[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill-nvidia-riva-asr":3,"mdc-3s9cyk-key":34,"related-repo-nvidia-riva-asr":4610,"related-org-nvidia-riva-asr":4674},{"slug":4,"name":4,"fn":5,"description":6,"org":7,"tags":11,"stars":23,"repoUrl":24,"updatedAt":25,"license":26,"forks":27,"topics":28,"repo":29,"sourceUrl":32,"mdContent":33},"riva-asr","run and test Riva ASR models","Use when the user wants to deploy, run, or test an ASR (speech-to-text) Riva NIM — cloud-hosted (build.nvidia.com) or self-hosted Parakeet\u002FCanary\u002FWhisper.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},"nvidia","NVIDIA","https:\u002F\u002Fpexgzepcugksgbtrxkhf.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Forg-logos\u002Fnvidia.png",[12,14,17,20],{"name":9,"slug":8,"type":13},"tag",{"name":15,"slug":16,"type":13},"AI Infrastructure","ai-infrastructure",{"name":18,"slug":19,"type":13},"Testing","testing",{"name":21,"slug":22,"type":13},"Speech","speech",6,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fdigital-health-skills","2026-07-14T05:35:49.042673","Apache-2.0",1,[],{"repoUrl":24,"stars":23,"forks":27,"topics":30,"description":31},[],"NVIDIA Digital Health skills repo for agent-guided healthcare AI workflows, including evaluation, data generation, model adaptation, deployment guidance, and developer best practices.","https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fdigital-health-skills\u002Ftree\u002FHEAD\u002Fskills\u002Friva-asr","---\nname: \"riva-asr\"\nlicense: \"Apache-2.0\"\ndescription: \"Use when the user wants to deploy, run, or test an ASR (speech-to-text) Riva NIM — cloud-hosted (build.nvidia.com) or self-hosted Parakeet\u002FCanary\u002FWhisper.\"\nmetadata:\n  author: \"Mayank Jain \u003Cmayjain@nvidia.com>\"\n  team: riva\n  tags:\n    - nvidia\n    - riva\n    - nim\n    - asr\n    - speech-to-text\n    - parakeet\n    - canary\n    - whisper\n    - grpc\n    - http\n    - cloud\n    - nvcf\n  domain: ml\n  version: \"1.0.0\"\n---\n\n# Riva ASR NIM\n\n> **Agent:** When walking the user through a multi-step workflow, announce each step before presenting it: **Step N\u002FM — Step Title** (e.g., \"**Step 1\u002F4 — Set Model Variables**\").\n\n---\n\n## Purpose\n\nDeploy and run NVIDIA Riva ASR (speech-to-text) NIMs. Supports cloud-hosted\ninference via build.nvidia.com (no GPU required) and self-hosted deployment on\nyour own GPU using Docker. Covers streaming and offline transcription, word\nboosting, and performance benchmarking.\n\n## Workflow\n\nChoose **Option A** (cloud) for quick testing without a GPU, or **Option B** (self-hosted) for production. Self-hosted follows a 4-step process: set model variables → run container → verify health → run inference.\n\n## Prerequisites\n\n- Complete `riva-nim-setup` before self-hosted deployment: NVIDIA Container Toolkit, `NGC_API_KEY` exported, Docker logged in to `nvcr.io`\n- Cloud-hosted inference: `pip install -U nvidia-riva-client` and a valid `NVIDIA_API_KEY`\n- Not sure which model to use? Run `riva-model-selection` first\n\n## Instructions\n\n1. **Pick a path**: cloud inference (Option A) or self-hosted NIM (Option B).\n2. **For cloud inference (Option A)**: install `nvidia-riva-client`, set `NVIDIA_API_KEY`, pick a function ID, run `transcribe_file.py` (streaming) or `transcribe_file_offline.py` (offline) against `grpc.nvcf.nvidia.com:443` with `--use-ssl`.\n3. **For self-hosted (Option B)**: look up `CONTAINER_ID` and `NIM_TAGS_SELECTOR` from the ASR support matrix, mount a model cache dir with `chmod 700`, then follow Steps 1–4 below (set vars, run container, verify readiness, run inference).\n\n## Option A — Cloud-Hosted Inference (build.nvidia.com)\n\n**Setup:** `pip install -U nvidia-riva-client` + clone https:\u002F\u002Fgithub.com\u002Fnvidia-riva\u002Fpython-clients and `cd` into it.\n\n**Auth:** Set `NVIDIA_API_KEY` from https:\u002F\u002Fbuild.nvidia.com (different from `NGC_API_KEY`).\n\n**Server:** `grpc.nvcf.nvidia.com:443` — always pass `--use-ssl`.\n\n**Canonical command (streaming models):**\n\n```bash\npython python-clients\u002Fscripts\u002Fasr\u002Ftranscribe_file.py \\\n    --server grpc.nvcf.nvidia.com:443 --use-ssl \\\n    --metadata function-id \"\u003CFUNCTION_ID>\" \\\n    --metadata \"authorization\" \"Bearer $NVIDIA_API_KEY\" \\\n    --language-code \u003CLANG_CODE> \\\n    --input-file \u002Fpath\u002Fto\u002Faudio.wav\n```\n\nFor offline-only models use `transcribe_file_offline.py` instead. For Whisper translation add `--custom-configuration \"task:translate\"`.\n\n**Word timestamps:** All models except Whisper and Canary support `--word-time-offsets`. Accuracy ranking: CTC > TDT > RNNT.\n\n**Note:** Both cloud and self-hosted scripts use `--input-file`, not `--audio-file`.\n\n### Function IDs\n\nIf a function-id no longer works, fetch the current one from `https:\u002F\u002Fbuild.nvidia.com\u002F\u003Corg>\u002F\u003Cmodel>\u002Fapi`.\n\n| Model | Build Page slug | Function ID | Script | Lang code |\n|-------|----------------|-------------|--------|-----------|\n| Nemotron ASR Streaming | `nvidia\u002Fnemotron-asr-streaming` | `bb0837de-8c7b-481f-9ec8-ef5663e9c1fa` | `transcribe_file.py` | *(omit)* |\n| Parakeet 1.1B RNNT Multilingual | `nvidia\u002Fparakeet-1_1b-rnnt-multilingual-asr` | `71203149-d3b7-4460-8231-1be2543a1fca` | `transcribe_file.py` | `multi` |\n| Parakeet TDT 0.6B v2 | `nvidia\u002Fparakeet-tdt-0_6b-v2` | `d3fe9151-442b-4204-a70d-5fcc597fd610` | `transcribe_file_offline.py` | `en-US` |\n| Whisper Large v3 | `openai\u002Fwhisper-large-v3` | `b702f636-f60c-4a3d-a6f4-f3568c13bd7d` | `transcribe_file_offline.py` | `en` \u002F `multi` |\n| Parakeet CTC 1.1B English | `nvidia\u002Fparakeet-ctc-1_1b-asr` | `1598d209-5e27-4d3c-8079-4751568b1081` | `transcribe_file.py` | `en-US` |\n| Parakeet CTC 0.6B Spanish | `nvidia\u002Fparakeet-ctc-0_6b-es` | `a9eeee8f-b509-4712-b19d-194361fa5f31` | `transcribe_file.py` | `es-US` |\n| Parakeet CTC 0.6B Vietnamese | `nvidia\u002Fparakeet-ctc-0_6b-vi` | `f3dff2bb-99f9-403d-a5f1-f574a757deb0` | `transcribe_file.py` | `vi-VN` |\n| Parakeet CTC 0.6B Mandarin (Simplified) | `nvidia\u002Fparakeet-ctc-0_6b-zh-cn` | `9add5ef7-322e-47e0-ad7a-5653fb8d259b` | `transcribe_file.py` | `zh-CN` |\n| Parakeet CTC 0.6B Taiwanese (Traditional) | `nvidia\u002Fparakeet-ctc-0_6b-zh-tw` | `8473f56d-51ef-473c-bb26-efd4f5def2bf` | `transcribe_file.py` | `zh-TW` |\n\n---\n\n## Option B — Self-Hosted NIM Deployment\n\nComplete `riva-nim-setup` first: NVIDIA Container Toolkit, `NGC_API_KEY` exported, Docker logged in to `nvcr.io`. Driver and VRAM minimums: see https:\u002F\u002Fdocs.nvidia.com\u002Fnim\u002Fspeech\u002Flatest\u002Freference\u002Fsupport-matrix\u002Fasr.html. Latency and throughput benchmarks: https:\u002F\u002Fdocs.nvidia.com\u002Fnim\u002Fspeech\u002Flatest\u002Freference\u002Fperformances\u002Fasr\u002Fperformance.html\n\n## Step 1 — Set Model Variables\n\nGet the current `CONTAINER_ID` and `NIM_TAGS_SELECTOR` values from the ASR support matrix (includes all models, modes, VRAM, and deployment profiles):\nhttps:\u002F\u002Fdocs.nvidia.com\u002Fnim\u002Fspeech\u002Flatest\u002Freference\u002Fsupport-matrix\u002Fasr.html\n\n```bash\nexport CONTAINER_ID=\u003Ccontainer-id-from-support-matrix>\nexport NIM_TAGS_SELECTOR=\"\u003Cselector-from-support-matrix>\"\n```\n\n`NIM_TAGS_SELECTOR` pattern: `name=\u003Cmodel-name>,mode=\u003Cstr|offline|all>[,model_type=\u003Cprebuilt|rmir>]`\n\n**Prebuilt vs RMIR:** The NIM auto-detects your GPU on startup. For well-known GPUs (A100, H100, select Blackwell), it pulls a prebuilt model repo — a tarball of TensorRT engines already compiled for that GPU via `riva-build` + `riva-deploy`. For unsupported GPUs, it falls back to RMIR (Riva Model Intermediate Representation), which is a portable format that gets compiled into TensorRT engines on your local GPU at first run (slower startup, same runtime performance). You rarely need to set `model_type` explicitly — omit it and the NIM picks the right one automatically.\n\n## Step 2 — Run the Container\n\n```bash\nexport LOCAL_NIM_CACHE=~\u002F.cache\u002Fnim\nmkdir -p $LOCAL_NIM_CACHE && chmod 700 $LOCAL_NIM_CACHE\n\ndocker run -it --rm --name=$CONTAINER_ID \\\n  --runtime=nvidia \\\n  --gpus '\"device=0\"' \\\n  --shm-size=8GB \\\n  -e NGC_API_KEY \\\n  -e NIM_TAGS_SELECTOR \\\n  -e NIM_HTTP_API_PORT=9000 \\\n  -e NIM_GRPC_API_PORT=50051 \\\n  -p 9000:9000 \\\n  -p 50051:50051 \\\n  -v $LOCAL_NIM_CACHE:\u002Fopt\u002Fnim\u002F.cache \\\n  nvcr.io\u002Fnim\u002Fnvidia\u002F$CONTAINER_ID:latest\n```\n\nOmit `-v $LOCAL_NIM_CACHE:\u002Fopt\u002Fnim\u002F.cache` to skip caching (re-downloads model on every run).\n> **Security note:** `NGC_API_KEY` passed via `-e NGC_API_KEY` inherits from the shell environment. For production, use Docker secrets or a secrets manager instead of env vars; avoid storing API keys in shell history or plaintext config files.\n\n### RMIR Model (Export + Re-run Pattern)\n\n```bash\nexport NIM_EXPORT_PATH=~\u002Fnim_export\nmkdir -p $NIM_EXPORT_PATH && chmod 700 $NIM_EXPORT_PATH\nexport NIM_TAGS_SELECTOR=\"name=parakeet-1-1b-ctc-en-us,mode=str,model_type=rmir\"\n\n# Step 1: Export\ndocker run -it --rm --name=$CONTAINER_ID \\\n  --runtime=nvidia --gpus '\"device=0\"' --shm-size=8GB \\\n  -e NGC_API_KEY -e NIM_TAGS_SELECTOR \\\n  -e NIM_HTTP_API_PORT=9000 -e NIM_GRPC_API_PORT=50051 \\\n  -p 9000:9000 -p 50051:50051 \\\n  -v $NIM_EXPORT_PATH:\u002Fopt\u002Fnim\u002Fexport \\\n  -e NIM_EXPORT_PATH=\u002Fopt\u002Fnim\u002Fexport \\\n  nvcr.io\u002Fnim\u002Fnvidia\u002F$CONTAINER_ID:latest\n\n# Step 2: Run from export\ndocker run -it --rm --name=$CONTAINER_ID \\\n  --runtime=nvidia --gpus '\"device=0\"' --shm-size=8GB \\\n  -e NGC_API_KEY -e NIM_TAGS_SELECTOR \\\n  -e NIM_DISABLE_MODEL_DOWNLOAD=true \\\n  -e NIM_HTTP_API_PORT=9000 -e NIM_GRPC_API_PORT=50051 \\\n  -p 9000:9000 -p 50051:50051 \\\n  -v $NIM_EXPORT_PATH:\u002Fopt\u002Fnim\u002Fexport \\\n  -e NIM_EXPORT_PATH=\u002Fopt\u002Fnim\u002Fexport \\\n  nvcr.io\u002Fnim\u002Fnvidia\u002F$CONTAINER_ID:latest\n```\n\n## Step 3 — Verify Readiness\n\n```bash\ncurl -X GET http:\u002F\u002Flocalhost:9000\u002Fv1\u002Fhealth\u002Fready\n# Expected: {\"status\":\"ready\"}\n```\n\n## Step 4 — Run Inference\n\n### Streaming ASR (Python — gRPC)\n\n```bash\npython3 python-clients\u002Fscripts\u002Fasr\u002Ftranscribe_file.py \\\n  --server 0.0.0.0:50051 \\\n  --input-file \u002Fpath\u002Fto\u002Faudio.wav \\\n  --language-code en-US\n```\n\nWith diarization and word timestamps:\n\n```bash\npython3 python-clients\u002Fscripts\u002Fasr\u002Ftranscribe_file.py \\\n  --server 0.0.0.0:50051 \\\n  --input-file \u002Fpath\u002Fto\u002Faudio.wav \\\n  --language-code en-US \\\n  --speaker-diarization \\\n  --word-time-offsets\n```\n\nFor real-time microphone streaming:\n\n```bash\npython3 python-clients\u002Fscripts\u002Fasr\u002Ftranscribe_mic.py \\\n  --server 0.0.0.0:50051\n```\n\n### Offline Transcription (Python — gRPC)\n\n```bash\npython3 python-clients\u002Fscripts\u002Fasr\u002Ftranscribe_file_offline.py \\\n  --server 0.0.0.0:50051 \\\n  --input-file \u002Fpath\u002Fto\u002Faudio.wav \\\n  --language-code en-US\n```\n\n### HTTP API\n\n```bash\ncurl -X POST http:\u002F\u002Flocalhost:9000\u002Fv1\u002Faudio\u002Ftranscriptions \\\n  -H \"Content-Type: multipart\u002Fform-data\" \\\n  -F \"file=@\u002Fpath\u002Fto\u002Faudio.wav\" \\\n  -F \"model=asr\"\n```\n\n### C++ Client\n\n```bash\n# Build (requires Bazel in cpp-clients\u002F)\ncd cpp-clients\nbazel build \u002F\u002Friva\u002Fclients\u002Fasr:riva_asr_client\n\n# Run\n.\u002Fbazel-bin\u002Friva\u002Fclients\u002Fasr\u002Friva_asr_client \\\n  --server=0.0.0.0:50051 \\\n  --audio-file=\u002Fpath\u002Fto\u002Faudio.wav\n```\n\n### WebSocket (AudioCodes \u002F Telephony)\n\nThe `websocket-bridge` repo provides a Node.js bridge for AudioCodes-compatible telephony integrations:\n\n```bash\ncd websocket-bridge\nnpm install\nnode src\u002Findex.js --riva-server=localhost:50051\n```\n\n## Port Reference\n\n| Port | Protocol | Use |\n|------|----------|-----|\n| 9000 | HTTP | REST API, health check |\n| 50051 | gRPC | Python\u002FC++ client inference |\n\n## Examples\n\nThese are end-to-end use cases that combine features covered above; they are not repeats of the canonical commands in Option A or Step 4.\n\n**Whisper translation — non-English audio → English transcript (cloud):**\n```bash\npython python-clients\u002Fscripts\u002Fasr\u002Ftranscribe_file_offline.py \\\n    --server grpc.nvcf.nvidia.com:443 --use-ssl \\\n    --metadata function-id \"b702f636-f60c-4a3d-a6f4-f3568c13bd7d\" \\\n    --metadata authorization \"Bearer $NVIDIA_API_KEY\" \\\n    --custom-configuration \"task:translate\" \\\n    --input-file spanish_audio.wav\n```\n\n**Streaming with diarization and word timestamps (self-hosted):**\n```bash\npython3 python-clients\u002Fscripts\u002Fasr\u002Ftranscribe_file.py \\\n  --server 0.0.0.0:50051 --input-file meeting.wav --language-code en-US \\\n  --speaker-diarization --word-time-offsets\n```\n\n**Domain-specific vocabulary via token boosting (CTC, self-hosted):**\n```bash\npython3 python-clients\u002Fscripts\u002Fasr\u002Ftranscribe_file.py \\\n  --server 0.0.0.0:50051 --input-file clinical_dictation.wav \\\n  --language-code en-US \\\n  --boosted-lm-words 'Abloopar:▁a\u002Fb\u002Flo\u002Fop\u002Fer' \\\n  --boosted-lm-score 80\n```\n\n\n## Troubleshooting\n\n- **Wrong `NIM_TAGS_SELECTOR`** — if the selector doesn't match any available profile, the container exits. Check the support matrix for exact tag values.\n- **GPU device index** — `--gpus '\"device=0\"'` targets GPU 0. Adjust for multi-GPU hosts.\n- **Port 8000 conflict** — avoid `NIM_HTTP_API_PORT=8000`; use 9000 (default).\n- **Parakeet TDT and word timestamps** — only available via gRPC with `enable_word_time_offsets=True` in the recognition config.\n\n## Customization\n\n### Word Boosting\n\nBias the model toward domain-specific words at request time. Pass each word as a separate flag:\n\n```bash\npython scripts\u002Fasr\u002Ftranscribe_file.py \\\n  --server 0.0.0.0:50051 \\\n  --input-file audio.wav \\\n  --boosted-lm-words 'AntiBERTa' \\\n  --boosted-lm-words 'Abloopar' \\\n  --boosted-lm-score 20\n```\n\nBoost score ranges: **CTC**: 20–100 (negative scores discourage words) | **RNNT\u002FTDT**: 0.5–2.0 (single score for all words, no negatives, no OOV).\n\n### Token Boosting (CTC only)\n\nMaps the tokens the model *predicts* to the word you *want*. Use when word boosting alone can't overcome the acoustic model (e.g. multi-syllable OOV terms).\n\n**Step 1 — Get the model tokenizer from the running container:**\n\n```bash\ndocker exec \u003Ccontainer> find \u002Fdata\u002Fmodels -name \"*tokenizer.model\" | head -5\ndocker cp \u003Ccontainer>:\u002Fdata\u002Fmodels\u002F\u003Cmodel-dir>\u002F1\u002F\u003Chash>_tokenizer.model \u002Ftmp\u002Ftokenizer.model\n```\n\n**Step 2 — Generate token mapping with SentencePiece:**\n\n```python\nimport sentencepiece as spm\ns = spm.SentencePieceProcessor(model_file='\u002Ftmp\u002Ftokenizer.model')\n\nword_asr_predicts = \"ablooper\"       # what the model currently outputs\nword_asr_should_predict = \"Abloopar\" # what you want\n\ntokens = s.encode(word_asr_predicts, out_type=str)\nprint(word_asr_should_predict + ':' + '\u002F'.join(tokens))\n# e.g. Abloopar:▁a\u002Fb\u002Flo\u002Fop\u002Fer\n```\n\n**Step 3 — Pass mapping as boosted word:**\n\n```bash\n--boosted-lm-words 'Abloopar:▁a\u002Fb\u002Flo\u002Fop\u002Fer' --boosted-lm-score 80\n```\n\n**Important:** Always use the actual model tokenizer from inside the container — test tokenizers generate different token sequences and the mapping will silently fail.\n\n## Performance Benchmarking (Self-Hosted)\n\nUse `riva_streaming_asr_client` — a **pre-built binary available in PATH inside the NIM container**. Do not use the Python script from python-clients; run the binary via `docker exec`.\n\nA sample LibriSpeech wav file is bundled at `\u002Fopt\u002Friva\u002Fexamples\u002Fasr_lib\u002F1272-135031-0000.wav` inside the container.\n\n### Streaming Models\n\nRun at increasing concurrency levels (1, 2, 4, 8, …). Set `num_iterations` to 3× `num_parallel_requests` for stable results.\n\n```bash\nexport N=4  # num parallel streams — sweep: 1, 2, 4, 8, ...\n\ndocker exec \u003Ccontainer_name> riva_streaming_asr_client \\\n  --riva_uri=0.0.0.0:50051 \\\n  --language_code=en-US \\\n  --audio_file=\u002Fopt\u002Friva\u002Fexamples\u002Fasr_lib\u002F1272-135031-0000.wav \\\n  --chunk_duration_ms=160 \\\n  --simulate_realtime=true \\\n  --automatic_punctuation=true \\\n  --num_parallel_requests=$N \\\n  --num_iterations=$((3 * N)) \\\n  --word_time_offsets=false \\\n  --print_transcripts=false \\\n  --interim_results=false \\\n  --output_filename=\u002Ftmp\u002Foutput.json\n```\n\n### Offline Models\n\n```bash\nexport N=4  # num parallel requests — sweep: 1, 2, 4, 8, ...\n\ndocker exec \u003Ccontainer_name> riva_streaming_asr_client \\\n  --riva_uri=0.0.0.0:50051 \\\n  --language_code=en-US \\\n  --audio_file=\u002Fopt\u002Friva\u002Fexamples\u002Fasr_lib\u002F1272-135031-0000.wav \\\n  --automatic_punctuation=true \\\n  --num_parallel_requests=$N \\\n  --num_iterations=$((3 * N)) \\\n  --word_time_offsets=false \\\n  --print_transcripts=false \\\n  --interim_results=false \\\n  --output_filename=\u002Ftmp\u002Foutput.json\n```\n\nNote: Omit `--chunk_duration_ms` and `--simulate_realtime` for offline models — they process the full audio in one shot, not in streaming chunks.