIA-NeMo/Speech Agent Skills author: NVIDIA-NeMo Apache-2.0 1 file body ≈ 6 193 tokens Open the sourcegithub.com↗ analyzed 6 d ago

Fine-tune a SALM speech-LLM checkpoint (speech encoder + projection + LLM decoder) with NeMo speechlm2 on a new domain or task, and evaluate it with vLLM.

As a process D 47/100 · Unfinished process — References files that are not bundled: examples/speechlm2/finetuning/docs/fine-tuning-guide.md, examples/speechlm2/finetuning/docs/results.md, examples/speechlm2/finetuning/docs/medical-asr-synthetic-data.md

AnalyzerAI and agentstype and topics are labelled automatically from the skill text
Runs in: Claude / Claude Code ChatGPT / Codex Gemini CLI Cursor GitHub Copilot / VS Code Qwen Code Kimi Code CLI Z.ai (GLM) DeepSeek (Deep Code) Mistral Vibe LangChain (deepagents) OpenClaw Goose OpenCode
JSON
Technical rating
C
86/100
safety, quality, tests
Safety 60%
100
Quality 40%
65
Run on models
none yet
Process rating
D
47/100
Unfinished process
References files that are not bundled: examples/speechlm2/finetuning/docs/fine-tuning-guide.md, examples/speechlm2/finetuning/docs/results.md, examples/speechlm2/finetuning/docs/medical-asr-synthetic-data.md
Tools and files w 18
0
When it triggers w 12
20
Inputs and preconditions w 11
30
the three weakest of ten parameters · all ten

How to improve

  1. The SKILL.md body is over 5,000 tokens: move reference detail into references/ and load it when needed.
  2. The text references files that are not there: add them or drop the references.
For the model run — optional
  • Your own cases (evals/evals.json, 4–6 real requests with expected answers): the full check would then run those instead of a model-drafted suite.
  • A spec.yaml with trigger phrases and assertions — a behaviour contract for CI; `skilltest init` writes a template.

Guard findings · 0

✓ No critical or high findings

Files scanned: 1. Evidence is masked. Grey chips explain why severity was lowered.

Against the Agent Skills spec

  • warning body-long SKILL.md body ≈ 6193 tokens (recommended < 5000); move details to references/
  • warning missing-ref reference to a missing file: examples/speechlm2/finetuning/docs/fine-tuning-guide.md
  • warning missing-ref reference to a missing file: examples/speechlm2/finetuning/docs/results.md
  • warning missing-ref reference to a missing file: examples/speechlm2/finetuning/docs/medical-asr-synthetic-data.md
  • warning missing-ref reference to a missing file: examples/speechlm2/finetuning/README.md
  • warning missing-ref reference to a missing file: examples/speechlm2/finetuning/sdg/
  • warning missing-ref reference to a missing file: examples/speechlm2/conf/salm_automodel.yaml
  • warning missing-ref reference to a missing file: examples/speechlm2/salm_train.py
  • warning missing-ref reference to a missing file: examples/speechlm2/to_hf.py
  • warning missing-ref reference to a missing file: examples/speechlm2/finetuning/

Process rating: all ten parameters 47/100

Will not run. References files that are not bundled: examples/speechlm2/finetuning/docs/fine-tuning-guide.md, examples/speechlm2/finetuning/docs/results.md, examples/speechlm2/finetuning/docs/medical-asr-synthetic-data.md
  • 0Tools and files. 9 referenced file(s) missing: examples/speechlm2/finetuning/docs/fine-tuning-guide.md, examples/speechlm2/finetuning/docs/results.md, examples/speechlm2/finetuning/docs/medical-asr-synthetic-data.md
  • 20When it triggers. No condition that starts the skill
  • 30Inputs and preconditions. Does not say what the process needs to start
  • 30Running it twice. 18 mutating operations with no state check
  • 50Failures and branches. 0 branches, has a failure section
  • 60Result and completion. Output format stated, no completion criterion
  • 70Execution cost. Instruction body is 6193 tokens
  • 85Steps. 71 steps, 1 vague phrases
  • 100Consistency. Name and required fields are in place
  • 100Progress reporting. Reports progress
  • medium Safety rules and hard prohibitions inside a skill: they belong in the system prompt, here they protect nothing
  • low 14 top-level sections: this looks like several domains in one skill
  • low The response is described with custom markup (4 tags): a typed call is more reliable

Everything here is measured from the skill text rather than judged by a model, so the numbers are checkable. A parameter weighs more when it is a more common reason for the process to stall.

Quality signals

  • +5Description has no quoted example phrases that should trigger the skill
  • +4Description does not say when NOT to use the skill (false activations)
  • +1No license
  • +2Single-language instructions
  • +3Description length 504: enough signal without eating the budget
  • +4Structure: 16 headings
  • +3Step-by-step instructions: 71 items
  • +3Output format is stated explicitly
  • +4Has examples (7 code blocks)

Quality base 70; lint remarks subtract, signals add up to 100. Result: 65.