\n\n**Key flags:**\n\n| Flag | Description |\n|------|-------------|\n| `--chunk_duration_ms` | Match your deployed `chunk_size` in ms — streaming models only |\n| `--simulate_realtime` | Throttle audio to real-time speed — streaming models only |\n| `--num_parallel_requests` | Concurrent streams; sweep 1→2→4→8→… to find throughput peak |\n| `--num_iterations` | Total requests; use 3× `num_parallel_requests` for stable results |\n| `--print_transcripts=false` | Suppress transcripts for clean benchmark output |\n\n**Output metrics:**\n\n| Metric | Description |\n|--------|-------------|\n| Median \u002F 90th \u002F 95th \u002F 99th latency | Time from chunk sent to partial transcript received (ms) |\n| Throughput (RTFX) | Audio processed per second of wall time; >1.0 = faster than real-time |\n\n\n## Limitations\n\n- x86_64 architecture only — ARM is not supported\n- Self-hosted deployment requires an NVIDIA AI Enterprise license\n- Cloud-hosted inference requires an active `NVIDIA_API_KEY` and internet access\n- Word\u002Ftoken boosting is advisory — does not guarantee transcription of boosted terms\n- TDT models have a silence-based end-of-utterance threshold in streaming mode\n\n## Next Steps\n\n- Customize ASR pipeline (VAD, diarization, language model): see `riva-pipelines`\n- Deploy a custom-trained model: see `riva-asr-custom`\n- Check system requirements: see `riva-ops`\n\n",{"data":35,"body":52},{"name":4,"license":26,"description":6,"metadata":36},{"author":37,"team":38,"tags":39,"domain":50,"version":51},"Mayank Jain \u003Cmayjain@nvidia.com>","riva",[8,38,40,41,42,43,44,45,46,47,48,49],"nim","asr","speech-to-text","parakeet","canary","whisper","grpc","http","cloud","nvcf","ml","1.0.0",{"type":53,"children":54},"root",[55,64,94,98,105,110,116,135,141,206,212,317,323,357,389,411,419,599,618,636,661,668,680,1116,1119,1125,1163,1169,1192,1259,1275,1309,1315,1635,1648,1675,1681,2213,2219,2257,2263,2269,2339,2344,2432,2437,2472,2478,2542,2548,2650,2656,2749,2755,2768,2818,2824,2885,2891,2896,2904,3041,3049,3117,3125,3224,3230,3302,3308,3314,3319,3433,3452,3458,3477,3485,3649,3657,3735,3743,3779,3789,3795,3822,3835,3841,3862,4109,4115,4318,4338,4346,4466,4474,4521,4527,4562,4568,4604],{"type":56,"tag":57,"props":58,"children":60},"element","h1",{"id":59},"riva-asr-nim",[61],{"type":62,"value":63},"text","Riva ASR NIM",{"type":56,"tag":65,"props":66,"children":67},"blockquote",{},[68],{"type":56,"tag":69,"props":70,"children":71},"p",{},[72,78,80,85,87,92],{"type":56,"tag":73,"props":74,"children":75},"strong",{},[76],{"type":62,"value":77},"Agent:",{"type":62,"value":79}," When walking the user through a multi-step workflow, announce each step before presenting it: ",{"type":56,"tag":73,"props":81,"children":82},{},[83],{"type":62,"value":84},"Step N\u002FM — Step Title",{"type":62,"value":86}," (e.g., \"",{"type":56,"tag":73,"props":88,"children":89},{},[90],{"type":62,"value":91},"Step 1\u002F4 — Set Model Variables",{"type":62,"value":93},"\").",{"type":56,"tag":95,"props":96,"children":97},"hr",{},[],{"type":56,"tag":99,"props":100,"children":102},"h2",{"id":101},"purpose",[103],{"type":62,"value":104},"Purpose",{"type":56,"tag":69,"props":106,"children":107},{},[108],{"type":62,"value":109},"Deploy and run NVIDIA Riva ASR (speech-to-text) NIMs. Supports cloud-hosted\ninference via build.nvidia.com (no GPU required) and self-hosted deployment on\nyour own GPU using Docker. Covers streaming and offline transcription, word\nboosting, and performance benchmarking.",{"type":56,"tag":99,"props":111,"children":113},{"id":112},"workflow",[114],{"type":62,"value":115},"Workflow",{"type":56,"tag":69,"props":117,"children":118},{},[119,121,126,128,133],{"type":62,"value":120},"Choose ",{"type":56,"tag":73,"props":122,"children":123},{},[124],{"type":62,"value":125},"Option A",{"type":62,"value":127}," (cloud) for quick testing without a GPU, or ",{"type":56,"tag":73,"props":129,"children":130},{},[131],{"type":62,"value":132},"Option B",{"type":62,"value":134}," (self-hosted) for production. Self-hosted follows a 4-step process: set model variables → run container → verify health → run inference.",{"type":56,"tag":99,"props":136,"children":138},{"id":137},"prerequisites",[139],{"type":62,"value":140},"Prerequisites",{"type":56,"tag":142,"props":143,"children":144},"ul",{},[145,174,193],{"type":56,"tag":146,"props":147,"children":148},"li",{},[149,151,158,160,166,168],{"type":62,"value":150},"Complete ",{"type":56,"tag":152,"props":153,"children":155},"code",{"className":154},[],[156],{"type":62,"value":157},"riva-nim-setup",{"type":62,"value":159}," before self-hosted deployment: NVIDIA Container Toolkit, ",{"type":56,"tag":152,"props":161,"children":163},{"className":162},[],[164],{"type":62,"value":165},"NGC_API_KEY",{"type":62,"value":167}," exported, Docker logged in to ",{"type":56,"tag":152,"props":169,"children":171},{"className":170},[],[172],{"type":62,"value":173},"nvcr.io",{"type":56,"tag":146,"props":175,"children":176},{},[177,179,185,187],{"type":62,"value":178},"Cloud-hosted inference: ",{"type":56,"tag":152,"props":180,"children":182},{"className":181},[],[183],{"type":62,"value":184},"pip install -U nvidia-riva-client",{"type":62,"value":186}," and a valid ",{"type":56,"tag":152,"props":188,"children":190},{"className":189},[],[191],{"type":62,"value":192},"NVIDIA_API_KEY",{"type":56,"tag":146,"props":194,"children":195},{},[196,198,204],{"type":62,"value":197},"Not sure which model to use? Run ",{"type":56,"tag":152,"props":199,"children":201},{"className":200},[],[202],{"type":62,"value":203},"riva-model-selection",{"type":62,"value":205}," first",{"type":56,"tag":99,"props":207,"children":209},{"id":208},"instructions",[210],{"type":62,"value":211},"Instructions",{"type":56,"tag":213,"props":214,"children":215},"ol",{},[216,226,283],{"type":56,"tag":146,"props":217,"children":218},{},[219,224],{"type":56,"tag":73,"props":220,"children":221},{},[222],{"type":62,"value":223},"Pick a path",{"type":62,"value":225},": cloud inference (Option A) or self-hosted NIM (Option B).",{"type":56,"tag":146,"props":227,"children":228},{},[229,234,236,242,244,249,251,257,259,265,267,273,275,281],{"type":56,"tag":73,"props":230,"children":231},{},[232],{"type":62,"value":233},"For cloud inference (Option A)",{"type":62,"value":235},": install ",{"type":56,"tag":152,"props":237,"children":239},{"className":238},[],[240],{"type":62,"value":241},"nvidia-riva-client",{"type":62,"value":243},", set ",{"type":56,"tag":152,"props":245,"children":247},{"className":246},[],[248],{"type":62,"value":192},{"type":62,"value":250},", pick a function ID, run ",{"type":56,"tag":152,"props":252,"children":254},{"className":253},[],[255],{"type":62,"value":256},"transcribe_file.py",{"type":62,"value":258}," (streaming) or ",{"type":56,"tag":152,"props":260,"children":262},{"className":261},[],[263],{"type":62,"value":264},"transcribe_file_offline.py",{"type":62,"value":266}," (offline) against ",{"type":56,"tag":152,"props":268,"children":270},{"className":269},[],[271],{"type":62,"value":272},"grpc.nvcf.nvidia.com:443",{"type":62,"value":274}," with ",{"type":56,"tag":152,"props":276,"children":278},{"className":277},[],[279],{"type":62,"value":280},"--use-ssl",{"type":62,"value":282},".",{"type":56,"tag":146,"props":284,"children":285},{},[286,291,293,299,301,307,309,315],{"type":56,"tag":73,"props":287,"children":288},{},[289],{"type":62,"value":290},"For self-hosted (Option B)",{"type":62,"value":292},": look up ",{"type":56,"tag":152,"props":294,"children":296},{"className":295},[],[297],{"type":62,"value":298},"CONTAINER_ID",{"type":62,"value":300}," and ",{"type":56,"tag":152,"props":302,"children":304},{"className":303},[],[305],{"type":62,"value":306},"NIM_TAGS_SELECTOR",{"type":62,"value":308}," from the ASR support matrix, mount a model cache dir with ",{"type":56,"tag":152,"props":310,"children":312},{"className":311},[],[313],{"type":62,"value":314},"chmod 700",{"type":62,"value":316},", then follow Steps 1–4 below (set vars, run container, verify readiness, run inference).",{"type":56,"tag":99,"props":318,"children":320},{"id":319},"option-a-cloud-hosted-inference-buildnvidiacom",[321],{"type":62,"value":322},"Option A — Cloud-Hosted Inference (build.nvidia.com)",{"type":56,"tag":69,"props":324,"children":325},{},[326,331,333,338,340,348,349,355],{"type":56,"tag":73,"props":327,"children":328},{},[329],{"type":62,"value":330},"Setup:",{"type":62,"value":332}," ",{"type":56,"tag":152,"props":334,"children":336},{"className":335},[],[337],{"type":62,"value":184},{"type":62,"value":339}," + clone ",{"type":56,"tag":341,"props":342,"children":346},"a",{"href":343,"rel":344},"https:\u002F\u002Fgithub.com\u002Fnvidia-riva\u002Fpython-clients",[345],"nofollow",[347],{"type":62,"value":343},{"type":62,"value":300},{"type":56,"tag":152,"props":350,"children":352},{"className":351},[],[353],{"type":62,"value":354},"cd",{"type":62,"value":356}," into it.",{"type":56,"tag":69,"props":358,"children":359},{},[360,365,367,372,374,380,382,387],{"type":56,"tag":73,"props":361,"children":362},{},[363],{"type":62,"value":364},"Auth:",{"type":62,"value":366}," Set ",{"type":56,"tag":152,"props":368,"children":370},{"className":369},[],[371],{"type":62,"value":192},{"type":62,"value":373}," from ",{"type":56,"tag":341,"props":375,"children":378},{"href":376,"rel":377},"https:\u002F\u002Fbuild.nvidia.com",[345],[379],{"type":62,"value":376},{"type":62,"value":381}," (different from ",{"type":56,"tag":152,"props":383,"children":385},{"className":384},[],[386],{"type":62,"value":165},{"type":62,"value":388},").",{"type":56,"tag":69,"props":390,"children":391},{},[392,397,398,403,405,410],{"type":56,"tag":73,"props":393,"children":394},{},[395],{"type":62,"value":396},"Server:",{"type":62,"value":332},{"type":56,"tag":152,"props":399,"children":401},{"className":400},[],[402],{"type":62,"value":272},{"type":62,"value":404}," — always pass ",{"type":56,"tag":152,"props":406,"children":408},{"className":407},[],[409],{"type":62,"value":280},{"type":62,"value":282},{"type":56,"tag":69,"props":412,"children":413},{},[414],{"type":56,"tag":73,"props":415,"children":416},{},[417],{"type":62,"value":418},"Canonical command (streaming models):",{"type":56,"tag":420,"props":421,"children":426},"pre",{"className":422,"code":423,"language":424,"meta":425,"style":425},"language-bash shiki shiki-themes material-theme-lighter material-theme material-theme-palenight","python python-clients\u002Fscripts\u002Fasr\u002Ftranscribe_file.py \\\n    --server grpc.nvcf.nvidia.com:443 --use-ssl \\\n    --metadata function-id \"\u003CFUNCTION_ID>\" \\\n    --metadata \"authorization\" \"Bearer $NVIDIA_API_KEY\" \\\n    --language-code \u003CLANG_CODE> \\\n    --input-file \u002Fpath\u002Fto\u002Faudio.wav\n","bash","",[427],{"type":56,"tag":152,"props":428,"children":429},{"__ignoreMap":425},[430,453,476,510,553,586],{"type":56,"tag":431,"props":432,"children":434},"span",{"class":433,"line":27},"line",[435,441,447],{"type":56,"tag":431,"props":436,"children":438},{"style":437},"--shiki-light:#E2931D;--shiki-default:#FFCB6B;--shiki-dark:#FFCB6B",[439],{"type":62,"value":440},"python",{"type":56,"tag":431,"props":442,"children":444},{"style":443},"--shiki-light:#91B859;--shiki-default:#C3E88D;--shiki-dark:#C3E88D",[445],{"type":62,"value":446}," python-clients\u002Fscripts\u002Fasr\u002Ftranscribe_file.py",{"type":56,"tag":431,"props":448,"children":450},{"style":449},"--shiki-light:#90A4AE;--shiki-default:#EEFFFF;--shiki-dark:#BABED8",[451],{"type":62,"value":452}," \\\n",{"type":56,"tag":431,"props":454,"children":456},{"class":433,"line":455},2,[457,462,467,472],{"type":56,"tag":431,"props":458,"children":459},{"style":443},[460],{"type":62,"value":461},"    --server",{"type":56,"tag":431,"props":463,"children":464},{"style":443},[465],{"type":62,"value":466}," grpc.nvcf.nvidia.com:443",{"type":56,"tag":431,"props":468,"children":469},{"style":443},[470],{"type":62,"value":471}," --use-ssl",{"type":56,"tag":431,"props":473,"children":474},{"style":449},[475],{"type":62,"value":452},{"type":56,"tag":431,"props":477,"children":479},{"class":433,"line":478},3,[480,485,490,496,501,506],{"type":56,"tag":431,"props":481,"children":482},{"style":443},[483],{"type":62,"value":484},"    --metadata",{"type":56,"tag":431,"props":486,"children":487},{"style":443},[488],{"type":62,"value":489}," function-id",{"type":56,"tag":431,"props":491,"children":493},{"style":492},"--shiki-light:#39ADB5;--shiki-default:#89DDFF;--shiki-dark:#89DDFF",[494],{"type":62,"value":495}," \"",{"type":56,"tag":431,"props":497,"children":498},{"style":443},[499],{"type":62,"value":500},"\u003CFUNCTION_ID>",{"type":56,"tag":431,"props":502,"children":503},{"style":492},[504],{"type":62,"value":505},"\"",{"type":56,"tag":431,"props":507,"children":508},{"style":449},[509],{"type":62,"value":452},{"type":56,"tag":431,"props":511,"children":513},{"class":433,"line":512},4,[514,518,522,527,531,535,540,545,549],{"type":56,"tag":431,"props":515,"children":516},{"style":443},[517],{"type":62,"value":484},{"type":56,"tag":431,"props":519,"children":520},{"style":492},[521],{"type":62,"value":495},{"type":56,"tag":431,"props":523,"children":524},{"style":443},[525],{"type":62,"value":526},"authorization",{"type":56,"tag":431,"props":528,"children":529},{"style":492},[530],{"type":62,"value":505},{"type":56,"tag":431,"props":532,"children":533},{"style":492},[534],{"type":62,"value":495},{"type":56,"tag":431,"props":536,"children":537},{"style":443},[538],{"type":62,"value":539},"Bearer ",{"type":56,"tag":431,"props":541,"children":542},{"style":449},[543],{"type":62,"value":544},"$NVIDIA_API_KEY",{"type":56,"tag":431,"props":546,"children":547},{"style":492},[548],{"type":62,"value":505},{"type":56,"tag":431,"props":550,"children":551},{"style":449},[552],{"type":62,"value":452},{"type":56,"tag":431,"props":554,"children":556},{"class":433,"line":555},5,[557,562,567,572,577,582],{"type":56,"tag":431,"props":558,"children":559},{"style":443},[560],{"type":62,"value":561},"    --language-code",{"type":56,"tag":431,"props":563,"children":564},{"style":492},[565],{"type":62,"value":566}," \u003C",{"type":56,"tag":431,"props":568,"children":569},{"style":443},[570],{"type":62,"value":571},"LANG_COD",{"type":56,"tag":431,"props":573,"children":574},{"style":449},[575],{"type":62,"value":576},"E",{"type":56,"tag":431,"props":578,"children":579},{"style":492},[580],{"type":62,"value":581},">",{"type":56,"tag":431,"props":583,"children":584},{"style":449},[585],{"type":62,"value":452},{"type":56,"tag":431,"props":587,"children":588},{"class":433,"line":23},[589,594],{"type":56,"tag":431,"props":590,"children":591},{"style":443},[592],{"type":62,"value":593},"    --input-file",{"type":56,"tag":431,"props":595,"children":596},{"style":443},[597],{"type":62,"value":598}," \u002Fpath\u002Fto\u002Faudio.wav\n",{"type":56,"tag":69,"props":600,"children":601},{},[602,604,609,611,617],{"type":62,"value":603},"For offline-only models use ",{"type":56,"tag":152,"props":605,"children":607},{"className":606},[],[608],{"type":62,"value":264},{"type":62,"value":610}," instead. For Whisper translation add ",{"type":56,"tag":152,"props":612,"children":614},{"className":613},[],[615],{"type":62,"value":616},"--custom-configuration \"task:translate\"",{"type":62,"value":282},{"type":56,"tag":69,"props":619,"children":620},{},[621,626,628,634],{"type":56,"tag":73,"props":622,"children":623},{},[624],{"type":62,"value":625},"Word timestamps:",{"type":62,"value":627}," All models except Whisper and Canary support ",{"type":56,"tag":152,"props":629,"children":631},{"className":630},[],[632],{"type":62,"value":633},"--word-time-offsets",{"type":62,"value":635},". Accuracy ranking: CTC > TDT > RNNT.",{"type":56,"tag":69,"props":637,"children":638},{},[639,644,646,652,654,660],{"type":56,"tag":73,"props":640,"children":641},{},[642],{"type":62,"value":643},"Note:",{"type":62,"value":645}," Both cloud and self-hosted scripts use ",{"type":56,"tag":152,"props":647,"children":649},{"className":648},[],[650],{"type":62,"value":651},"--input-file",{"type":62,"value":653},", not ",{"type":56,"tag":152,"props":655,"children":657},{"className":656},[],[658],{"type":62,"value":659},"--audio-file",{"type":62,"value":282},{"type":56,"tag":662,"props":663,"children":665},"h3",{"id":664},"function-ids",[666],{"type":62,"value":667},"Function IDs",{"type":56,"tag":69,"props":669,"children":670},{},[671,673,679],{"type":62,"value":672},"If a function-id no longer works, fetch the current one from ",{"type":56,"tag":152,"props":674,"children":676},{"className":675},[],[677],{"type":62,"value":678},"https:\u002F\u002Fbuild.nvidia.com\u002F\u003Corg>\u002F\u003Cmodel>\u002Fapi",{"type":62,"value":282},{"type":56,"tag":681,"props":682,"children":683},"table",{},[684,718],{"type":56,"tag":685,"props":686,"children":687},"thead",{},[688],{"type":56,"tag":689,"props":690,"children":691},"tr",{},[692,698,703,708,713],{"type":56,"tag":693,"props":694,"children":695},"th",{},[696],{"type":62,"value":697},"Model",{"type":56,"tag":693,"props":699,"children":700},{},[701],{"type":62,"value":702},"Build Page slug",{"type":56,"tag":693,"props":704,"children":705},{},[706],{"type":62,"value":707},"Function ID",{"type":56,"tag":693,"props":709,"children":710},{},[711],{"type":62,"value":712},"Script",{"type":56,"tag":693,"props":714,"children":715},{},[716],{"type":62,"value":717},"Lang code",{"type":56,"tag":719,"props":720,"children":721},"tbody",{},[722,766,809,852,902,944,987,1030,1073],{"type":56,"tag":689,"props":723,"children":724},{},[725,731,740,749,757],{"type":56,"tag":726,"props":727,"children":728},"td",{},[729],{"type":62,"value":730},"Nemotron ASR Streaming",{"type":56,"tag":726,"props":732,"children":733},{},[734],{"type":56,"tag":152,"props":735,"children":737},{"className":736},[],[738],{"type":62,"value":739},"nvidia\u002Fnemotron-asr-streaming",{"type":56,"tag":726,"props":741,"children":742},{},[743],{"type":56,"tag":152,"props":744,"children":746},{"className":745},[],[747],{"type":62,"value":748},"bb0837de-8c7b-481f-9ec8-ef5663e9c1fa",{"type":56,"tag":726,"props":750,"children":751},{},[752],{"type":56,"tag":152,"props":753,"children":755},{"className":754},[],[756],{"type":62,"value":256},{"type":56,"tag":726,"props":758,"children":759},{},[760],{"type":56,"tag":761,"props":762,"children":763},"em",{},[764],{"type":62,"value":765},"(omit)",{"type":56,"tag":689,"props":767,"children":768},{},[769,774,783,792,800],{"type":56,"tag":726,"props":770,"children":771},{},[772],{"type":62,"value":773},"Parakeet 1.1B RNNT Multilingual",{"type":56,"tag":726,"props":775,"children":776},{},[777],{"type":56,"tag":152,"props":778,"children":780},{"className":779},[],[781],{"type":62,"value":782},"nvidia\u002Fparakeet-1_1b-rnnt-multilingual-asr",{"type":56,"tag":726,"props":784,"children":785},{},[786],{"type":56,"tag":152,"props":787,"children":789},{"className":788},[],[790],{"type":62,"value":791},"71203149-d3b7-4460-8231-1be2543a1fca",{"type":56,"tag":726,"props":793,"children":794},{},[795],{"type":56,"tag":152,"props":796,"children":798},{"className":797},[],[799],{"type":62,"value":256},{"type":56,"tag":726,"props":801,"children":802},{},[803],{"type":56,"tag":152,"props":804,"children":806},{"className":805},[],[807],{"type":62,"value":808},"multi",{"type":56,"tag":689,"props":810,"children":811},{},[812,817,826,835,843],{"type":56,"tag":726,"props":813,"children":814},{},[815],{"type":62,"value":816},"Parakeet TDT 0.6B v2",{"type":56,"tag":726,"props":818,"children":819},{},[820],{"type":56,"tag":152,"props":821,"children":823},{"className":822},[],[824],{"type":62,"value":825},"nvidia\u002Fparakeet-tdt-0_6b-v2",{"type":56,"tag":726,"props":827,"children":828},{},[829],{"type":56,"tag":152,"props":830,"children":832},{"className":831},[],[833],{"type":62,"value":834},"d3fe9151-442b-4204-a70d-5fcc597fd610",{"type":56,"tag":726,"props":836,"children":837},{},[838],{"type":56,"tag":152,"props":839,"children":841},{"className":840},[],[842],{"type":62,"value":264},{"type":56,"tag":726,"props":844,"children":845},{},[846],{"type":56,"tag":152,"props":847,"children":849},{"className":848},[],[850],{"type":62,"value":851},"en-US",{"type":56,"tag":689,"props":853,"children":854},{},[855,860,869,878,886],{"type":56,"tag":726,"props":856,"children":857},{},[858],{"type":62,"value":859},"Whisper Large v3",{"type":56,"tag":726,"props":861,"children":862},{},[863],{"type":56,"tag":152,"props":864,"children":866},{"className":865},[],[867],{"type":62,"value":868},"openai\u002Fwhisper-large-v3",{"type":56,"tag":726,"props":870,"children":871},{},[872],{"type":56,"tag":152,"props":873,"children":875},{"className":874},[],[876],{"type":62,"value":877},"b702f636-f60c-4a3d-a6f4-f3568c13bd7d",{"type":56,"tag":726,"props":879,"children":880},{},[881],{"type":56,"tag":152,"props":882,"children":884},{"className":883},[],[885],{"type":62,"value":264},{"type":56,"tag":726,"props":887,"children":888},{},[889,895,897],{"type":56,"tag":152,"props":890,"children":892},{"className":891},[],[893],{"type":62,"value":894},"en",{"type":62,"value":896}," \u002F ",{"type":56,"tag":152,"props":898,"children":900},{"className":899},[],[901],{"type":62,"value":808},{"type":56,"tag":689,"props":903,"children":904},{},[905,910,919,928,936],{"type":56,"tag":726,"props":906,"children":907},{},[908],{"type":62,"value":909},"Parakeet CTC 1.1B English",{"type":56,"tag":726,"props":911,"children":912},{},[913],{"type":56,"tag":152,"props":914,"children":916},{"className":915},[],[917],{"type":62,"value":918},"nvidia\u002Fparakeet-ctc-1_1b-asr",{"type":56,"tag":726,"props":920,"children":921},{},[922],{"type":56,"tag":152,"props":923,"children":925},{"className":924},[],[926],{"type":62,"value":927},"1598d209-5e27-4d3c-8079-4751568b1081",{"type":56,"tag":726,"props":929,"children":930},{},[931],{"type":56,"tag":152,"props":932,"children":934},{"className":933},[],[935],{"type":62,"value":256},{"type":56,"tag":726,"props":937,"children":938},{},[939],{"type":56,"tag":152,"props":940,"children":942},{"className":941},[],[943],{"type":62,"value":851},{"type":56,"tag":689,"props":945,"children":946},{},[947,952,961,970,978],{"type":56,"tag":726,"props":948,"children":949},{},[950],{"type":62,"value":951},"Parakeet CTC 0.6B Spanish",{"type":56,"tag":726,"props":953,"children":954},{},[955],{"type":56,"tag":152,"props":956,"children":958},{"className":957},[],[959],{"type":62,"value":960},"nvidia\u002Fparakeet-ctc-0_6b-es",{"type":56,"tag":726,"props":962,"children":963},{},[964],{"type":56,"tag":152,"props":965,"children":967},{"className":966},[],[968],{"type":62,"value":969},"a9eeee8f-b509-4712-b19d-194361fa5f31",{"type":56,"tag":726,"props":971,"children":972},{},[973],{"type":56,"tag":152,"props":974,"children":976},{"className":975},[],[977],{"type":62,"value":256},{"type":56,"tag":726,"props":979,"children":980},{},[981],{"type":56,"tag":152,"props":982,"children":984},{"className":983},[],[985],{"type":62,"value":986},"es-US",{"type":56,"tag":689,"props":988,"children":989},{},[990,995,1004,1013,1021],{"type":56,"tag":726,"props":991,"children":992},{},[993],{"type":62,"value":994},"Parakeet CTC 0.6B Vietnamese",{"type":56,"tag":726,"props":996,"children":997},{},[998],{"type":56,"tag":152,"props":999,"children":1001},{"className":1000},[],[1002],{"type":62,"value":1003},"nvidia\u002Fparakeet-ctc-0_6b-vi",{"type":56,"tag":726,"props":1005,"children":1006},{},[1007],{"type":56,"tag":152,"props":1008,"children":1010},{"className":1009},[],[1011],{"type":62,"value":1012},"f3dff2bb-99f9-403d-a5f1-f574a757deb0",{"type":56,"tag":726,"props":1014,"children":1015},{},[1016],{"type":56,"tag":152,"props":1017,"children":1019},{"className":1018},[],[1020],{"type":62,"value":256},{"type":56,"tag":726,"props":1022,"children":1023},{},[1024],{"type":56,"tag":152,"props":1025,"children":1027},{"className":1026},[],[1028],{"type":62,"value":1029},"vi-VN",{"type":56,"tag":689,"props":1031,"children":1032},{},[1033,1038,1047,1056,1064],{"type":56,"tag":726,"props":1034,"children":1035},{},[1036],{"type":62,"value":1037},"Parakeet CTC 0.6B Mandarin (Simplified)",{"type":56,"tag":726,"props":1039,"children":1040},{},[1041],{"type":56,"tag":152,"props":1042,"children":1044},{"className":1043},[],[1045],{"type":62,"value":1046},"nvidia\u002Fparakeet-ctc-0_6b-zh-cn",{"type":56,"tag":726,"props":1048,"children":1049},{},[1050],{"type":56,"tag":152,"props":1051,"children":1053},{"className":1052},[],[1054],{"type":62,"value":1055},"9add5ef7-322e-47e0-ad7a-5653fb8d259b",{"type":56,"tag":726,"props":1057,"children":1058},{},[1059],{"type":56,"tag":152,"props":1060,"children":1062},{"className":1061},[],[1063],{"type":62,"value":256},{"type":56,"tag":726,"props":1065,"children":1066},{},[1067],{"type":56,"tag":152,"props":1068,"children":1070},{"className":1069},[],[1071],{"type":62,"value":1072},"zh-CN",{"type":56,"tag":689,"props":1074,"children":1075},{},[1076,1081,1090,1099,1107],{"type":56,"tag":726,"props":1077,"children":1078},{},[1079],{"type":62,"value":1080},"Parakeet CTC 0.6B Taiwanese (Traditional)",{"type":56,"tag":726,"props":1082,"children":1083},{},[1084],{"type":56,"tag":152,"props":1085,"children":1087},{"className":1086},[],[1088],{"type":62,"value":1089},"nvidia\u002Fparakeet-ctc-0_6b-zh-tw",{"type":56,"tag":726,"props":1091,"children":1092},{},[1093],{"type":56,"tag":152,"props":1094,"children":1096},{"className":1095},[],[1097],{"type":62,"value":1098},"8473f56d-51ef-473c-bb26-efd4f5def2bf",{"type":56,"tag":726,"props":1100,"children":1101},{},[1102],{"type":56,"tag":152,"props":1103,"children":1105},{"className":1104},[],[1106],{"type":62,"value":256},{"type":56,"tag":726,"props":1108,"children":1109},{},[1110],{"type":56,"tag":152,"props":1111,"children":1113},{"className":1112},[],[1114],{"type":62,"value":1115},"zh-TW",{"type":56,"tag":95,"props":1117,"children":1118},{},[],{"type":56,"tag":99,"props":1120,"children":1122},{"id":1121},"option-b-self-hosted-nim-deployment",[1123],{"type":62,"value":1124},"Option B — Self-Hosted NIM Deployment",{"type":56,"tag":69,"props":1126,"children":1127},{},[1128,1129,1134,1136,1141,1142,1147,1149,1155,1157],{"type":62,"value":150},{"type":56,"tag":152,"props":1130,"children":1132},{"className":1131},[],[1133],{"type":62,"value":157},{"type":62,"value":1135}," first: NVIDIA Container Toolkit, ",{"type":56,"tag":152,"props":1137,"children":1139},{"className":1138},[],[1140],{"type":62,"value":165},{"type":62,"value":167},{"type":56,"tag":152,"props":1143,"children":1145},{"className":1144},[],[1146],{"type":62,"value":173},{"type":62,"value":1148},". Driver and VRAM minimums: see ",{"type":56,"tag":341,"props":1150,"children":1153},{"href":1151,"rel":1152},"https:\u002F\u002Fdocs.nvidia.com\u002Fnim\u002Fspeech\u002Flatest\u002Freference\u002Fsupport-matrix\u002Fasr.html",[345],[1154],{"type":62,"value":1151},{"type":62,"value":1156},". Latency and throughput benchmarks: ",{"type":56,"tag":341,"props":1158,"children":1161},{"href":1159,"rel":1160},"https:\u002F\u002Fdocs.nvidia.com\u002Fnim\u002Fspeech\u002Flatest\u002Freference\u002Fperformances\u002Fasr\u002Fperformance.html",[345],[1162],{"type":62,"value":1159},{"type":56,"tag":99,"props":1164,"children":1166},{"id":1165},"step-1-set-model-variables",[1167],{"type":62,"value":1168},"Step 1 — Set Model Variables",{"type":56,"tag":69,"props":1170,"children":1171},{},[1172,1174,1179,1180,1185,1187],{"type":62,"value":1173},"Get the current ",{"type":56,"tag":152,"props":1175,"children":1177},{"className":1176},[],[1178],{"type":62,"value":298},{"type":62,"value":300},{"type":56,"tag":152,"props":1181,"children":1183},{"className":1182},[],[1184],{"type":62,"value":306},{"type":62,"value":1186}," values from the ASR support matrix (includes all models, modes, VRAM, and deployment profiles):\n",{"type":56,"tag":341,"props":1188,"children":1190},{"href":1151,"rel":1189},[345],[1191],{"type":62,"value":1151},{"type":56,"tag":420,"props":1193,"children":1195},{"className":422,"code":1194,"language":424,"meta":425,"style":425},"export CONTAINER_ID=\u003Ccontainer-id-from-support-matrix>\nexport NIM_TAGS_SELECTOR=\"\u003Cselector-from-support-matrix>\"\n",[1196],{"type":56,"tag":152,"props":1197,"children":1198},{"__ignoreMap":425},[1199,1228],{"type":56,"tag":431,"props":1200,"children":1201},{"class":433,"line":27},[1202,1208,1213,1218,1223],{"type":56,"tag":431,"props":1203,"children":1205},{"style":1204},"--shiki-light:#9C3EDA;--shiki-default:#C792EA;--shiki-dark:#C792EA",[1206],{"type":62,"value":1207},"export",{"type":56,"tag":431,"props":1209,"children":1210},{"style":449},[1211],{"type":62,"value":1212}," CONTAINER_ID",{"type":56,"tag":431,"props":1214,"children":1215},{"style":492},[1216],{"type":62,"value":1217},"=\u003C",{"type":56,"tag":431,"props":1219,"children":1220},{"style":449},[1221],{"type":62,"value":1222},"container-id-from-support-matrix",{"type":56,"tag":431,"props":1224,"children":1225},{"style":492},[1226],{"type":62,"value":1227},">\n",{"type":56,"tag":431,"props":1229,"children":1230},{"class":433,"line":455},[1231,1235,1240,1245,1249,1254],{"type":56,"tag":431,"props":1232,"children":1233},{"style":1204},[1234],{"type":62,"value":1207},{"type":56,"tag":431,"props":1236,"children":1237},{"style":449},[1238],{"type":62,"value":1239}," NIM_TAGS_SELECTOR",{"type":56,"tag":431,"props":1241,"children":1242},{"style":492},[1243],{"type":62,"value":1244},"=",{"type":56,"tag":431,"props":1246,"children":1247},{"style":492},[1248],{"type":62,"value":505},{"type":56,"tag":431,"props":1250,"children":1251},{"style":443},[1252],{"type":62,"value":1253},"\u003Cselector-from-support-matrix>",{"type":56,"tag":431,"props":1255,"children":1256},{"style":492},[1257],{"type":62,"value":1258},"\"\n",{"type":56,"tag":69,"props":1260,"children":1261},{},[1262,1267,1269],{"type":56,"tag":152,"props":1263,"children":1265},{"className":1264},[],[1266],{"type":62,"value":306},{"type":62,"value":1268}," pattern: ",{"type":56,"tag":152,"props":1270,"children":1272},{"className":1271},[],[1273],{"type":62,"value":1274},"name=\u003Cmodel-name>,mode=\u003Cstr|offline|all>[,model_type=\u003Cprebuilt|rmir>]",{"type":56,"tag":69,"props":1276,"children":1277},{},[1278,1283,1285,1291,1293,1299,1301,1307],{"type":56,"tag":73,"props":1279,"children":1280},{},[1281],{"type":62,"value":1282},"Prebuilt vs RMIR:",{"type":62,"value":1284}," The NIM auto-detects your GPU on startup. For well-known GPUs (A100, H100, select Blackwell), it pulls a prebuilt model repo — a tarball of TensorRT engines already compiled for that GPU via ",{"type":56,"tag":152,"props":1286,"children":1288},{"className":1287},[],[1289],{"type":62,"value":1290},"riva-build",{"type":62,"value":1292}," + ",{"type":56,"tag":152,"props":1294,"children":1296},{"className":1295},[],[1297],{"type":62,"value":1298},"riva-deploy",{"type":62,"value":1300},". For unsupported GPUs, it falls back to RMIR (Riva Model Intermediate Representation), which is a portable format that gets compiled into TensorRT engines on your local GPU at first run (slower startup, same runtime performance). You rarely need to set ",{"type":56,"tag":152,"props":1302,"children":1304},{"className":1303},[],[1305],{"type":62,"value":1306},"model_type",{"type":62,"value":1308}," explicitly — omit it and the NIM picks the right one automatically.",{"type":56,"tag":99,"props":1310,"children":1312},{"id":1311},"step-2-run-the-container",[1313],{"type":62,"value":1314},"Step 2 — Run the Container",{"type":56,"tag":420,"props":1316,"children":1318},{"className":422,"code":1317,"language":424,"meta":425,"style":425},"export LOCAL_NIM_CACHE=~\u002F.cache\u002Fnim\nmkdir -p $LOCAL_NIM_CACHE && chmod 700 $LOCAL_NIM_CACHE\n\ndocker run -it --rm --name=$CONTAINER_ID \\\n  --runtime=nvidia \\\n  --gpus '\"device=0\"' \\\n  --shm-size=8GB \\\n  -e NGC_API_KEY \\\n  -e NIM_TAGS_SELECTOR \\\n  -e NIM_HTTP_API_PORT=9000 \\\n  -e NIM_GRPC_API_PORT=50051 \\\n  -p 9000:9000 \\\n  -p 50051:50051 \\\n  -v $LOCAL_NIM_CACHE:\u002Fopt\u002Fnim\u002F.cache \\\n  nvcr.io\u002Fnim\u002Fnvidia\u002F$CONTAINER_ID:latest\n",[1319],{"type":56,"tag":152,"props":1320,"children":1321},{"__ignoreMap":425},[1322,1344,1383,1392,1429,1441,1468,1481,1499,1515,1537,1559,1577,1594,1617],{"type":56,"tag":431,"props":1323,"children":1324},{"class":433,"line":27},[1325,1329,1334,1339],{"type":56,"tag":431,"props":1326,"children":1327},{"style":1204},[1328],{"type":62,"value":1207},{"type":56,"tag":431,"props":1330,"children":1331},{"style":449},[1332],{"type":62,"value":1333}," LOCAL_NIM_CACHE",{"type":56,"tag":431,"props":1335,"children":1336},{"style":492},[1337],{"type":62,"value":1338},"=~",{"type":56,"tag":431,"props":1340,"children":1341},{"style":449},[1342],{"type":62,"value":1343},"\u002F.cache\u002Fnim\n",{"type":56,"tag":431,"props":1345,"children":1346},{"class":433,"line":455},[1347,1352,1357,1362,1367,1372,1378],{"type":56,"tag":431,"props":1348,"children":1349},{"style":437},[1350],{"type":62,"value":1351},"mkdir",{"type":56,"tag":431,"props":1353,"children":1354},{"style":443},[1355],{"type":62,"value":1356}," -p",{"type":56,"tag":431,"props":1358,"children":1359},{"style":449},[1360],{"type":62,"value":1361}," $LOCAL_NIM_CACHE ",{"type":56,"tag":431,"props":1363,"children":1364},{"style":492},[1365],{"type":62,"value":1366},"&&",{"type":56,"tag":431,"props":1368,"children":1369},{"style":437},[1370],{"type":62,"value":1371}," chmod",{"type":56,"tag":431,"props":1373,"children":1375},{"style":1374},"--shiki-light:#F76D47;--shiki-default:#F78C6C;--shiki-dark:#F78C6C",[1376],{"type":62,"value":1377}," 700",{"type":56,"tag":431,"props":1379,"children":1380},{"style":449},[1381],{"type":62,"value":1382}," $LOCAL_NIM_CACHE\n",{"type":56,"tag":431,"props":1384,"children":1385},{"class":433,"line":478},[1386],{"type":56,"tag":431,"props":1387,"children":1389},{"emptyLinePlaceholder":1388},true,[1390],{"type":62,"value":1391},"\n",{"type":56,"tag":431,"props":1393,"children":1394},{"class":433,"line":512},[1395,1400,1405,1410,1415,1420,1425],{"type":56,"tag":431,"props":1396,"children":1397},{"style":437},[1398],{"type":62,"value":1399},"docker",{"type":56,"tag":431,"props":1401,"children":1402},{"style":443},[1403],{"type":62,"value":1404}," run",{"type":56,"tag":431,"props":1406,"children":1407},{"style":443},[1408],{"type":62,"value":1409}," -it",{"type":56,"tag":431,"props":1411,"children":1412},{"style":443},[1413],{"type":62,"value":1414}," --rm",{"type":56,"tag":431,"props":1416,"children":1417},{"style":443},[1418],{"type":62,"value":1419}," --name=",{"type":56,"tag":431,"props":1421,"children":1422},{"style":449},[1423],{"type":62,"value":1424},"$CONTAINER_ID",{"type":56,"tag":431,"props":1426,"children":1427},{"style":449},[1428],{"type":62,"value":452},{"type":56,"tag":431,"props":1430,"children":1431},{"class":433,"line":555},[1432,1437],{"type":56,"tag":431,"props":1433,"children":1434},{"style":443},[1435],{"type":62,"value":1436},"  --runtime=nvidia",{"type":56,"tag":431,"props":1438,"children":1439},{"style":449},[1440],{"type":62,"value":452},{"type":56,"tag":431,"props":1442,"children":1443},{"class":433,"line":23},[1444,1449,1454,1459,1464],{"type":56,"tag":431,"props":1445,"children":1446},{"style":443},[1447],{"type":62,"value":1448},"  --gpus",{"type":56,"tag":431,"props":1450,"children":1451},{"style":492},[1452],{"type":62,"value":1453}," '",{"type":56,"tag":431,"props":1455,"children":1456},{"style":443},[1457],{"type":62,"value":1458},"\"device=0\"",{"type":56,"tag":431,"props":1460,"children":1461},{"style":492},[1462],{"type":62,"value":1463},"'",{"type":56,"tag":431,"props":1465,"children":1466},{"style":449},[1467],{"type":62,"value":452},{"type":56,"tag":431,"props":1469,"children":1471},{"class":433,"line":1470},7,[1472,1477],{"type":56,"tag":431,"props":1473,"children":1474},{"style":443},[1475],{"type":62,"value":1476},"  --shm-size=8GB",{"type":56,"tag":431,"props":1478,"children":1479},{"style":449},[1480],{"type":62,"value":452},{"type":56,"tag":431,"props":1482,"children":1484},{"class":433,"line":1483},8,[1485,1490,1495],{"type":56,"tag":431,"props":1486,"children":1487},{"style":443},[1488],{"type":62,"value":1489},"  -e",{"type":56,"tag":431,"props":1491,"children":1492},{"style":443},[1493],{"type":62,"value":1494}," NGC_API_KEY",{"type":56,"tag":431,"props":1496,"children":1497},{"style":449},[1498],{"type":62,"value":452},{"type":56,"tag":431,"props":1500,"children":1502},{"class":433,"line":1501},9,[1503,1507,1511],{"type":56,"tag":431,"props":1504,"children":1505},{"style":443},[1506],{"type":62,"value":1489},{"type":56,"tag":431,"props":1508,"children":1509},{"style":443},[1510],{"type":62,"value":1239},{"type":56,"tag":431,"props":1512,"children":1513},{"style":449},[1514],{"type":62,"value":452},{"type":56,"tag":431,"props":1516,"children":1518},{"class":433,"line":1517},10,[1519,1523,1528,1533],{"type":56,"tag":431,"props":1520,"children":1521},{"style":443},[1522],{"type":62,"value":1489},{"type":56,"tag":431,"props":1524,"children":1525},{"style":443},[1526],{"type":62,"value":1527}," NIM_HTTP_API_PORT=",{"type":56,"tag":431,"props":1529,"children":1530},{"style":1374},[1531],{"type":62,"value":1532},"9000",{"type":56,"tag":431,"props":1534,"children":1535},{"style":449},[1536],{"type":62,"value":452},{"type":56,"tag":431,"props":1538,"children":1540},{"class":433,"line":1539},11,[1541,1545,1550,1555],{"type":56,"tag":431,"props":1542,"children":1543},{"style":443},[1544],{"type":62,"value":1489},{"type":56,"tag":431,"props":1546,"children":1547},{"style":443},[1548],{"type":62,"value":1549}," NIM_GRPC_API_PORT=",{"type":56,"tag":431,"props":1551,"children":1552},{"style":1374},[1553],{"type":62,"value":1554},"50051",{"type":56,"tag":431,"props":1556,"children":1557},{"style":449},[1558],{"type":62,"value":452},{"type":56,"tag":431,"props":1560,"children":1562},{"class":433,"line":1561},12,[1563,1568,1573],{"type":56,"tag":431,"props":1564,"children":1565},{"style":443},[1566],{"type":62,"value":1567},"  -p",{"type":56,"tag":431,"props":1569,"children":1570},{"style":443},[1571],{"type":62,"value":1572}," 9000:9000",{"type":56,"tag":431,"props":1574,"children":1575},{"style":449},[1576],{"type":62,"value":452},{"type":56,"tag":431,"props":1578,"children":1580},{"class":433,"line":1579},13,[1581,1585,1590],{"type":56,"tag":431,"props":1582,"children":1583},{"style":443},[1584],{"type":62,"value":1567},{"type":56,"tag":431,"props":1586,"children":1587},{"style":443},[1588],{"type":62,"value":1589}," 50051:50051",{"type":56,"tag":431,"props":1591,"children":1592},{"style":449},[1593],{"type":62,"value":452},{"type":56,"tag":431,"props":1595,"children":1597},{"class":433,"line":1596},14,[1598,1603,1608,1613],{"type":56,"tag":431,"props":1599,"children":1600},{"style":443},[1601],{"type":62,"value":1602},"  -v",{"type":56,"tag":431,"props":1604,"children":1605},{"style":449},[1606],{"type":62,"value":1607}," $LOCAL_NIM_CACHE",{"type":56,"tag":431,"props":1609,"children":1610},{"style":443},[1611],{"type":62,"value":1612},":\u002Fopt\u002Fnim\u002F.cache",{"type":56,"tag":431,"props":1614,"children":1615},{"style":449},[1616],{"type":62,"value":452},{"type":56,"tag":431,"props":1618,"children":1620},{"class":433,"line":1619},15,[1621,1626,1630],{"type":56,"tag":431,"props":1622,"children":1623},{"style":443},[1624],{"type":62,"value":1625},"  nvcr.io\u002Fnim\u002Fnvidia\u002F",{"type":56,"tag":431,"props":1627,"children":1628},{"style":449},[1629],{"type":62,"value":1424},{"type":56,"tag":431,"props":1631,"children":1632},{"style":443},[1633],{"type":62,"value":1634},":latest\n",{"type":56,"tag":69,"props":1636,"children":1637},{},[1638,1640,1646],{"type":62,"value":1639},"Omit ",{"type":56,"tag":152,"props":1641,"children":1643},{"className":1642},[],[1644],{"type":62,"value":1645},"-v $LOCAL_NIM_CACHE:\u002Fopt\u002Fnim\u002F.cache",{"type":62,"value":1647}," to skip caching (re-downloads model on every run).",{"type":56,"tag":65,"props":1649,"children":1650},{},[1651],{"type":56,"tag":69,"props":1652,"children":1653},{},[1654,1659,1660,1665,1667,1673],{"type":56,"tag":73,"props":1655,"children":1656},{},[1657],{"type":62,"value":1658},"Security note:",{"type":62,"value":332},{"type":56,"tag":152,"props":1661,"children":1663},{"className":1662},[],[1664],{"type":62,"value":165},{"type":62,"value":1666}," passed via ",{"type":56,"tag":152,"props":1668,"children":1670},{"className":1669},[],[1671],{"type":62,"value":1672},"-e NGC_API_KEY",{"type":62,"value":1674}," inherits from the shell environment. For production, use Docker secrets or a secrets manager instead of env vars; avoid storing API keys in shell history or plaintext config files.",{"type":56,"tag":662,"props":1676,"children":1678},{"id":1677},"rmir-model-export-re-run-pattern",[1679],{"type":62,"value":1680},"RMIR Model (Export + Re-run Pattern)",{"type":56,"tag":420,"props":1682,"children":1684},{"className":422,"code":1683,"language":424,"meta":425,"style":425},"export NIM_EXPORT_PATH=~\u002Fnim_export\nmkdir -p $NIM_EXPORT_PATH && chmod 700 $NIM_EXPORT_PATH\nexport NIM_TAGS_SELECTOR=\"name=parakeet-1-1b-ctc-en-us,mode=str,model_type=rmir\"\n\n# Step 1: Export\ndocker run -it --rm --name=$CONTAINER_ID \\\n  --runtime=nvidia --gpus '\"device=0\"' --shm-size=8GB \\\n  -e NGC_API_KEY -e NIM_TAGS_SELECTOR \\\n  -e NIM_HTTP_API_PORT=9000 -e NIM_GRPC_API_PORT=50051 \\\n  -p 9000:9000 -p 50051:50051 \\\n  -v $NIM_EXPORT_PATH:\u002Fopt\u002Fnim\u002Fexport \\\n  -e NIM_EXPORT_PATH=\u002Fopt\u002Fnim\u002Fexport \\\n  nvcr.io\u002Fnim\u002Fnvidia\u002F$CONTAINER_ID:latest\n\n# Step 2: Run from export\ndocker run -it --rm --name=$CONTAINER_ID \\\n  --runtime=nvidia --gpus '\"device=0\"' --shm-size=8GB \\\n  -e NGC_API_KEY -e NIM_TAGS_SELECTOR \\\n  -e NIM_DISABLE_MODEL_DOWNLOAD=true \\\n  -e NIM_HTTP_API_PORT=9000 -e NIM_GRPC_API_PORT=50051 \\\n  -p 9000:9000 -p 50051:50051 \\\n  -v $NIM_EXPORT_PATH:\u002Fopt\u002Fnim\u002Fexport \\\n  -e NIM_EXPORT_PATH=\u002Fopt\u002Fnim\u002Fexport \\\n  nvcr.io\u002Fnim\u002Fnvidia\u002F$CONTAINER_ID:latest\n",[1685],{"type":56,"tag":152,"props":1686,"children":1687},{"__ignoreMap":425},[1688,1709,1742,1770,1777,1786,1817,1850,1874,1905,1928,1949,1965,1980,1987,1995,2027,2059,2083,2105,2137,2161,2181,2197],{"type":56,"tag":431,"props":1689,"children":1690},{"class":433,"line":27},[1691,1695,1700,1704],{"type":56,"tag":431,"props":1692,"children":1693},{"style":1204},[1694],{"type":62,"value":1207},{"type":56,"tag":431,"props":1696,"children":1697},{"style":449},[1698],{"type":62,"value":1699}," NIM_EXPORT_PATH",{"type":56,"tag":431,"props":1701,"children":1702},{"style":492},[1703],{"type":62,"value":1338},{"type":56,"tag":431,"props":1705,"children":1706},{"style":449},[1707],{"type":62,"value":1708},"\u002Fnim_export\n",{"type":56,"tag":431,"props":1710,"children":1711},{"class":433,"line":455},[1712,1716,1720,1725,1729,1733,1737],{"type":56,"tag":431,"props":1713,"children":1714},{"style":437},[1715],{"type":62,"value":1351},{"type":56,"tag":431,"props":1717,"children":1718},{"style":443},[1719],{"type":62,"value":1356},{"type":56,"tag":431,"props":1721,"children":1722},{"style":449},[1723],{"type":62,"value":1724}," $NIM_EXPORT_PATH ",{"type":56,"tag":431,"props":1726,"children":1727},{"style":492},[1728],{"type":62,"value":1366},{"type":56,"tag":431,"props":1730,"children":1731},{"style":437},[1732],{"type":62,"value":1371},{"type":56,"tag":431,"props":1734,"children":1735},{"style":1374},[1736],{"type":62,"value":1377},{"type":56,"tag":431,"props":1738,"children":1739},{"style":449},[1740],{"type":62,"value":1741}," $NIM_EXPORT_PATH\n",{"type":56,"tag":431,"props":1743,"children":1744},{"class":433,"line":478},[1745,1749,1753,1757,1761,1766],{"type":56,"tag":431,"props":1746,"children":1747},{"style":1204},[1748],{"type":62,"value":1207},{"type":56,"tag":431,"props":1750,"children":1751},{"style":449},[1752],{"type":62,"value":1239},{"type":56,"tag":431,"props":1754,"children":1755},{"style":492},[1756],{"type":62,"value":1244},{"type":56,"tag":431,"props":1758,"children":1759},{"style":492},[1760],{"type":62,"value":505},{"type":56,"tag":431,"props":1762,"children":1763},{"style":443},[1764],{"type":62,"value":1765},"name=parakeet-1-1b-ctc-en-us,mode=str,model_type=rmir",{"type":56,"tag":431,"props":1767,"children":1768},{"style":492},[1769],{"type":62,"value":1258},{"type":56,"tag":431,"props":1771,"children":1772},{"class":433,"line":512},[1773],{"type":56,"tag":431,"props":1774,"children":1775},{"emptyLinePlaceholder":1388},[1776],{"type":62,"value":1391},{"type":56,"tag":431,"props":1778,"children":1779},{"class":433,"line":555},[1780],{"type":56,"tag":431,"props":1781,"children":1783},{"style":1782},"--shiki-light:#90A4AE;--shiki-light-font-style:italic;--shiki-default:#546E7A;--shiki-default-font-style:italic;--shiki-dark:#676E95;--shiki-dark-font-style:italic",[1784],{"type":62,"value":1785},"# Step 1: Export\n",{"type":56,"tag":431,"props":1787,"children":1788},{"class":433,"line":23},[1789,1793,1797,1801,1805,1809,1813],{"type":56,"tag":431,"props":1790,"children":1791},{"style":437},[1792],{"type":62,"value":1399},{"type":56,"tag":431,"props":1794,"children":1795},{"style":443},[1796],{"type":62,"value":1404},{"type":56,"tag":431,"props":1798,"children":1799},{"style":443},[1800],{"type":62,"value":1409},{"type":56,"tag":431,"props":1802,"children":1803},{"style":443},[1804],{"type":62,"value":1414},{"type":56,"tag":431,"props":1806,"children":1807},{"style":443},[1808],{"type":62,"value":1419},{"type":56,"tag":431,"props":1810,"children":1811},{"style":449},[1812],{"type":62,"value":1424},{"type":56,"tag":431,"props":1814,"children":1815},{"style":449},[1816],{"type":62,"value":452},{"type":56,"tag":431,"props":1818,"children":1819},{"class":433,"line":1470},[1820,1824,1829,1833,1837,1841,1846],{"type":56,"tag":431,"props":1821,"children":1822},{"style":443},[1823],{"type":62,"value":1436},{"type":56,"tag":431,"props":1825,"children":1826},{"style":443},[1827],{"type":62,"value":1828}," --gpus",{"type":56,"tag":431,"props":1830,"children":1831},{"style":492},[1832],{"type":62,"value":1453},{"type":56,"tag":431,"props":1834,"children":1835},{"style":443},[1836],{"type":62,"value":1458},{"type":56,"tag":431,"props":1838,"children":1839},{"style":492},[1840],{"type":62,"value":1463},{"type":56,"tag":431,"props":1842,"children":1843},{"style":443},[1844],{"type":62,"value":1845}," --shm-size=8GB",{"type":56,"tag":431,"props":1847,"children":1848},{"style":449},[1849],{"type":62,"value":452},{"type":56,"tag":431,"props":1851,"children":1852},{"class":433,"line":1483},[1853,1857,1861,1866,1870],{"type":56,"tag":431,"props":1854,"children":1855},{"style":443},[1856],{"type":62,"value":1489},{"type":56,"tag":431,"props":1858,"children":1859},{"style":443},[1860],{"type":62,"value":1494},{"type":56,"tag":431,"props":1862,"children":1863},{"style":443},[1864],{"type":62,"value":1865}," -e",{"type":56,"tag":431,"props":1867,"children":1868},{"style":443},[1869],{"type":62,"value":1239},{"type":56,"tag":431,"props":1871,"children":1872},{"style":449},[1873],{"type":62,"value":452},{"type":56,"tag":431,"props":1875,"children":1876},{"class":433,"line":1501},[1877,1881,1885,1889,1893,1897,1901],{"type":56,"tag":431,"props":1878,"children":1879},{"style":443},[1880],{"type":62,"value":1489},{"type":56,"tag":431,"props":1882,"children":1883},{"style":443},[1884],{"type":62,"value":1527},{"type":56,"tag":431,"props":1886,"children":1887},{"style":1374},[1888],{"type":62,"value":1532},{"type":56,"tag":431,"props":1890,"children":1891},{"style":443},[1892],{"type":62,"value":1865},{"type":56,"tag":431,"props":1894,"children":1895},{"style":443},[1896],{"type":62,"value":1549},{"type":56,"tag":431,"props":1898,"children":1899},{"style":1374},[1900],{"type":62,"value":1554},{"type":56,"tag":431,"props":1902,"children":1903},{"style":449},[1904],{"type":62,"value":452},{"type":56,"tag":431,"props":1906,"children":1907},{"class":433,"line":1517},[1908,1912,1916,1920,1924],{"type":56,"tag":431,"props":1909,"children":1910},{"style":443},[1911],{"type":62,"value":1567},{"type":56,"tag":431,"props":1913,"children":1914},{"style":443},[1915],{"type":62,"value":1572},{"type":56,"tag":431,"props":1917,"children":1918},{"style":443},[1919],{"type":62,"value":1356},{"type":56,"tag":431,"props":1921,"children":1922},{"style":443},[1923],{"type":62,"value":1589},{"type":56,"tag":431,"props":1925,"children":1926},{"style":449},[1927],{"type":62,"value":452},{"type":56,"tag":431,"props":1929,"children":1930},{"class":433,"line":1539},[1931,1935,1940,1945],{"type":56,"tag":431,"props":1932,"children":1933},{"style":443},[1934],{"type":62,"value":1602},{"type":56,"tag":431,"props":1936,"children":1937},{"style":449},[1938],{"type":62,"value":1939}," $NIM_EXPORT_PATH",{"type":56,"tag":431,"props":1941,"children":1942},{"style":443},[1943],{"type":62,"value":1944},":\u002Fopt\u002Fnim\u002Fexport",{"type":56,"tag":431,"props":1946,"children":1947},{"style":449},[1948],{"type":62,"value":452},{"type":56,"tag":431,"props":1950,"children":1951},{"class":433,"line":1561},[1952,1956,1961],{"type":56,"tag":431,"props":1953,"children":1954},{"style":443},[1955],{"type":62,"value":1489},{"type":56,"tag":431,"props":1957,"children":1958},{"style":443},[1959],{"type":62,"value":1960}," NIM_EXPORT_PATH=\u002Fopt\u002Fnim\u002Fexport",{"type":56,"tag":431,"props":1962,"children":1963},{"style":449},[1964],{"type":62,"value":452},{"type":56,"tag":431,"props":1966,"children":1967},{"class":433,"line":1579},[1968,1972,1976],{"type":56,"tag":431,"props":1969,"children":1970},{"style":443},[1971],{"type":62,"value":1625},{"type":56,"tag":431,"props":1973,"children":1974},{"style":449},[1975],{"type":62,"value":1424},{"type":56,"tag":431,"props":1977,"children":1978},{"style":443},[1979],{"type":62,"value":1634},{"type":56,"tag":431,"props":1981,"children":1982},{"class":433,"line":1596},[1983],{"type":56,"tag":431,"props":1984,"children":1985},{"emptyLinePlaceholder":1388},[1986],{"type":62,"value":1391},{"type":56,"tag":431,"props":1988,"children":1989},{"class":433,"line":1619},[1990],{"type":56,"tag":431,"props":1991,"children":1992},{"style":1782},[1993],{"type":62,"value":1994},"# Step 2: Run from export\n",{"type":56,"tag":431,"props":1996,"children":1998},{"class":433,"line":1997},16,[1999,2003,2007,2011,2015,2019,2023],{"type":56,"tag":431,"props":2000,"children":2001},{"style":437},[2002],{"type":62,"value":1399},{"type":56,"tag":431,"props":2004,"children":2005},{"style":443},[2006],{"type":62,"value":1404},{"type":56,"tag":431,"props":2008,"children":2009},{"style":443},[2010],{"type":62,"value":1409},{"type":56,"tag":431,"props":2012,"children":2013},{"style":443},[2014],{"type":62,"value":1414},{"type":56,"tag":431,"props":2016,"children":2017},{"style":443},[2018],{"type":62,"value":1419},{"type":56,"tag":431,"props":2020,"children":2021},{"style":449},[2022],{"type":62,"value":1424},{"type":56,"tag":431,"props":2024,"children":2025},{"style":449},[2026],{"type":62,"value":452},{"type":56,"tag":431,"props":2028,"children":2030},{"class":433,"line":2029},17,[2031,2035,2039,2043,2047,2051,2055],{"type":56,"tag":431,"props":2032,"children":2033},{"style":443},[2034],{"type":62,"value":1436},{"type":56,"tag":431,"props":2036,"children":2037},{"style":443},[2038],{"type":62,"value":1828},{"type":56,"tag":431,"props":2040,"children":2041},{"style":492},[2042],{"type":62,"value":1453},{"type":56,"tag":431,"props":2044,"children":2045},{"style":443},[2046],{"type":62,"value":1458},{"type":56,"tag":431,"props":2048,"children":2049},{"style":492},[2050],{"type":62,"value":1463},{"type":56,"tag":431,"props":2052,"children":2053},{"style":443},[2054],{"type":62,"value":1845},{"type":56,"tag":431,"props":2056,"children":2057},{"style":449},[2058],{"type":62,"value":452},{"type":56,"tag":431,"props":2060,"children":2062},{"class":433,"line":2061},18,[2063,2067,2071,2075,2079],{"type":56,"tag":431,"props":2064,"children":2065},{"style":443},[2066],{"type":62,"value":1489},{"type":56,"tag":431,"props":2068,"children":2069},{"style":443},[2070],{"type":62,"value":1494},{"type":56,"tag":431,"props":2072,"children":2073},{"style":443},[2074],{"type":62,"value":1865},{"type":56,"tag":431,"props":2076,"children":2077},{"style":443},[2078],{"type":62,"value":1239},{"type":56,"tag":431,"props":2080,"children":2081},{"style":449},[2082],{"type":62,"value":452},{"type":56,"tag":431,"props":2084,"children":2086},{"class":433,"line":2085},19,[2087,2091,2096,2101],{"type":56,"tag":431,"props":2088,"children":2089},{"style":443},[2090],{"type":62,"value":1489},{"type":56,"tag":431,"props":2092,"children":2093},{"style":443},[2094],{"type":62,"value":2095}," NIM_DISABLE_MODEL_DOWNLOAD=",{"type":56,"tag":431,"props":2097,"children":2098},{"style":492},[2099],{"type":62,"value":2100},"true",{"type":56,"tag":431,"props":2102,"children":2103},{"style":449},[2104],{"type":62,"value":452},{"type":56,"tag":431,"props":2106,"children":2108},{"class":433,"line":2107},20,[2109,2113,2117,2121,2125,2129,2133],{"type":56,"tag":431,"props":2110,"children":2111},{"style":443},[2112],{"type":62,"value":1489},{"type":56,"tag":431,"props":2114,"children":2115},{"style":443},[2116],{"type":62,"value":1527},{"type":56,"tag":431,"props":2118,"children":2119},{"style":1374},[2120],{"type":62,"value":1532},{"type":56,"tag":431,"props":2122,"children":2123},{"style":443},[2124],{"type":62,"value":1865},{"type":56,"tag":431,"props":2126,"children":2127},{"style":443},[2128],{"type":62,"value":1549},{"type":56,"tag":431,"props":2130,"children":2131},{"style":1374},[2132],{"type":62,"value":1554},{"type":56,"tag":431,"props":2134,"children":2135},{"style":449},[2136],{"type":62,"value":452},{"type":56,"tag":431,"props":2138,"children":2140},{"class":433,"line":2139},21,[2141,2145,2149,2153,2157],{"type":56,"tag":431,"props":2142,"children":2143},{"style":443},[2144],{"type":62,"value":1567},{"type":56,"tag":431,"props":2146,"children":2147},{"style":443},[2148],{"type":62,"value":1572},{"type":56,"tag":431,"props":2150,"children":2151},{"style":443},[2152],{"type":62,"value":1356},{"type":56,"tag":431,"props":2154,"children":2155},{"style":443},[2156],{"type":62,"value":1589},{"type":56,"tag":431,"props":2158,"children":2159},{"style":449},[2160],{"type":62,"value":452},{"type":56,"tag":431,"props":2162,"children":2164},{"class":433,"line":2163},22,[2165,2169,2173,2177],{"type":56,"tag":431,"props":2166,"children":2167},{"style":443},[2168],{"type":62,"value":1602},{"type":56,"tag":431,"props":2170,"children":2171},{"style":449},[2172],{"type":62,"value":1939},{"type":56,"tag":431,"props":2174,"children":2175},{"style":443},[2176],{"type":62,"value":1944},{"type":56,"tag":431,"props":2178,"children":2179},{"style":449},[2180],{"type":62,"value":452},{"type":56,"tag":431,"props":2182,"children":2184},{"class":433,"line":2183},23,[2185,2189,2193],{"type":56,"tag":431,"props":2186,"children":2187},{"style":443},[2188],{"type":62,"value":1489},{"type":56,"tag":431,"props":2190,"children":2191},{"style":443},[2192],{"type":62,"value":1960},{"type":56,"tag":431,"props":2194,"children":2195},{"style":449},[2196],{"type":62,"value":452},{"type":56,"tag":431,"props":2198,"children":2200},{"class":433,"line":2199},24,[2201,2205,2209],{"type":56,"tag":431,"props":2202,"children":2203},{"style":443},[2204],{"type":62,"value":1625},{"type":56,"tag":431,"props":2206,"children":2207},{"style":449},[2208],{"type":62,"value":1424},{"type":56,"tag":431,"props":2210,"children":2211},{"style":443},[2212],{"type":62,"value":1634},{"type":56,"tag":99,"props":2214,"children":2216},{"id":2215},"step-3-verify-readiness",[2217],{"type":62,"value":2218},"Step 3 — Verify Readiness",{"type":56,"tag":420,"props":2220,"children":2222},{"className":422,"code":2221,"language":424,"meta":425,"style":425},"curl -X GET http:\u002F\u002Flocalhost:9000\u002Fv1\u002Fhealth\u002Fready\n# Expected: {\"status\":\"ready\"}\n",[2223],{"type":56,"tag":152,"props":2224,"children":2225},{"__ignoreMap":425},[2226,2249],{"type":56,"tag":431,"props":2227,"children":2228},{"class":433,"line":27},[2229,2234,2239,2244],{"type":56,"tag":431,"props":2230,"children":2231},{"style":437},[2232],{"type":62,"value":2233},"curl",{"type":56,"tag":431,"props":2235,"children":2236},{"style":443},[2237],{"type":62,"value":2238}," -X",{"type":56,"tag":431,"props":2240,"children":2241},{"style":443},[2242],{"type":62,"value":2243}," GET",{"type":56,"tag":431,"props":2245,"children":2246},{"style":443},[2247],{"type":62,"value":2248}," http:\u002F\u002Flocalhost:9000\u002Fv1\u002Fhealth\u002Fready\n",{"type":56,"tag":431,"props":2250,"children":2251},{"class":433,"line":455},[2252],{"type":56,"tag":431,"props":2253,"children":2254},{"style":1782},[2255],{"type":62,"value":2256},"# Expected: {\"status\":\"ready\"}\n",{"type":56,"tag":99,"props":2258,"children":2260},{"id":2259},"step-4-run-inference",[2261],{"type":62,"value":2262},"Step 4 — Run Inference",{"type":56,"tag":662,"props":2264,"children":2266},{"id":2265},"streaming-asr-python-grpc",[2267],{"type":62,"value":2268},"Streaming ASR (Python — gRPC)",{"type":56,"tag":420,"props":2270,"children":2272},{"className":422,"code":2271,"language":424,"meta":425,"style":425},"python3 python-clients\u002Fscripts\u002Fasr\u002Ftranscribe_file.py \\\n  --server 0.0.0.0:50051 \\\n  --input-file \u002Fpath\u002Fto\u002Faudio.wav \\\n  --language-code en-US\n",[2273],{"type":56,"tag":152,"props":2274,"children":2275},{"__ignoreMap":425},[2276,2292,2309,2326],{"type":56,"tag":431,"props":2277,"children":2278},{"class":433,"line":27},[2279,2284,2288],{"type":56,"tag":431,"props":2280,"children":2281},{"style":437},[2282],{"type":62,"value":2283},"python3",{"type":56,"tag":431,"props":2285,"children":2286},{"style":443},[2287],{"type":62,"value":446},{"type":56,"tag":431,"props":2289,"children":2290},{"style":449},[2291],{"type":62,"value":452},{"type":56,"tag":431,"props":2293,"children":2294},{"class":433,"line":455},[2295,2300,2305],{"type":56,"tag":431,"props":2296,"children":2297},{"style":443},[2298],{"type":62,"value":2299},"  --server",{"type":56,"tag":431,"props":2301,"children":2302},{"style":443},[2303],{"type":62,"value":2304}," 0.0.0.0:50051",{"type":56,"tag":431,"props":2306,"children":2307},{"style":449},[2308],{"type":62,"value":452},{"type":56,"tag":431,"props":2310,"children":2311},{"class":433,"line":478},[2312,2317,2322],{"type":56,"tag":431,"props":2313,"children":2314},{"style":443},[2315],{"type":62,"value":2316},"  --input-file",{"type":56,"tag":431,"props":2318,"children":2319},{"style":443},[2320],{"type":62,"value":2321}," \u002Fpath\u002Fto\u002Faudio.wav",{"type":56,"tag":431,"props":2323,"children":2324},{"style":449},[2325],{"type":62,"value":452},{"type":56,"tag":431,"props":2327,"children":2328},{"class":433,"line":512},[2329,2334],{"type":56,"tag":431,"props":2330,"children":2331},{"style":443},[2332],{"type":62,"value":2333},"  --language-code",{"type":56,"tag":431,"props":2335,"children":2336},{"style":443},[2337],{"type":62,"value":2338}," en-US\n",{"type":56,"tag":69,"props":2340,"children":2341},{},[2342],{"type":62,"value":2343},"With diarization and word timestamps:",{"type":56,"tag":420,"props":2345,"children":2347},{"className":422,"code":2346,"language":424,"meta":425,"style":425},"python3 python-clients\u002Fscripts\u002Fasr\u002Ftranscribe_file.py \\\n  --server 0.0.0.0:50051 \\\n  --input-file \u002Fpath\u002Fto\u002Faudio.wav \\\n  --language-code en-US \\\n  --speaker-diarization \\\n  --word-time-offsets\n",[2348],{"type":56,"tag":152,"props":2349,"children":2350},{"__ignoreMap":425},[2351,2366,2381,2396,2412,2424],{"type":56,"tag":431,"props":2352,"children":2353},{"class":433,"line":27},[2354,2358,2362],{"type":56,"tag":431,"props":2355,"children":2356},{"style":437},[2357],{"type":62,"value":2283},{"type":56,"tag":431,"props":2359,"children":2360},{"style":443},[2361],{"type":62,"value":446},{"type":56,"tag":431,"props":2363,"children":2364},{"style":449},[2365],{"type":62,"value":452},{"type":56,"tag":431,"props":2367,"children":2368},{"class":433,"line":455},[2369,2373,2377],{"type":56,"tag":431,"props":2370,"children":2371},{"style":443},[2372],{"type":62,"value":2299},{"type":56,"tag":431,"props":2374,"children":2375},{"style":443},[2376],{"type":62,"value":2304},{"type":56,"tag":431,"props":2378,"children":2379},{"style":449},[2380],{"type":62,"value":452},{"type":56,"tag":431,"props":2382,"children":2383},{"class":433,"line":478},[2384,2388,2392],{"type":56,"tag":431,"props":2385,"children":2386},{"style":443},[2387],{"type":62,"value":2316},{"type":56,"tag":431,"props":2389,"children":2390},{"style":443},[2391],{"type":62,"value":2321},{"type":56,"tag":431,"props":2393,"children":2394},{"style":449},[2395],{"type":62,"value":452},{"type":56,"tag":431,"props":2397,"children":2398},{"class":433,"line":512},[2399,2403,2408],{"type":56,"tag":431,"props":2400,"children":2401},{"style":443},[2402],{"type":62,"value":2333},{"type":56,"tag":431,"props":2404,"children":2405},{"style":443},[2406],{"type":62,"value":2407}," en-US",{"type":56,"tag":431,"props":2409,"children":2410},{"style":449},[2411],{"type":62,"value":452},{"type":56,"tag":431,"props":2413,"children":2414},{"class":433,"line":555},[2415,2420],{"type":56,"tag":431,"props":2416,"children":2417},{"style":443},[2418],{"type":62,"value":2419},"  --speaker-diarization",{"type":56,"tag":431,"props":2421,"children":2422},{"style":449},[2423],{"type":62,"value":452},{"type":56,"tag":431,"props":2425,"children":2426},{"class":433,"line":23},[2427],{"type":56,"tag":431,"props":2428,"children":2429},{"style":443},[2430],{"type":62,"value":2431},"  --word-time-offsets\n",{"type":56,"tag":69,"props":2433,"children":2434},{},[2435],{"type":62,"value":2436},"For real-time microphone streaming:",{"type":56,"tag":420,"props":2438,"children":2440},{"className":422,"code":2439,"language":424,"meta":425,"style":425},"python3 python-clients\u002Fscripts\u002Fasr\u002Ftranscribe_mic.py \\\n  --server 0.0.0.0:50051\n",[2441],{"type":56,"tag":152,"props":2442,"children":2443},{"__ignoreMap":425},[2444,2460],{"type":56,"tag":431,"props":2445,"children":2446},{"class":433,"line":27},[2447,2451,2456],{"type":56,"tag":431,"props":2448,"children":2449},{"style":437},[2450],{"type":62,"value":2283},{"type":56,"tag":431,"props":2452,"children":2453},{"style":443},[2454],{"type":62,"value":2455}," python-clients\u002Fscripts\u002Fasr\u002Ftranscribe_mic.py",{"type":56,"tag":431,"props":2457,"children":2458},{"style":449},[2459],{"type":62,"value":452},{"type":56,"tag":431,"props":2461,"children":2462},{"class":433,"line":455},[2463,2467],{"type":56,"tag":431,"props":2464,"children":2465},{"style":443},[2466],{"type":62,"value":2299},{"type":56,"tag":431,"props":2468,"children":2469},{"style":443},[2470],{"type":62,"value":2471}," 0.0.0.0:50051\n",{"type":56,"tag":662,"props":2473,"children":2475},{"id":2474},"offline-transcription-python-grpc",[2476],{"type":62,"value":2477},"Offline Transcription (Python — gRPC)",{"type":56,"tag":420,"props":2479,"children":2481},{"className":422,"code":2480,"language":424,"meta":425,"style":425},"python3 python-clients\u002Fscripts\u002Fasr\u002Ftranscribe_file_offline.py \\\n  --server 0.0.0.0:50051 \\\n  --input-file \u002Fpath\u002Fto\u002Faudio.wav \\\n  --language-code en-US\n",[2482],{"type":56,"tag":152,"props":2483,"children":2484},{"__ignoreMap":425},[2485,2501,2516,2531],{"type":56,"tag":431,"props":2486,"children":2487},{"class":433,"line":27},[2488,2492,2497],{"type":56,"tag":431,"props":2489,"children":2490},{"style":437},[2491],{"type":62,"value":2283},{"type":56,"tag":431,"props":2493,"children":2494},{"style":443},[2495],{"type":62,"value":2496}," python-clients\u002Fscripts\u002Fasr\u002Ftranscribe_file_offline.py",{"type":56,"tag":431,"props":2498,"children":2499},{"style":449},[2500],{"type":62,"value":452},{"type":56,"tag":431,"props":2502,"children":2503},{"class":433,"line":455},[2504,2508,2512],{"type":56,"tag":431,"props":2505,"children":2506},{"style":443},[2507],{"type":62,"value":2299},{"type":56,"tag":431,"props":2509,"children":2510},{"style":443},[2511],{"type":62,"value":2304},{"type":56,"tag":431,"props":2513,"children":2514},{"style":449},[2515],{"type":62,"value":452},{"type":56,"tag":431,"props":2517,"children":2518},{"class":433,"line":478},[2519,2523,2527],{"type":56,"tag":431,"props":2520,"children":2521},{"style":443},[2522],{"type":62,"value":2316},{"type":56,"tag":431,"props":2524,"children":2525},{"style":443},[2526],{"type":62,"value":2321},{"type":56,"tag":431,"props":2528,"children":2529},{"style":449},[2530],{"type":62,"value":452},{"type":56,"tag":431,"props":2532,"children":2533},{"class":433,"line":512},[2534,2538],{"type":56,"tag":431,"props":2535,"children":2536},{"style":443},[2537],{"type":62,"value":2333},{"type":56,"tag":431,"props":2539,"children":2540},{"style":443},[2541],{"type":62,"value":2338},{"type":56,"tag":662,"props":2543,"children":2545},{"id":2544},"http-api",[2546],{"type":62,"value":2547},"HTTP API",{"type":56,"tag":420,"props":2549,"children":2551},{"className":422,"code":2550,"language":424,"meta":425,"style":425},"curl -X POST http:\u002F\u002Flocalhost:9000\u002Fv1\u002Faudio\u002Ftranscriptions \\\n  -H \"Content-Type: multipart\u002Fform-data\" \\\n  -F \"file=@\u002Fpath\u002Fto\u002Faudio.wav\" \\\n  -F \"model=asr\"\n",[2552],{"type":56,"tag":152,"props":2553,"children":2554},{"__ignoreMap":425},[2555,2580,2605,2630],{"type":56,"tag":431,"props":2556,"children":2557},{"class":433,"line":27},[2558,2562,2566,2571,2576],{"type":56,"tag":431,"props":2559,"children":2560},{"style":437},[2561],{"type":62,"value":2233},{"type":56,"tag":431,"props":2563,"children":2564},{"style":443},[2565],{"type":62,"value":2238},{"type":56,"tag":431,"props":2567,"children":2568},{"style":443},[2569],{"type":62,"value":2570}," POST",{"type":56,"tag":431,"props":2572,"children":2573},{"style":443},[2574],{"type":62,"value":2575}," http:\u002F\u002Flocalhost:9000\u002Fv1\u002Faudio\u002Ftranscriptions",{"type":56,"tag":431,"props":2577,"children":2578},{"style":449},[2579],{"type":62,"value":452},{"type":56,"tag":431,"props":2581,"children":2582},{"class":433,"line":455},[2583,2588,2592,2597,2601],{"type":56,"tag":431,"props":2584,"children":2585},{"style":443},[2586],{"type":62,"value":2587},"  -H",{"type":56,"tag":431,"props":2589,"children":2590},{"style":492},[2591],{"type":62,"value":495},{"type":56,"tag":431,"props":2593,"children":2594},{"style":443},[2595],{"type":62,"value":2596},"Content-Type: multipart\u002Fform-data",{"type":56,"tag":431,"props":2598,"children":2599},{"style":492},[2600],{"type":62,"value":505},{"type":56,"tag":431,"props":2602,"children":2603},{"style":449},[2604],{"type":62,"value":452},{"type":56,"tag":431,"props":2606,"children":2607},{"class":433,"line":478},[2608,2613,2617,2622,2626],{"type":56,"tag":431,"props":2609,"children":2610},{"style":443},[2611],{"type":62,"value":2612},"  -F",{"type":56,"tag":431,"props":2614,"children":2615},{"style":492},[2616],{"type":62,"value":495},{"type":56,"tag":431,"props":2618,"children":2619},{"style":443},[2620],{"type":62,"value":2621},"file=@\u002Fpath\u002Fto\u002Faudio.wav",{"type":56,"tag":431,"props":2623,"children":2624},{"style":492},[2625],{"type":62,"value":505},{"type":56,"tag":431,"props":2627,"children":2628},{"style":449},[2629],{"type":62,"value":452},{"type":56,"tag":431,"props":2631,"children":2632},{"class":433,"line":512},[2633,2637,2641,2646],{"type":56,"tag":431,"props":2634,"children":2635},{"style":443},[2636],{"type":62,"value":2612},{"type":56,"tag":431,"props":2638,"children":2639},{"style":492},[2640],{"type":62,"value":495},{"type":56,"tag":431,"props":2642,"children":2643},{"style":443},[2644],{"type":62,"value":2645},"model=asr",{"type":56,"tag":431,"props":2647,"children":2648},{"style":492},[2649],{"type":62,"value":1258},{"type":56,"tag":662,"props":2651,"children":2653},{"id":2652},"c-client",[2654],{"type":62,"value":2655},"C++ Client",{"type":56,"tag":420,"props":2657,"children":2659},{"className":422,"code":2658,"language":424,"meta":425,"style":425},"# Build (requires Bazel in cpp-clients\u002F)\ncd cpp-clients\nbazel build \u002F\u002Friva\u002Fclients\u002Fasr:riva_asr_client\n\n# Run\n.\u002Fbazel-bin\u002Friva\u002Fclients\u002Fasr\u002Friva_asr_client \\\n  --server=0.0.0.0:50051 \\\n  --audio-file=\u002Fpath\u002Fto\u002Faudio.wav\n",[2660],{"type":56,"tag":152,"props":2661,"children":2662},{"__ignoreMap":425},[2663,2671,2684,2702,2709,2717,2729,2741],{"type":56,"tag":431,"props":2664,"children":2665},{"class":433,"line":27},[2666],{"type":56,"tag":431,"props":2667,"children":2668},{"style":1782},[2669],{"type":62,"value":2670},"# Build (requires Bazel in cpp-clients\u002F)\n",{"type":56,"tag":431,"props":2672,"children":2673},{"class":433,"line":455},[2674,2679],{"type":56,"tag":431,"props":2675,"children":2677},{"style":2676},"--shiki-light:#6182B8;--shiki-default:#82AAFF;--shiki-dark:#82AAFF",[2678],{"type":62,"value":354},{"type":56,"tag":431,"props":2680,"children":2681},{"style":443},[2682],{"type":62,"value":2683}," cpp-clients\n",{"type":56,"tag":431,"props":2685,"children":2686},{"class":433,"line":478},[2687,2692,2697],{"type":56,"tag":431,"props":2688,"children":2689},{"style":437},[2690],{"type":62,"value":2691},"bazel",{"type":56,"tag":431,"props":2693,"children":2694},{"style":443},[2695],{"type":62,"value":2696}," build",{"type":56,"tag":431,"props":2698,"children":2699},{"style":443},[2700],{"type":62,"value":2701}," \u002F\u002Friva\u002Fclients\u002Fasr:riva_asr_client\n",{"type":56,"tag":431,"props":2703,"children":2704},{"class":433,"line":512},[2705],{"type":56,"tag":431,"props":2706,"children":2707},{"emptyLinePlaceholder":1388},[2708],{"type":62,"value":1391},{"type":56,"tag":431,"props":2710,"children":2711},{"class":433,"line":555},[2712],{"type":56,"tag":431,"props":2713,"children":2714},{"style":1782},[2715],{"type":62,"value":2716},"# Run\n",{"type":56,"tag":431,"props":2718,"children":2719},{"class":433,"line":23},[2720,2725],{"type":56,"tag":431,"props":2721,"children":2722},{"style":437},[2723],{"type":62,"value":2724},".\u002Fbazel-bin\u002Friva\u002Fclients\u002Fasr\u002Friva_asr_client",{"type":56,"tag":431,"props":2726,"children":2727},{"style":449},[2728],{"type":62,"value":452},{"type":56,"tag":431,"props":2730,"children":2731},{"class":433,"line":1470},[2732,2737],{"type":56,"tag":431,"props":2733,"children":2734},{"style":443},[2735],{"type":62,"value":2736},"  --server=0.0.0.0:50051",{"type":56,"tag":431,"props":2738,"children":2739},{"style":449},[2740],{"type":62,"value":452},{"type":56,"tag":431,"props":2742,"children":2743},{"class":433,"line":1483},[2744],{"type":56,"tag":431,"props":2745,"children":2746},{"style":443},[2747],{"type":62,"value":2748},"  --audio-file=\u002Fpath\u002Fto\u002Faudio.wav\n",{"type":56,"tag":662,"props":2750,"children":2752},{"id":2751},"websocket-audiocodes-telephony",[2753],{"type":62,"value":2754},"WebSocket (AudioCodes \u002F Telephony)",{"type":56,"tag":69,"props":2756,"children":2757},{},[2758,2760,2766],{"type":62,"value":2759},"The ",{"type":56,"tag":152,"props":2761,"children":2763},{"className":2762},[],[2764],{"type":62,"value":2765},"websocket-bridge",{"type":62,"value":2767}," repo provides a Node.js bridge for AudioCodes-compatible telephony integrations:",{"type":56,"tag":420,"props":2769,"children":2771},{"className":422,"code":2770,"language":424,"meta":425,"style":425},"cd websocket-bridge\nnpm install\nnode src\u002Findex.js --riva-server=localhost:50051\n",[2772],{"type":56,"tag":152,"props":2773,"children":2774},{"__ignoreMap":425},[2775,2787,2800],{"type":56,"tag":431,"props":2776,"children":2777},{"class":433,"line":27},[2778,2782],{"type":56,"tag":431,"props":2779,"children":2780},{"style":2676},[2781],{"type":62,"value":354},{"type":56,"tag":431,"props":2783,"children":2784},{"style":443},[2785],{"type":62,"value":2786}," websocket-bridge\n",{"type":56,"tag":431,"props":2788,"children":2789},{"class":433,"line":455},[2790,2795],{"type":56,"tag":431,"props":2791,"children":2792},{"style":437},[2793],{"type":62,"value":2794},"npm",{"type":56,"tag":431,"props":2796,"children":2797},{"style":443},[2798],{"type":62,"value":2799}," install\n",{"type":56,"tag":431,"props":2801,"children":2802},{"class":433,"line":478},[2803,2808,2813],{"type":56,"tag":431,"props":2804,"children":2805},{"style":437},[2806],{"type":62,"value":2807},"node",{"type":56,"tag":431,"props":2809,"children":2810},{"style":443},[2811],{"type":62,"value":2812}," src\u002Findex.js",{"type":56,"tag":431,"props":2814,"children":2815},{"style":443},[2816],{"type":62,"value":2817}," --riva-server=localhost:50051\n",{"type":56,"tag":99,"props":2819,"children":2821},{"id":2820},"port-reference",[2822],{"type":62,"value":2823},"Port Reference",{"type":56,"tag":681,"props":2825,"children":2826},{},[2827,2848],{"type":56,"tag":685,"props":2828,"children":2829},{},[2830],{"type":56,"tag":689,"props":2831,"children":2832},{},[2833,2838,2843],{"type":56,"tag":693,"props":2834,"children":2835},{},[2836],{"type":62,"value":2837},"Port",{"type":56,"tag":693,"props":2839,"children":2840},{},[2841],{"type":62,"value":2842},"Protocol",{"type":56,"tag":693,"props":2844,"children":2845},{},[2846],{"type":62,"value":2847},"Use",{"type":56,"tag":719,"props":2849,"children":2850},{},[2851,2868],{"type":56,"tag":689,"props":2852,"children":2853},{},[2854,2858,2863],{"type":56,"tag":726,"props":2855,"children":2856},{},[2857],{"type":62,"value":1532},{"type":56,"tag":726,"props":2859,"children":2860},{},[2861],{"type":62,"value":2862},"HTTP",{"type":56,"tag":726,"props":2864,"children":2865},{},[2866],{"type":62,"value":2867},"REST API, health check",{"type":56,"tag":689,"props":2869,"children":2870},{},[2871,2875,2880],{"type":56,"tag":726,"props":2872,"children":2873},{},[2874],{"type":62,"value":1554},{"type":56,"tag":726,"props":2876,"children":2877},{},[2878],{"type":62,"value":2879},"gRPC",{"type":56,"tag":726,"props":2881,"children":2882},{},[2883],{"type":62,"value":2884},"Python\u002FC++ client inference",{"type":56,"tag":99,"props":2886,"children":2888},{"id":2887},"examples",[2889],{"type":62,"value":2890},"Examples",{"type":56,"tag":69,"props":2892,"children":2893},{},[2894],{"type":62,"value":2895},"These are end-to-end use cases that combine features covered above; they are not repeats of the canonical commands in Option A or Step 4.",{"type":56,"tag":69,"props":2897,"children":2898},{},[2899],{"type":56,"tag":73,"props":2900,"children":2901},{},[2902],{"type":62,"value":2903},"Whisper translation — non-English audio → English transcript (cloud):",{"type":56,"tag":420,"props":2905,"children":2907},{"className":422,"code":2906,"language":424,"meta":425,"style":425},"python python-clients\u002Fscripts\u002Fasr\u002Ftranscribe_file_offline.py \\\n    --server grpc.nvcf.nvidia.com:443 --use-ssl \\\n    --metadata function-id \"b702f636-f60c-4a3d-a6f4-f3568c13bd7d\" \\\n    --metadata authorization \"Bearer $NVIDIA_API_KEY\" \\\n    --custom-configuration \"task:translate\" \\\n    --input-file spanish_audio.wav\n",[2908],{"type":56,"tag":152,"props":2909,"children":2910},{"__ignoreMap":425},[2911,2926,2945,2972,3004,3029],{"type":56,"tag":431,"props":2912,"children":2913},{"class":433,"line":27},[2914,2918,2922],{"type":56,"tag":431,"props":2915,"children":2916},{"style":437},[2917],{"type":62,"value":440},{"type":56,"tag":431,"props":2919,"children":2920},{"style":443},[2921],{"type":62,"value":2496},{"type":56,"tag":431,"props":2923,"children":2924},{"style":449},[2925],{"type":62,"value":452},{"type":56,"tag":431,"props":2927,"children":2928},{"class":433,"line":455},[2929,2933,2937,2941],{"type":56,"tag":431,"props":2930,"children":2931},{"style":443},[2932],{"type":62,"value":461},{"type":56,"tag":431,"props":2934,"children":2935},{"style":443},[2936],{"type":62,"value":466},{"type":56,"tag":431,"props":2938,"children":2939},{"style":443},[2940],{"type":62,"value":471},{"type":56,"tag":431,"props":2942,"children":2943},{"style":449},[2944],{"type":62,"value":452},{"type":56,"tag":431,"props":2946,"children":2947},{"class":433,"line":478},[2948,2952,2956,2960,2964,2968],{"type":56,"tag":431,"props":2949,"children":2950},{"style":443},[2951],{"type":62,"value":484},{"type":56,"tag":431,"props":2953,"children":2954},{"style":443},[2955],{"type":62,"value":489},{"type":56,"tag":431,"props":2957,"children":2958},{"style":492},[2959],{"type":62,"value":495},{"type":56,"tag":431,"props":2961,"children":2962},{"style":443},[2963],{"type":62,"value":877},{"type":56,"tag":431,"props":2965,"children":2966},{"style":492},[2967],{"type":62,"value":505},{"type":56,"tag":431,"props":2969,"children":2970},{"style":449},[2971],{"type":62,"value":452},{"type":56,"tag":431,"props":2973,"children":2974},{"class":433,"line":512},[2975,2979,2984,2988,2992,2996,3000],{"type":56,"tag":431,"props":2976,"children":2977},{"style":443},[2978],{"type":62,"value":484},{"type":56,"tag":431,"props":2980,"children":2981},{"style":443},[2982],{"type":62,"value":2983}," authorization",{"type":56,"tag":431,"props":2985,"children":2986},{"style":492},[2987],{"type":62,"value":495},{"type":56,"tag":431,"props":2989,"children":2990},{"style":443},[2991],{"type":62,"value":539},{"type":56,"tag":431,"props":2993,"children":2994},{"style":449},[2995],{"type":62,"value":544},{"type":56,"tag":431,"props":2997,"children":2998},{"style":492},[2999],{"type":62,"value":505},{"type":56,"tag":431,"props":3001,"children":3002},{"style":449},[3003],{"type":62,"value":452},{"type":56,"tag":431,"props":3005,"children":3006},{"class":433,"line":555},[3007,3012,3016,3021,3025],{"type":56,"tag":431,"props":3008,"children":3009},{"style":443},[3010],{"type":62,"value":3011},"    --custom-configuration",{"type":56,"tag":431,"props":3013,"children":3014},{"style":492},[3015],{"type":62,"value":495},{"type":56,"tag":431,"props":3017,"children":3018},{"style":443},[3019],{"type":62,"value":3020},"task:translate",{"type":56,"tag":431,"props":3022,"children":3023},{"style":492},[3024],{"type":62,"value":505},{"type":56,"tag":431,"props":3026,"children":3027},{"style":449},[3028],{"type":62,"value":452},{"type":56,"tag":431,"props":3030,"children":3031},{"class":433,"line":23},[3032,3036],{"type":56,"tag":431,"props":3033,"children":3034},{"style":443},[3035],{"type":62,"value":593},{"type":56,"tag":431,"props":3037,"children":3038},{"style":443},[3039],{"type":62,"value":3040}," spanish_audio.wav\n",{"type":56,"tag":69,"props":3042,"children":3043},{},[3044],{"type":56,"tag":73,"props":3045,"children":3046},{},[3047],{"type":62,"value":3048},"Streaming with diarization and word timestamps (self-hosted):",{"type":56,"tag":420,"props":3050,"children":3052},{"className":422,"code":3051,"language":424,"meta":425,"style":425},"python3 python-clients\u002Fscripts\u002Fasr\u002Ftranscribe_file.py \\\n  --server 0.0.0.0:50051 --input-file meeting.wav --language-code en-US \\\n  --speaker-diarization --word-time-offsets\n",[3053],{"type":56,"tag":152,"props":3054,"children":3055},{"__ignoreMap":425},[3056,3071,3105],{"type":56,"tag":431,"props":3057,"children":3058},{"class":433,"line":27},[3059,3063,3067],{"type":56,"tag":431,"props":3060,"children":3061},{"style":437},[3062],{"type":62,"value":2283},{"type":56,"tag":431,"props":3064,"children":3065},{"style":443},[3066],{"type":62,"value":446},{"type":56,"tag":431,"props":3068,"children":3069},{"style":449},[3070],{"type":62,"value":452},{"type":56,"tag":431,"props":3072,"children":3073},{"class":433,"line":455},[3074,3078,3082,3087,3092,3097,3101],{"type":56,"tag":431,"props":3075,"children":3076},{"style":443},[3077],{"type":62,"value":2299},{"type":56,"tag":431,"props":3079,"children":3080},{"style":443},[3081],{"type":62,"value":2304},{"type":56,"tag":431,"props":3083,"children":3084},{"style":443},[3085],{"type":62,"value":3086}," --input-file",{"type":56,"tag":431,"props":3088,"children":3089},{"style":443},[3090],{"type":62,"value":3091}," meeting.wav",{"type":56,"tag":431,"props":3093,"children":3094},{"style":443},[3095],{"type":62,"value":3096}," --language-code",{"type":56,"tag":431,"props":3098,"children":3099},{"style":443},[3100],{"type":62,"value":2407},{"type":56,"tag":431,"props":3102,"children":3103},{"style":449},[3104],{"type":62,"value":452},{"type":56,"tag":431,"props":3106,"children":3107},{"class":433,"line":478},[3108,3112],{"type":56,"tag":431,"props":3109,"children":3110},{"style":443},[3111],{"type":62,"value":2419},{"type":56,"tag":431,"props":3113,"children":3114},{"style":443},[3115],{"type":62,"value":3116}," --word-time-offsets\n",{"type":56,"tag":69,"props":3118,"children":3119},{},[3120],{"type":56,"tag":73,"props":3121,"children":3122},{},[3123],{"type":62,"value":3124},"Domain-specific vocabulary via token boosting (CTC, self-hosted):",{"type":56,"tag":420,"props":3126,"children":3128},{"className":422,"code":3127,"language":424,"meta":425,"style":425},"python3 python-clients\u002Fscripts\u002Fasr\u002Ftranscribe_file.py \\\n  --server 0.0.0.0:50051 --input-file clinical_dictation.wav \\\n  --language-code en-US \\\n  --boosted-lm-words 'Abloopar:▁a\u002Fb\u002Flo\u002Fop\u002Fer' \\\n  --boosted-lm-score 80\n",[3129],{"type":56,"tag":152,"props":3130,"children":3131},{"__ignoreMap":425},[3132,3147,3171,3186,3211],{"type":56,"tag":431,"props":3133,"children":3134},{"class":433,"line":27},[3135,3139,3143],{"type":56,"tag":431,"props":3136,"children":3137},{"style":437},[3138],{"type":62,"value":2283},{"type":56,"tag":431,"props":3140,"children":3141},{"style":443},[3142],{"type":62,"value":446},{"type":56,"tag":431,"props":3144,"children":3145},{"style":449},[3146],{"type":62,"value":452},{"type":56,"tag":431,"props":3148,"children":3149},{"class":433,"line":455},[3150,3154,3158,3162,3167],{"type":56,"tag":431,"props":3151,"children":3152},{"style":443},[3153],{"type":62,"value":2299},{"type":56,"tag":431,"props":3155,"children":3156},{"style":443},[3157],{"type":62,"value":2304},{"type":56,"tag":431,"props":3159,"children":3160},{"style":443},[3161],{"type":62,"value":3086},{"type":56,"tag":431,"props":3163,"children":3164},{"style":443},[3165],{"type":62,"value":3166}," clinical_dictation.wav",{"type":56,"tag":431,"props":3168,"children":3169},{"style":449},[3170],{"type":62,"value":452},{"type":56,"tag":431,"props":3172,"children":3173},{"class":433,"line":478},[3174,3178,3182],{"type":56,"tag":431,"props":3175,"children":3176},{"style":443},[3177],{"type":62,"value":2333},{"type":56,"tag":431,"props":3179,"children":3180},{"style":443},[3181],{"type":62,"value":2407},{"type":56,"tag":431,"props":3183,"children":3184},{"style":449},[3185],{"type":62,"value":452},{"type":56,"tag":431,"props":3187,"children":3188},{"class":433,"line":512},[3189,3194,3198,3203,3207],{"type":56,"tag":431,"props":3190,"children":3191},{"style":443},[3192],{"type":62,"value":3193},"  --boosted-lm-words",{"type":56,"tag":431,"props":3195,"children":3196},{"style":492},[3197],{"type":62,"value":1453},{"type":56,"tag":431,"props":3199,"children":3200},{"style":443},[3201],{"type":62,"value":3202},"Abloopar:▁a\u002Fb\u002Flo\u002Fop\u002Fer",{"type":56,"tag":431,"props":3204,"children":3205},{"style":492},[3206],{"type":62,"value":1463},{"type":56,"tag":431,"props":3208,"children":3209},{"style":449},[3210],{"type":62,"value":452},{"type":56,"tag":431,"props":3212,"children":3213},{"class":433,"line":555},[3214,3219],{"type":56,"tag":431,"props":3215,"children":3216},{"style":443},[3217],{"type":62,"value":3218},"  --boosted-lm-score",{"type":56,"tag":431,"props":3220,"children":3221},{"style":1374},[3222],{"type":62,"value":3223}," 80\n",{"type":56,"tag":99,"props":3225,"children":3227},{"id":3226},"troubleshooting",[3228],{"type":62,"value":3229},"Troubleshooting",{"type":56,"tag":142,"props":3231,"children":3232},{},[3233,3248,3266,3284],{"type":56,"tag":146,"props":3234,"children":3235},{},[3236,3246],{"type":56,"tag":73,"props":3237,"children":3238},{},[3239,3241],{"type":62,"value":3240},"Wrong ",{"type":56,"tag":152,"props":3242,"children":3244},{"className":3243},[],[3245],{"type":62,"value":306},{"type":62,"value":3247}," — if the selector doesn't match any available profile, the container exits. Check the support matrix for exact tag values.",{"type":56,"tag":146,"props":3249,"children":3250},{},[3251,3256,3258,3264],{"type":56,"tag":73,"props":3252,"children":3253},{},[3254],{"type":62,"value":3255},"GPU device index",{"type":62,"value":3257}," — ",{"type":56,"tag":152,"props":3259,"children":3261},{"className":3260},[],[3262],{"type":62,"value":3263},"--gpus '\"device=0\"'",{"type":62,"value":3265}," targets GPU 0. Adjust for multi-GPU hosts.",{"type":56,"tag":146,"props":3267,"children":3268},{},[3269,3274,3276,3282],{"type":56,"tag":73,"props":3270,"children":3271},{},[3272],{"type":62,"value":3273},"Port 8000 conflict",{"type":62,"value":3275}," — avoid ",{"type":56,"tag":152,"props":3277,"children":3279},{"className":3278},[],[3280],{"type":62,"value":3281},"NIM_HTTP_API_PORT=8000",{"type":62,"value":3283},"; use 9000 (default).",{"type":56,"tag":146,"props":3285,"children":3286},{},[3287,3292,3294,3300],{"type":56,"tag":73,"props":3288,"children":3289},{},[3290],{"type":62,"value":3291},"Parakeet TDT and word timestamps",{"type":62,"value":3293}," — only available via gRPC with ",{"type":56,"tag":152,"props":3295,"children":3297},{"className":3296},[],[3298],{"type":62,"value":3299},"enable_word_time_offsets=True",{"type":62,"value":3301}," in the recognition config.",{"type":56,"tag":99,"props":3303,"children":3305},{"id":3304},"customization",[3306],{"type":62,"value":3307},"Customization",{"type":56,"tag":662,"props":3309,"children":3311},{"id":3310},"word-boosting",[3312],{"type":62,"value":3313},"Word Boosting",{"type":56,"tag":69,"props":3315,"children":3316},{},[3317],{"type":62,"value":3318},"Bias the model toward domain-specific words at request time. Pass each word as a separate flag:",{"type":56,"tag":420,"props":3320,"children":3322},{"className":422,"code":3321,"language":424,"meta":425,"style":425},"python scripts\u002Fasr\u002Ftranscribe_file.py \\\n  --server 0.0.0.0:50051 \\\n  --input-file audio.wav \\\n  --boosted-lm-words 'AntiBERTa' \\\n  --boosted-lm-words 'Abloopar' \\\n  --boosted-lm-score 20\n",[3323],{"type":56,"tag":152,"props":3324,"children":3325},{"__ignoreMap":425},[3326,3342,3357,3373,3397,3421],{"type":56,"tag":431,"props":3327,"children":3328},{"class":433,"line":27},[3329,3333,3338],{"type":56,"tag":431,"props":3330,"children":3331},{"style":437},[3332],{"type":62,"value":440},{"type":56,"tag":431,"props":3334,"children":3335},{"style":443},[3336],{"type":62,"value":3337}," scripts\u002Fasr\u002Ftranscribe_file.py",{"type":56,"tag":431,"props":3339,"children":3340},{"style":449},[3341],{"type":62,"value":452},{"type":56,"tag":431,"props":3343,"children":3344},{"class":433,"line":455},[3345,3349,3353],{"type":56,"tag":431,"props":3346,"children":3347},{"style":443},[3348],{"type":62,"value":2299},{"type":56,"tag":431,"props":3350,"children":3351},{"style":443},[3352],{"type":62,"value":2304},{"type":56,"tag":431,"props":3354,"children":3355},{"style":449},[3356],{"type":62,"value":452},{"type":56,"tag":431,"props":3358,"children":3359},{"class":433,"line":478},[3360,3364,3369],{"type":56,"tag":431,"props":3361,"children":3362},{"style":443},[3363],{"type":62,"value":2316},{"type":56,"tag":431,"props":3365,"children":3366},{"style":443},[3367],{"type":62,"value":3368}," audio.wav",{"type":56,"tag":431,"props":3370,"children":3371},{"style":449},[3372],{"type":62,"value":452},{"type":56,"tag":431,"props":3374,"children":3375},{"class":433,"line":512},[3376,3380,3384,3389,3393],{"type":56,"tag":431,"props":3377,"children":3378},{"style":443},[3379],{"type":62,"value":3193},{"type":56,"tag":431,"props":3381,"children":3382},{"style":492},[3383],{"type":62,"value":1453},{"type":56,"tag":431,"props":3385,"children":3386},{"style":443},[3387],{"type":62,"value":3388},"AntiBERTa",{"type":56,"tag":431,"props":3390,"children":3391},{"style":492},[3392],{"type":62,"value":1463},{"type":56,"tag":431,"props":3394,"children":3395},{"style":449},[3396],{"type":62,"value":452},{"type":56,"tag":431,"props":3398,"children":3399},{"class":433,"line":555},[3400,3404,3408,3413,3417],{"type":56,"tag":431,"props":3401,"children":3402},{"style":443},[3403],{"type":62,"value":3193},{"type":56,"tag":431,"props":3405,"children":3406},{"style":492},[3407],{"type":62,"value":1453},{"type":56,"tag":431,"props":3409,"children":3410},{"style":443},[3411],{"type":62,"value":3412},"Abloopar",{"type":56,"tag":431,"props":3414,"children":3415},{"style":492},[3416],{"type":62,"value":1463},{"type":56,"tag":431,"props":3418,"children":3419},{"style":449},[3420],{"type":62,"value":452},{"type":56,"tag":431,"props":3422,"children":3423},{"class":433,"line":23},[3424,3428],{"type":56,"tag":431,"props":3425,"children":3426},{"style":443},[3427],{"type":62,"value":3218},{"type":56,"tag":431,"props":3429,"children":3430},{"style":1374},[3431],{"type":62,"value":3432}," 20\n",{"type":56,"tag":69,"props":3434,"children":3435},{},[3436,3438,3443,3445,3450],{"type":62,"value":3437},"Boost score ranges: ",{"type":56,"tag":73,"props":3439,"children":3440},{},[3441],{"type":62,"value":3442},"CTC",{"type":62,"value":3444},": 20–100 (negative scores discourage words) | ",{"type":56,"tag":73,"props":3446,"children":3447},{},[3448],{"type":62,"value":3449},"RNNT\u002FTDT",{"type":62,"value":3451},": 0.5–2.0 (single score for all words, no negatives, no OOV).",{"type":56,"tag":662,"props":3453,"children":3455},{"id":3454},"token-boosting-ctc-only",[3456],{"type":62,"value":3457},"Token Boosting (CTC only)",{"type":56,"tag":69,"props":3459,"children":3460},{},[3461,3463,3468,3470,3475],{"type":62,"value":3462},"Maps the tokens the model ",{"type":56,"tag":761,"props":3464,"children":3465},{},[3466],{"type":62,"value":3467},"predicts",{"type":62,"value":3469}," to the word you ",{"type":56,"tag":761,"props":3471,"children":3472},{},[3473],{"type":62,"value":3474},"want",{"type":62,"value":3476},". Use when word boosting alone can't overcome the acoustic model (e.g. multi-syllable OOV terms).",{"type":56,"tag":69,"props":3478,"children":3479},{},[3480],{"type":56,"tag":73,"props":3481,"children":3482},{},[3483],{"type":62,"value":3484},"Step 1 — Get the model tokenizer from the running container:",{"type":56,"tag":420,"props":3486,"children":3488},{"className":422,"code":3487,"language":424,"meta":425,"style":425},"docker exec \u003Ccontainer> find \u002Fdata\u002Fmodels -name \"*tokenizer.model\" | head -5\ndocker cp \u003Ccontainer>:\u002Fdata\u002Fmodels\u002F\u003Cmodel-dir>\u002F1\u002F\u003Chash>_tokenizer.model \u002Ftmp\u002Ftokenizer.model\n",[3489],{"type":56,"tag":152,"props":3490,"children":3491},{"__ignoreMap":425},[3492,3565],{"type":56,"tag":431,"props":3493,"children":3494},{"class":433,"line":27},[3495,3499,3504,3508,3513,3518,3522,3527,3532,3537,3541,3546,3550,3555,3560],{"type":56,"tag":431,"props":3496,"children":3497},{"style":437},[3498],{"type":62,"value":1399},{"type":56,"tag":431,"props":3500,"children":3501},{"style":443},[3502],{"type":62,"value":3503}," exec",{"type":56,"tag":431,"props":3505,"children":3506},{"style":492},[3507],{"type":62,"value":566},{"type":56,"tag":431,"props":3509,"children":3510},{"style":443},[3511],{"type":62,"value":3512},"containe",{"type":56,"tag":431,"props":3514,"children":3515},{"style":449},[3516],{"type":62,"value":3517},"r",{"type":56,"tag":431,"props":3519,"children":3520},{"style":492},[3521],{"type":62,"value":581},{"type":56,"tag":431,"props":3523,"children":3524},{"style":443},[3525],{"type":62,"value":3526}," find",{"type":56,"tag":431,"props":3528,"children":3529},{"style":443},[3530],{"type":62,"value":3531}," \u002Fdata\u002Fmodels",{"type":56,"tag":431,"props":3533,"children":3534},{"style":443},[3535],{"type":62,"value":3536}," -name",{"type":56,"tag":431,"props":3538,"children":3539},{"style":492},[3540],{"type":62,"value":495},{"type":56,"tag":431,"props":3542,"children":3543},{"style":443},[3544],{"type":62,"value":3545},"*tokenizer.model",{"type":56,"tag":431,"props":3547,"children":3548},{"style":492},[3549],{"type":62,"value":505},{"type":56,"tag":431,"props":3551,"children":3552},{"style":492},[3553],{"type":62,"value":3554}," |",{"type":56,"tag":431,"props":3556,"children":3557},{"style":437},[3558],{"type":62,"value":3559}," head",{"type":56,"tag":431,"props":3561,"children":3562},{"style":443},[3563],{"type":62,"value":3564}," -5\n",{"type":56,"tag":431,"props":3566,"children":3567},{"class":433,"line":455},[3568,3572,3577,3581,3585,3589,3593,3598,3603,3608,3612,3616,3621,3625,3630,3635,3639,3644],{"type":56,"tag":431,"props":3569,"children":3570},{"style":437},[3571],{"type":62,"value":1399},{"type":56,"tag":431,"props":3573,"children":3574},{"style":443},[3575],{"type":62,"value":3576}," cp",{"type":56,"tag":431,"props":3578,"children":3579},{"style":492},[3580],{"type":62,"value":566},{"type":56,"tag":431,"props":3582,"children":3583},{"style":443},[3584],{"type":62,"value":3512},{"type":56,"tag":431,"props":3586,"children":3587},{"style":449},[3588],{"type":62,"value":3517},{"type":56,"tag":431,"props":3590,"children":3591},{"style":492},[3592],{"type":62,"value":581},{"type":56,"tag":431,"props":3594,"children":3595},{"style":443},[3596],{"type":62,"value":3597},":\u002Fdata\u002Fmodels\u002F",{"type":56,"tag":431,"props":3599,"children":3600},{"style":492},[3601],{"type":62,"value":3602},"\u003C",{"type":56,"tag":431,"props":3604,"children":3605},{"style":443},[3606],{"type":62,"value":3607},"model-di",{"type":56,"tag":431,"props":3609,"children":3610},{"style":449},[3611],{"type":62,"value":3517},{"type":56,"tag":431,"props":3613,"children":3614},{"style":492},[3615],{"type":62,"value":581},{"type":56,"tag":431,"props":3617,"children":3618},{"style":443},[3619],{"type":62,"value":3620},"\u002F1\u002F",{"type":56,"tag":431,"props":3622,"children":3623},{"style":492},[3624],{"type":62,"value":3602},{"type":56,"tag":431,"props":3626,"children":3627},{"style":443},[3628],{"type":62,"value":3629},"has",{"type":56,"tag":431,"props":3631,"children":3632},{"style":449},[3633],{"type":62,"value":3634},"h",{"type":56,"tag":431,"props":3636,"children":3637},{"style":492},[3638],{"type":62,"value":581},{"type":56,"tag":431,"props":3640,"children":3641},{"style":443},[3642],{"type":62,"value":3643},"_tokenizer.model",{"type":56,"tag":431,"props":3645,"children":3646},{"style":443},[3647],{"type":62,"value":3648}," \u002Ftmp\u002Ftokenizer.model\n",{"type":56,"tag":69,"props":3650,"children":3651},{},[3652],{"type":56,"tag":73,"props":3653,"children":3654},{},[3655],{"type":62,"value":3656},"Step 2 — Generate token mapping with SentencePiece:",{"type":56,"tag":420,"props":3658,"children":3661},{"className":3659,"code":3660,"language":440,"meta":425,"style":425},"language-python shiki shiki-themes material-theme-lighter material-theme material-theme-palenight","import sentencepiece as spm\ns = spm.SentencePieceProcessor(model_file='\u002Ftmp\u002Ftokenizer.model')\n\nword_asr_predicts = \"ablooper\"       # what the model currently outputs\nword_asr_should_predict = \"Abloopar\" # what you want\n\ntokens = s.encode(word_asr_predicts, out_type=str)\nprint(word_asr_should_predict + ':' + '\u002F'.join(tokens))\n# e.g. Abloopar:▁a\u002Fb\u002Flo\u002Fop\u002Fer\n",[3662],{"type":56,"tag":152,"props":3663,"children":3664},{"__ignoreMap":425},[3665,3673,3681,3688,3696,3704,3711,3719,3727],{"type":56,"tag":431,"props":3666,"children":3667},{"class":433,"line":27},[3668],{"type":56,"tag":431,"props":3669,"children":3670},{},[3671],{"type":62,"value":3672},"import sentencepiece as spm\n",{"type":56,"tag":431,"props":3674,"children":3675},{"class":433,"line":455},[3676],{"type":56,"tag":431,"props":3677,"children":3678},{},[3679],{"type":62,"value":3680},"s = spm.SentencePieceProcessor(model_file='\u002Ftmp\u002Ftokenizer.model')\n",{"type":56,"tag":431,"props":3682,"children":3683},{"class":433,"line":478},[3684],{"type":56,"tag":431,"props":3685,"children":3686},{"emptyLinePlaceholder":1388},[3687],{"type":62,"value":1391},{"type":56,"tag":431,"props":3689,"children":3690},{"class":433,"line":512},[3691],{"type":56,"tag":431,"props":3692,"children":3693},{},[3694],{"type":62,"value":3695},"word_asr_predicts = \"ablooper\"       # what the model currently outputs\n",{"type":56,"tag":431,"props":3697,"children":3698},{"class":433,"line":555},[3699],{"type":56,"tag":431,"props":3700,"children":3701},{},[3702],{"type":62,"value":3703},"word_asr_should_predict = \"Abloopar\" # what you want\n",{"type":56,"tag":431,"props":3705,"children":3706},{"class":433,"line":23},[3707],{"type":56,"tag":431,"props":3708,"children":3709},{"emptyLinePlaceholder":1388},[3710],{"type":62,"value":1391},{"type":56,"tag":431,"props":3712,"children":3713},{"class":433,"line":1470},[3714],{"type":56,"tag":431,"props":3715,"children":3716},{},[3717],{"type":62,"value":3718},"tokens = s.encode(word_asr_predicts, out_type=str)\n",{"type":56,"tag":431,"props":3720,"children":3721},{"class":433,"line":1483},[3722],{"type":56,"tag":431,"props":3723,"children":3724},{},[3725],{"type":62,"value":3726},"print(word_asr_should_predict + ':' + '\u002F'.join(tokens))\n",{"type":56,"tag":431,"props":3728,"children":3729},{"class":433,"line":1501},[3730],{"type":56,"tag":431,"props":3731,"children":3732},{},[3733],{"type":62,"value":3734},"# e.g. Abloopar:▁a\u002Fb\u002Flo\u002Fop\u002Fer\n",{"type":56,"tag":69,"props":3736,"children":3737},{},[3738],{"type":56,"tag":73,"props":3739,"children":3740},{},[3741],{"type":62,"value":3742},"Step 3 — Pass mapping as boosted word:",{"type":56,"tag":420,"props":3744,"children":3746},{"className":422,"code":3745,"language":424,"meta":425,"style":425},"--boosted-lm-words 'Abloopar:▁a\u002Fb\u002Flo\u002Fop\u002Fer' --boosted-lm-score 80\n",[3747],{"type":56,"tag":152,"props":3748,"children":3749},{"__ignoreMap":425},[3750],{"type":56,"tag":431,"props":3751,"children":3752},{"class":433,"line":27},[3753,3758,3762,3766,3770,3775],{"type":56,"tag":431,"props":3754,"children":3755},{"style":437},[3756],{"type":62,"value":3757},"--boosted-lm-words",{"type":56,"tag":431,"props":3759,"children":3760},{"style":492},[3761],{"type":62,"value":1453},{"type":56,"tag":431,"props":3763,"children":3764},{"style":443},[3765],{"type":62,"value":3202},{"type":56,"tag":431,"props":3767,"children":3768},{"style":492},[3769],{"type":62,"value":1463},{"type":56,"tag":431,"props":3771,"children":3772},{"style":443},[3773],{"type":62,"value":3774}," --boosted-lm-score",{"type":56,"tag":431,"props":3776,"children":3777},{"style":1374},[3778],{"type":62,"value":3223},{"type":56,"tag":69,"props":3780,"children":3781},{},[3782,3787],{"type":56,"tag":73,"props":3783,"children":3784},{},[3785],{"type":62,"value":3786},"Important:",{"type":62,"value":3788}," Always use the actual model tokenizer from inside the container — test tokenizers generate different token sequences and the mapping will silently fail.",{"type":56,"tag":99,"props":3790,"children":3792},{"id":3791},"performance-benchmarking-self-hosted",[3793],{"type":62,"value":3794},"Performance Benchmarking (Self-Hosted)",{"type":56,"tag":69,"props":3796,"children":3797},{},[3798,3800,3806,3808,3813,3815,3821],{"type":62,"value":3799},"Use ",{"type":56,"tag":152,"props":3801,"children":3803},{"className":3802},[],[3804],{"type":62,"value":3805},"riva_streaming_asr_client",{"type":62,"value":3807}," — a ",{"type":56,"tag":73,"props":3809,"children":3810},{},[3811],{"type":62,"value":3812},"pre-built binary available in PATH inside the NIM container",{"type":62,"value":3814},". Do not use the Python script from python-clients; run the binary via ",{"type":56,"tag":152,"props":3816,"children":3818},{"className":3817},[],[3819],{"type":62,"value":3820},"docker exec",{"type":62,"value":282},{"type":56,"tag":69,"props":3823,"children":3824},{},[3825,3827,3833],{"type":62,"value":3826},"A sample LibriSpeech wav file is bundled at ",{"type":56,"tag":152,"props":3828,"children":3830},{"className":3829},[],[3831],{"type":62,"value":3832},"\u002Fopt\u002Friva\u002Fexamples\u002Fasr_lib\u002F1272-135031-0000.wav",{"type":62,"value":3834}," inside the container.",{"type":56,"tag":662,"props":3836,"children":3838},{"id":3837},"streaming-models",[3839],{"type":62,"value":3840},"Streaming Models",{"type":56,"tag":69,"props":3842,"children":3843},{},[3844,3846,3852,3854,3860],{"type":62,"value":3845},"Run at increasing concurrency levels (1, 2, 4, 8, …). Set ",{"type":56,"tag":152,"props":3847,"children":3849},{"className":3848},[],[3850],{"type":62,"value":3851},"num_iterations",{"type":62,"value":3853}," to 3× ",{"type":56,"tag":152,"props":3855,"children":3857},{"className":3856},[],[3858],{"type":62,"value":3859},"num_parallel_requests",{"type":62,"value":3861}," for stable results.",{"type":56,"tag":420,"props":3863,"children":3865},{"className":422,"code":3864,"language":424,"meta":425,"style":425},"export N=4  # num parallel streams — sweep: 1, 2, 4, 8, ...\n\ndocker exec \u003Ccontainer_name> riva_streaming_asr_client \\\n  --riva_uri=0.0.0.0:50051 \\\n  --language_code=en-US \\\n  --audio_file=\u002Fopt\u002Friva\u002Fexamples\u002Fasr_lib\u002F1272-135031-0000.wav \\\n  --chunk_duration_ms=160 \\\n  --simulate_realtime=true \\\n  --automatic_punctuation=true \\\n  --num_parallel_requests=$N \\\n  --num_iterations=$((3 * N)) \\\n  --word_time_offsets=false \\\n  --print_transcripts=false \\\n  --interim_results=false \\\n  --output_filename=\u002Ftmp\u002Foutput.json\n",[3866],{"type":56,"tag":152,"props":3867,"children":3868},{"__ignoreMap":425},[3869,3895,3902,3940,3952,3964,3976,3988,4000,4012,4029,4065,4077,4089,4101],{"type":56,"tag":431,"props":3870,"children":3871},{"class":433,"line":27},[3872,3876,3881,3885,3890],{"type":56,"tag":431,"props":3873,"children":3874},{"style":1204},[3875],{"type":62,"value":1207},{"type":56,"tag":431,"props":3877,"children":3878},{"style":449},[3879],{"type":62,"value":3880}," N",{"type":56,"tag":431,"props":3882,"children":3883},{"style":492},[3884],{"type":62,"value":1244},{"type":56,"tag":431,"props":3886,"children":3887},{"style":1374},[3888],{"type":62,"value":3889},"4",{"type":56,"tag":431,"props":3891,"children":3892},{"style":1782},[3893],{"type":62,"value":3894},"  # num parallel streams — sweep: 1, 2, 4, 8, ...\n",{"type":56,"tag":431,"props":3896,"children":3897},{"class":433,"line":455},[3898],{"type":56,"tag":431,"props":3899,"children":3900},{"emptyLinePlaceholder":1388},[3901],{"type":62,"value":1391},{"type":56,"tag":431,"props":3903,"children":3904},{"class":433,"line":478},[3905,3909,3913,3917,3922,3927,3931,3936],{"type":56,"tag":431,"props":3906,"children":3907},{"style":437},[3908],{"type":62,"value":1399},{"type":56,"tag":431,"props":3910,"children":3911},{"style":443},[3912],{"type":62,"value":3503},{"type":56,"tag":431,"props":3914,"children":3915},{"style":492},[3916],{"type":62,"value":566},{"type":56,"tag":431,"props":3918,"children":3919},{"style":443},[3920],{"type":62,"value":3921},"container_nam",{"type":56,"tag":431,"props":3923,"children":3924},{"style":449},[3925],{"type":62,"value":3926},"e",{"type":56,"tag":431,"props":3928,"children":3929},{"style":492},[3930],{"type":62,"value":581},{"type":56,"tag":431,"props":3932,"children":3933},{"style":443},[3934],{"type":62,"value":3935}," riva_streaming_asr_client",{"type":56,"tag":431,"props":3937,"children":3938},{"style":449},[3939],{"type":62,"value":452},{"type":56,"tag":431,"props":3941,"children":3942},{"class":433,"line":512},[3943,3948],{"type":56,"tag":431,"props":3944,"children":3945},{"style":443},[3946],{"type":62,"value":3947},"  --riva_uri=0.0.0.0:50051",{"type":56,"tag":431,"props":3949,"children":3950},{"style":449},[3951],{"type":62,"value":452},{"type":56,"tag":431,"props":3953,"children":3954},{"class":433,"line":555},[3955,3960],{"type":56,"tag":431,"props":3956,"children":3957},{"style":443},[3958],{"type":62,"value":3959},"  --language_code=en-US",{"type":56,"tag":431,"props":3961,"children":3962},{"style":449},[3963],{"type":62,"value":452},{"type":56,"tag":431,"props":3965,"children":3966},{"class":433,"line":23},[3967,3972],{"type":56,"tag":431,"props":3968,"children":3969},{"style":443},[3970],{"type":62,"value":3971},"  --audio_file=\u002Fopt\u002Friva\u002Fexamples\u002Fasr_lib\u002F1272-135031-0000.wav",{"type":56,"tag":431,"props":3973,"children":3974},{"style":449},[3975],{"type":62,"value":452},{"type":56,"tag":431,"props":3977,"children":3978},{"class":433,"line":1470},[3979,3984],{"type":56,"tag":431,"props":3980,"children":3981},{"style":443},[3982],{"type":62,"value":3983},"  --chunk_duration_ms=160",{"type":56,"tag":431,"props":3985,"children":3986},{"style":449},[3987],{"type":62,"value":452},{"type":56,"tag":431,"props":3989,"children":3990},{"class":433,"line":1483},[3991,3996],{"type":56,"tag":431,"props":3992,"children":3993},{"style":443},[3994],{"type":62,"value":3995},"  --simulate_realtime=true",{"type":56,"tag":431,"props":3997,"children":3998},{"style":449},[3999],{"type":62,"value":452},{"type":56,"tag":431,"props":4001,"children":4002},{"class":433,"line":1501},[4003,4008],{"type":56,"tag":431,"props":4004,"children":4005},{"style":443},[4006],{"type":62,"value":4007},"  --automatic_punctuation=true",{"type":56,"tag":431,"props":4009,"children":4010},{"style":449},[4011],{"type":62,"value":452},{"type":56,"tag":431,"props":4013,"children":4014},{"class":433,"line":1517},[4015,4020,4025],{"type":56,"tag":431,"props":4016,"children":4017},{"style":443},[4018],{"type":62,"value":4019},"  --num_parallel_requests=",{"type":56,"tag":431,"props":4021,"children":4022},{"style":449},[4023],{"type":62,"value":4024},"$N",{"type":56,"tag":431,"props":4026,"children":4027},{"style":449},[4028],{"type":62,"value":452},{"type":56,"tag":431,"props":4030,"children":4031},{"class":433,"line":1539},[4032,4037,4042,4047,4052,4056,4061],{"type":56,"tag":431,"props":4033,"children":4034},{"style":443},[4035],{"type":62,"value":4036},"  --num_iterations=",{"type":56,"tag":431,"props":4038,"children":4039},{"style":492},[4040],{"type":62,"value":4041},"$((",{"type":56,"tag":431,"props":4043,"children":4044},{"style":437},[4045],{"type":62,"value":4046},"3",{"type":56,"tag":431,"props":4048,"children":4049},{"style":449},[4050],{"type":62,"value":4051}," *",{"type":56,"tag":431,"props":4053,"children":4054},{"style":443},[4055],{"type":62,"value":3880},{"type":56,"tag":431,"props":4057,"children":4058},{"style":492},[4059],{"type":62,"value":4060},"))",{"type":56,"tag":431,"props":4062,"children":4063},{"style":449},[4064],{"type":62,"value":452},{"type":56,"tag":431,"props":4066,"children":4067},{"class":433,"line":1561},[4068,4073],{"type":56,"tag":431,"props":4069,"children":4070},{"style":443},[4071],{"type":62,"value":4072},"  --word_time_offsets=false",{"type":56,"tag":431,"props":4074,"children":4075},{"style":449},[4076],{"type":62,"value":452},{"type":56,"tag":431,"props":4078,"children":4079},{"class":433,"line":1579},[4080,4085],{"type":56,"tag":431,"props":4081,"children":4082},{"style":443},[4083],{"type":62,"value":4084},"  --print_transcripts=false",{"type":56,"tag":431,"props":4086,"children":4087},{"style":449},[4088],{"type":62,"value":452},{"type":56,"tag":431,"props":4090,"children":4091},{"class":433,"line":1596},[4092,4097],{"type":56,"tag":431,"props":4093,"children":4094},{"style":443},[4095],{"type":62,"value":4096},"  --interim_results=false",{"type":56,"tag":431,"props":4098,"children":4099},{"style":449},[4100],{"type":62,"value":452},{"type":56,"tag":431,"props":4102,"children":4103},{"class":433,"line":1619},[4104],{"type":56,"tag":431,"props":4105,"children":4106},{"style":443},[4107],{"type":62,"value":4108},"  --output_filename=\u002Ftmp\u002Foutput.json\n",{"type":56,"tag":662,"props":4110,"children":4112},{"id":4111},"offline-models",[4113],{"type":62,"value":4114},"Offline Models",{"type":56,"tag":420,"props":4116,"children":4118},{"className":422,"code":4117,"language":424,"meta":425,"style":425},"export N=4  # num parallel requests — sweep: 1, 2, 4, 8, ...\n\ndocker exec \u003Ccontainer_name> riva_streaming_asr_client \\\n  --riva_uri=0.0.0.0:50051 \\\n  --language_code=en-US \\\n  --audio_file=\u002Fopt\u002Friva\u002Fexamples\u002Fasr_lib\u002F1272-135031-0000.wav \\\n  --automatic_punctuation=true \\\n  --num_parallel_requests=$N \\\n  --num_iterations=$((3 * N)) \\\n  --word_time_offsets=false \\\n  --print_transcripts=false \\\n  --interim_results=false \\\n  --output_filename=\u002Ftmp\u002Foutput.json\n",[4119],{"type":56,"tag":152,"props":4120,"children":4121},{"__ignoreMap":425},[4122,4146,4153,4188,4199,4210,4221,4232,4247,4278,4289,4300,4311],{"type":56,"tag":431,"props":4123,"children":4124},{"class":433,"line":27},[4125,4129,4133,4137,4141],{"type":56,"tag":431,"props":4126,"children":4127},{"style":1204},[4128],{"type":62,"value":1207},{"type":56,"tag":431,"props":4130,"children":4131},{"style":449},[4132],{"type":62,"value":3880},{"type":56,"tag":431,"props":4134,"children":4135},{"style":492},[4136],{"type":62,"value":1244},{"type":56,"tag":431,"props":4138,"children":4139},{"style":1374},[4140],{"type":62,"value":3889},{"type":56,"tag":431,"props":4142,"children":4143},{"style":1782},[4144],{"type":62,"value":4145},"  # num parallel requests — sweep: 1, 2, 4, 8, ...\n",{"type":56,"tag":431,"props":4147,"children":4148},{"class":433,"line":455},[4149],{"type":56,"tag":431,"props":4150,"children":4151},{"emptyLinePlaceholder":1388},[4152],{"type":62,"value":1391},{"type":56,"tag":431,"props":4154,"children":4155},{"class":433,"line":478},[4156,4160,4164,4168,4172,4176,4180,4184],{"type":56,"tag":431,"props":4157,"children":4158},{"style":437},[4159],{"type":62,"value":1399},{"type":56,"tag":431,"props":4161,"children":4162},{"style":443},[4163],{"type":62,"value":3503},{"type":56,"tag":431,"props":4165,"children":4166},{"style":492},[4167],{"type":62,"value":566},{"type":56,"tag":431,"props":4169,"children":4170},{"style":443},[4171],{"type":62,"value":3921},{"type":56,"tag":431,"props":4173,"children":4174},{"style":449},[4175],{"type":62,"value":3926},{"type":56,"tag":431,"props":4177,"children":4178},{"style":492},[4179],{"type":62,"value":581},{"type":56,"tag":431,"props":4181,"children":4182},{"style":443},[4183],{"type":62,"value":3935},{"type":56,"tag":431,"props":4185,"children":4186},{"style":449},[4187],{"type":62,"value":452},{"type":56,"tag":431,"props":4189,"children":4190},{"class":433,"line":512},[4191,4195],{"type":56,"tag":431,"props":4192,"children":4193},{"style":443},[4194],{"type":62,"value":3947},{"type":56,"tag":431,"props":4196,"children":4197},{"style":449},[4198],{"type":62,"value":452},{"type":56,"tag":431,"props":4200,"children":4201},{"class":433,"line":555},[4202,4206],{"type":56,"tag":431,"props":4203,"children":4204},{"style":443},[4205],{"type":62,"value":3959},{"type":56,"tag":431,"props":4207,"children":4208},{"style":449},[4209],{"type":62,"value":452},{"type":56,"tag":431,"props":4211,"children":4212},{"class":433,"line":23},[4213,4217],{"type":56,"tag":431,"props":4214,"children":4215},{"style":443},[4216],{"type":62,"value":3971},{"type":56,"tag":431,"props":4218,"children":4219},{"style":449},[4220],{"type":62,"value":452},{"type":56,"tag":431,"props":4222,"children":4223},{"class":433,"line":1470},[4224,4228],{"type":56,"tag":431,"props":4225,"children":4226},{"style":443},[4227],{"type":62,"value":4007},{"type":56,"tag":431,"props":4229,"children":4230},{"style":449},[4231],{"type":62,"value":452},{"type":56,"tag":431,"props":4233,"children":4234},{"class":433,"line":1483},[4235,4239,4243],{"type":56,"tag":431,"props":4236,"children":4237},{"style":443},[4238],{"type":62,"value":4019},{"type":56,"tag":431,"props":4240,"children":4241},{"style":449},[4242],{"type":62,"value":4024},{"type":56,"tag":431,"props":4244,"children":4245},{"style":449},[4246],{"type":62,"value":452},{"type":56,"tag":431,"props":4248,"children":4249},{"class":433,"line":1501},[4250,4254,4258,4262,4266,4270,4274],{"type":56,"tag":431,"props":4251,"children":4252},{"style":443},[4253],{"type":62,"value":4036},{"type":56,"tag":431,"props":4255,"children":4256},{"style":492},[4257],{"type":62,"value":4041},{"type":56,"tag":431,"props":4259,"children":4260},{"style":437},[4261],{"type":62,"value":4046},{"type":56,"tag":431,"props":4263,"children":4264},{"style":449},[4265],{"type":62,"value":4051},{"type":56,"tag":431,"props":4267,"children":4268},{"style":443},[4269],{"type":62,"value":3880},{"type":56,"tag":431,"props":4271,"children":4272},{"style":492},[4273],{"type":62,"value":4060},{"type":56,"tag":431,"props":4275,"children":4276},{"style":449},[4277],{"type":62,"value":452},{"type":56,"tag":431,"props":4279,"children":4280},{"class":433,"line":1517},[4281,4285],{"type":56,"tag":431,"props":4282,"children":4283},{"style":443},[4284],{"type":62,"value":4072},{"type":56,"tag":431,"props":4286,"children":4287},{"style":449},[4288],{"type":62,"value":452},{"type":56,"tag":431,"props":4290,"children":4291},{"class":433,"line":1539},[4292,4296],{"type":56,"tag":431,"props":4293,"children":4294},{"style":443},[4295],{"type":62,"value":4084},{"type":56,"tag":431,"props":4297,"children":4298},{"style":449},[4299],{"type":62,"value":452},{"type":56,"tag":431,"props":4301,"children":4302},{"class":433,"line":1561},[4303,4307],{"type":56,"tag":431,"props":4304,"children":4305},{"style":443},[4306],{"type":62,"value":4096},{"type":56,"tag":431,"props":4308,"children":4309},{"style":449},[4310],{"type":62,"value":452},{"type":56,"tag":431,"props":4312,"children":4313},{"class":433,"line":1579},[4314],{"type":56,"tag":431,"props":4315,"children":4316},{"style":443},[4317],{"type":62,"value":4108},{"type":56,"tag":69,"props":4319,"children":4320},{},[4321,4323,4329,4330,4336],{"type":62,"value":4322},"Note: Omit ",{"type":56,"tag":152,"props":4324,"children":4326},{"className":4325},[],[4327],{"type":62,"value":4328},"--chunk_duration_ms",{"type":62,"value":300},{"type":56,"tag":152,"props":4331,"children":4333},{"className":4332},[],[4334],{"type":62,"value":4335},"--simulate_realtime",{"type":62,"value":4337}," for offline models — they process the full audio in one shot, not in streaming chunks.",{"type":56,"tag":69,"props":4339,"children":4340},{},[4341],{"type":56,"tag":73,"props":4342,"children":4343},{},[4344],{"type":62,"value":4345},"Key flags:",{"type":56,"tag":681,"props":4347,"children":4348},{},[4349,4365],{"type":56,"tag":685,"props":4350,"children":4351},{},[4352],{"type":56,"tag":689,"props":4353,"children":4354},{},[4355,4360],{"type":56,"tag":693,"props":4356,"children":4357},{},[4358],{"type":62,"value":4359},"Flag",{"type":56,"tag":693,"props":4361,"children":4362},{},[4363],{"type":62,"value":4364},"Description",{"type":56,"tag":719,"props":4366,"children":4367},{},[4368,4392,4408,4425,4449],{"type":56,"tag":689,"props":4369,"children":4370},{},[4371,4379],{"type":56,"tag":726,"props":4372,"children":4373},{},[4374],{"type":56,"tag":152,"props":4375,"children":4377},{"className":4376},[],[4378],{"type":62,"value":4328},{"type":56,"tag":726,"props":4380,"children":4381},{},[4382,4384,4390],{"type":62,"value":4383},"Match your deployed ",{"type":56,"tag":152,"props":4385,"children":4387},{"className":4386},[],[4388],{"type":62,"value":4389},"chunk_size",{"type":62,"value":4391}," in ms — streaming models only",{"type":56,"tag":689,"props":4393,"children":4394},{},[4395,4403],{"type":56,"tag":726,"props":4396,"children":4397},{},[4398],{"type":56,"tag":152,"props":4399,"children":4401},{"className":4400},[],[4402],{"type":62,"value":4335},{"type":56,"tag":726,"props":4404,"children":4405},{},[4406],{"type":62,"value":4407},"Throttle audio to real-time speed — streaming models only",{"type":56,"tag":689,"props":4409,"children":4410},{},[4411,4420],{"type":56,"tag":726,"props":4412,"children":4413},{},[4414],{"type":56,"tag":152,"props":4415,"children":4417},{"className":4416},[],[4418],{"type":62,"value":4419},"--num_parallel_requests",{"type":56,"tag":726,"props":4421,"children":4422},{},[4423],{"type":62,"value":4424},"Concurrent streams; sweep 1→2→4→8→… to find throughput peak",{"type":56,"tag":689,"props":4426,"children":4427},{},[4428,4437],{"type":56,"tag":726,"props":4429,"children":4430},{},[4431],{"type":56,"tag":152,"props":4432,"children":4434},{"className":4433},[],[4435],{"type":62,"value":4436},"--num_iterations",{"type":56,"tag":726,"props":4438,"children":4439},{},[4440,4442,4447],{"type":62,"value":4441},"Total requests; use 3× ",{"type":56,"tag":152,"props":4443,"children":4445},{"className":4444},[],[4446],{"type":62,"value":3859},{"type":62,"value":4448}," for stable results",{"type":56,"tag":689,"props":4450,"children":4451},{},[4452,4461],{"type":56,"tag":726,"props":4453,"children":4454},{},[4455],{"type":56,"tag":152,"props":4456,"children":4458},{"className":4457},[],[4459],{"type":62,"value":4460},"--print_transcripts=false",{"type":56,"tag":726,"props":4462,"children":4463},{},[4464],{"type":62,"value":4465},"Suppress transcripts for clean benchmark output",{"type":56,"tag":69,"props":4467,"children":4468},{},[4469],{"type":56,"tag":73,"props":4470,"children":4471},{},[4472],{"type":62,"value":4473},"Output metrics:",{"type":56,"tag":681,"props":4475,"children":4476},{},[4477,4492],{"type":56,"tag":685,"props":4478,"children":4479},{},[4480],{"type":56,"tag":689,"props":4481,"children":4482},{},[4483,4488],{"type":56,"tag":693,"props":4484,"children":4485},{},[4486],{"type":62,"value":4487},"Metric",{"type":56,"tag":693,"props":4489,"children":4490},{},[4491],{"type":62,"value":4364},{"type":56,"tag":719,"props":4493,"children":4494},{},[4495,4508],{"type":56,"tag":689,"props":4496,"children":4497},{},[4498,4503],{"type":56,"tag":726,"props":4499,"children":4500},{},[4501],{"type":62,"value":4502},"Median \u002F 90th \u002F 95th \u002F 99th latency",{"type":56,"tag":726,"props":4504,"children":4505},{},[4506],{"type":62,"value":4507},"Time from chunk sent to partial transcript received (ms)",{"type":56,"tag":689,"props":4509,"children":4510},{},[4511,4516],{"type":56,"tag":726,"props":4512,"children":4513},{},[4514],{"type":62,"value":4515},"Throughput (RTFX)",{"type":56,"tag":726,"props":4517,"children":4518},{},[4519],{"type":62,"value":4520},"Audio processed per second of wall time; >1.0 = faster than real-time",{"type":56,"tag":99,"props":4522,"children":4524},{"id":4523},"limitations",[4525],{"type":62,"value":4526},"Limitations",{"type":56,"tag":142,"props":4528,"children":4529},{},[4530,4535,4540,4552,4557],{"type":56,"tag":146,"props":4531,"children":4532},{},[4533],{"type":62,"value":4534},"x86_64 architecture only — ARM is not supported",{"type":56,"tag":146,"props":4536,"children":4537},{},[4538],{"type":62,"value":4539},"Self-hosted deployment requires an NVIDIA AI Enterprise license",{"type":56,"tag":146,"props":4541,"children":4542},{},[4543,4545,4550],{"type":62,"value":4544},"Cloud-hosted inference requires an active ",{"type":56,"tag":152,"props":4546,"children":4548},{"className":4547},[],[4549],{"type":62,"value":192},{"type":62,"value":4551}," and internet access",{"type":56,"tag":146,"props":4553,"children":4554},{},[4555],{"type":62,"value":4556},"Word\u002Ftoken boosting is advisory — does not guarantee transcription of boosted terms",{"type":56,"tag":146,"props":4558,"children":4559},{},[4560],{"type":62,"value":4561},"TDT models have a silence-based end-of-utterance threshold in streaming mode",{"type":56,"tag":99,"props":4563,"children":4565},{"id":4564},"next-steps",[4566],{"type":62,"value":4567},"Next Steps",{"type":56,"tag":142,"props":4569,"children":4570},{},[4571,4582,4593],{"type":56,"tag":146,"props":4572,"children":4573},{},[4574,4576],{"type":62,"value":4575},"Customize ASR pipeline (VAD, diarization, language model): see ",{"type":56,"tag":152,"props":4577,"children":4579},{"className":4578},[],[4580],{"type":62,"value":4581},"riva-pipelines",{"type":56,"tag":146,"props":4583,"children":4584},{},[4585,4587],{"type":62,"value":4586},"Deploy a custom-trained model: see ",{"type":56,"tag":152,"props":4588,"children":4590},{"className":4589},[],[4591],{"type":62,"value":4592},"riva-asr-custom",{"type":56,"tag":146,"props":4594,"children":4595},{},[4596,4598],{"type":62,"value":4597},"Check system requirements: see ",{"type":56,"tag":152,"props":4599,"children":4601},{"className":4600},[],[4602],{"type":62,"value":4603},"riva-ops",{"type":56,"tag":4605,"props":4606,"children":4607},"style",{},[4608],{"type":62,"value":4609},"html .light .shiki span {color: var(--shiki-light);background: var(--shiki-light-bg);font-style: var(--shiki-light-font-style);font-weight: var(--shiki-light-font-weight);text-decoration: var(--shiki-light-text-decoration);}html.light .shiki span {color: var(--shiki-light);background: var(--shiki-light-bg);font-style: var(--shiki-light-font-style);font-weight: var(--shiki-light-font-weight);text-decoration: var(--shiki-light-text-decoration);}html .default .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}html.dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}",{"items":4611,"total":555},[4612,4629,4636,4648,4659],{"slug":4613,"name":4613,"fn":4614,"description":4615,"org":4616,"tags":4617,"stars":23,"repoUrl":24,"updatedAt":4628},"nemo-clinical-data-designer","generate synthetic clinical datasets","Use when generating synthetic tabular datasets via Data Designer — sampler columns, LLM columns, custom generators. Not for ASR audio.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[4618,4621,4624,4627],{"name":4619,"slug":4620,"type":13},"Data Analysis","data-analysis",{"name":4622,"slug":4623,"type":13},"Datasets","datasets",{"name":4625,"slug":4626,"type":13},"Life Sciences","life-sciences",{"name":9,"slug":8,"type":13},"2026-07-14T05:35:42.770246",{"slug":4,"name":4,"fn":5,"description":6,"org":4630,"tags":4631,"stars":23,"repoUrl":24,"updatedAt":25},{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[4632,4633,4634,4635],{"name":15,"slug":16,"type":13},{"name":9,"slug":8,"type":13},{"name":21,"slug":22,"type":13},{"name":18,"slug":19,"type":13},{"slug":4592,"name":4592,"fn":4637,"description":4638,"org":4639,"tags":4640,"stars":23,"repoUrl":24,"updatedAt":4647},"deploy custom Riva ASR models","Use when the user wants to deploy a custom-trained ASR model as a Riva NIM, or convert a NeMo model via nemo2riva \u002F riva-build \u002F riva-deploy \u002F RMIR.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[4641,4642,4645,4646],{"name":15,"slug":16,"type":13},{"name":4643,"slug":4644,"type":13},"Deployment","deployment",{"name":9,"slug":8,"type":13},{"name":21,"slug":22,"type":13},"2026-07-14T05:35:44.009948",{"slug":157,"name":157,"fn":4649,"description":4650,"org":4651,"tags":4652,"stars":23,"repoUrl":24,"updatedAt":4658},"set up NVIDIA Riva Speech NIMs","Use when getting started with NVIDIA Riva Speech NIMs: NGC access, Docker login for nvcr.io, NVIDIA Container Toolkit, GPU verification, Riva Python client.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[4653,4654,4656,4657],{"name":15,"slug":16,"type":13},{"name":4655,"slug":1399,"type":13},"Docker",{"name":9,"slug":8,"type":13},{"name":21,"slug":22,"type":13},"2026-07-14T05:35:50.295877",{"slug":4660,"name":4660,"fn":4661,"description":4662,"org":4663,"tags":4664,"stars":23,"repoUrl":24,"updatedAt":4673},"riva-tts","deploy and run Riva TTS models","Use when the user wants to deploy, run, or test a TTS (speech-synthesis) Riva NIM — cloud-hosted (build.nvidia.com) or self-hosted Magpie \u002F voice cloning.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[4665,4666,4669,4670],{"name":15,"slug":16,"type":13},{"name":4667,"slug":4668,"type":13},"Audio","audio",{"name":9,"slug":8,"type":13},{"name":4671,"slug":4672,"type":13},"Text-to-Speech","text-to-speech","2026-07-14T05:35:45.262923",{"items":4675,"total":4829},[4676,4694,4709,4720,4732,4746,4759,4773,4784,4795,4809,4818],{"slug":4677,"name":4677,"fn":4678,"description":4679,"org":4680,"tags":4681,"stars":4691,"repoUrl":4692,"updatedAt":4693},"nemoclaw-user-guide","retrieve NemoClaw documentation and configuration","Guides human users' AI agents to the NemoClaw docs MCP server and canonical Fern documentation in Markdown form. Use when users ask how to install, configure, operate, troubleshoot, secure, or learn NemoClaw with an AI coding assistant. Trigger keywords - nemoclaw docs, use nemoclaw with ai agent, nemoclaw mcp docs, nemoclaw install help, nemoclaw quickstart, nemoclaw markdown docs, llms.txt, agent skills.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[4682,4685,4688],{"name":4683,"slug":4684,"type":13},"Documentation","documentation",{"name":4686,"slug":4687,"type":13},"MCP","mcp",{"name":4689,"slug":4690,"type":13},"Search","search",21777,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FNemoClaw","2026-07-20T06:00:01.461044",{"slug":4695,"name":4695,"fn":4696,"description":4697,"org":4698,"tags":4699,"stars":4706,"repoUrl":4707,"updatedAt":4708},"mcore-build-and-dependency","manage Megatron-LM development environments","Container-based dev environment setup and dependency management for Megatron-LM. Covers acquiring and launching the CI container, uv package management, and updating uv.lock.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[4700,4703,4704],{"name":4701,"slug":4702,"type":13},"Containers","containers",{"name":4643,"slug":4644,"type":13},{"name":4705,"slug":440,"type":13},"Python",17049,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FMegatron-LM","2026-07-27T06:06:11.249662",{"slug":4710,"name":4710,"fn":4711,"description":4712,"org":4713,"tags":4714,"stars":4706,"repoUrl":4707,"updatedAt":4719},"mcore-bump-base-image","update NVIDIA PyTorch base images","Bump the NVIDIA PyTorch base image (`nvcr.io\u002Fnvidia\u002Fpytorch:YY.MM-py3`) used by Megatron-LM CI. Covers the two pin sites (GitHub CI in `docker\u002F.ngc_version.dev` and GitLab CI in `.gitlab\u002Fstages\u002F01.build.yml`), the post-bump CI loop (re-run functional tests, refresh golden values, mark broken tests), and the gotchas that bit PRs",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[4715,4718],{"name":4716,"slug":4717,"type":13},"CI\u002FCD","ci-cd",{"name":4643,"slug":4644,"type":13},"2026-07-14T05:25:59.97109",{"slug":4721,"name":4721,"fn":4722,"description":4723,"org":4724,"tags":4725,"stars":4706,"repoUrl":4707,"updatedAt":4731},"mcore-cicd","manage CI\u002FCD pipelines for Megatron-LM","CI\u002FCD reference for Megatron-LM. Covers CI pipeline structure, PR scope labels, triggering internal GitLab CI (which force-pushes the current branch to a pull-request\u002FBRANCH ref — always dry-run and verify the destination first; never run against shared or protected branches), and CI failure investigation.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[4726,4727,4728],{"name":4716,"slug":4717,"type":13},{"name":4643,"slug":4644,"type":13},{"name":4729,"slug":4730,"type":13},"GitHub","github","2026-07-27T06:06:12.278222",{"slug":4733,"name":4733,"fn":4734,"description":4735,"org":4736,"tags":4737,"stars":4706,"repoUrl":4707,"updatedAt":4745},"mcore-create-issue","investigate CI failures and create issues","Investigate a failing GitHub Actions run or job and create a GitHub issue for the failure.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[4738,4741,4742],{"name":4739,"slug":4740,"type":13},"Debugging","debugging",{"name":4729,"slug":4730,"type":13},{"name":4743,"slug":4744,"type":13},"Triage","triage","2026-07-14T05:25:57.442089",{"slug":4747,"name":4747,"fn":4748,"description":4749,"org":4750,"tags":4751,"stars":4706,"repoUrl":4707,"updatedAt":4758},"mcore-linting-and-formatting","lint and format Megatron-LM code","Linting and formatting for Megatron-LM. Covers running autoformat.sh, tools (ruff, black, isort, pylint, mypy), and code style rules.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[4752,4755],{"name":4753,"slug":4754,"type":13},"Best Practices","best-practices",{"name":4756,"slug":4757,"type":13},"Code Analysis","code-analysis","2026-07-14T05:25:56.18433",{"slug":4760,"name":4760,"fn":4761,"description":4762,"org":4763,"tags":4764,"stars":4706,"repoUrl":4707,"updatedAt":4772},"mcore-migrate-gpt-to-hybrid","migrate Megatron-LM models to HybridModel","Migration guide for moving Megatron Core GPTModel checkpoints, model providers, training commands, and layer mappings to HybridModel.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[4765,4768,4771],{"name":4766,"slug":4767,"type":13},"Machine Learning","machine-learning",{"name":4769,"slug":4770,"type":13},"Migration","migration",{"name":9,"slug":8,"type":13},"2026-07-17T06:07:11.777011",{"slug":4774,"name":4774,"fn":4775,"description":4776,"org":4777,"tags":4778,"stars":4706,"repoUrl":4707,"updatedAt":4783},"mcore-onboard-gb200-1node-tests","onboard functional tests for GB200","Onboard 1-node GitHub MR functional tests for GB200 from existing mr-scoped 2-node tests.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[4779,4782],{"name":4780,"slug":4781,"type":13},"QA","qa",{"name":18,"slug":19,"type":13},"2026-07-14T05:25:53.673039",{"slug":4785,"name":4785,"fn":4786,"description":4787,"org":4788,"tags":4789,"stars":4706,"repoUrl":4707,"updatedAt":4794},"mcore-run-on-slurm","launch distributed training jobs on SLURM","How to launch distributed Megatron-LM training jobs on a SLURM cluster. Covers a minimal sbatch skeleton, environment-variable setup for torch.distributed.run, CUDA_DEVICE_MAX_CONNECTIONS rules across hardware and parallelism modes, container conventions, monitoring, and per-rank failure diagnosis.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[4790,4791],{"name":4643,"slug":4644,"type":13},{"name":4792,"slug":4793,"type":13},"Infrastructure","infrastructure","2026-07-14T05:25:49.362534",{"slug":4796,"name":4796,"fn":4797,"description":4798,"org":4799,"tags":4800,"stars":4706,"repoUrl":4707,"updatedAt":4808},"mcore-split-pr","split pull requests to reduce review load","Split a PR into multiple PRs to reduce the number of required CODEOWNERS reviewer groups.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[4801,4804,4805],{"name":4802,"slug":4803,"type":13},"Code Review","code-review",{"name":4729,"slug":4730,"type":13},{"name":4806,"slug":4807,"type":13},"Pull Requests","pull-requests","2026-07-14T05:26:01.226578",{"slug":4810,"name":4810,"fn":4811,"description":4812,"org":4813,"tags":4814,"stars":4706,"repoUrl":4707,"updatedAt":4817},"mcore-testing","run and manage Megatron-LM tests","Test system for Megatron-LM. Covers test layout, recipe YAML structure, adding and running unit and functional tests, golden values, marker filters, and CI parity.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[4815,4816],{"name":4780,"slug":4781,"type":13},{"name":18,"slug":19,"type":13},"2026-07-14T05:25:54.928983",{"slug":4819,"name":4819,"fn":4820,"description":4821,"org":4822,"tags":4823,"stars":4706,"repoUrl":4707,"updatedAt":4828},"nightly-sync","manage nightly main-to-dev sync workflows","Domain knowledge for the nightly main-to-dev sync workflow. Covers merge strategy, CI architecture, failure investigation, and known issues.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[4824,4827],{"name":4825,"slug":4826,"type":13},"Automation","automation",{"name":4716,"slug":4717,"type":13},"2026-07-30T05:29:03.275638",496]