fa32m_streaming_bpe1024_final.nemo — final NVIDIA NeMo archive.
final_stats.json — export metadata from the training box.
train32_full.py — training script used for the run.
benchmark_32m_gold69_fleurs.py — benchmark/eval harness.
manifest_stats.json — manifest row/hour stats.
benchmark_results/fa32m_full9669_gold69_fleurs_summary_20260614T225848Z.json — benchmark summary.
benchmark_results/fa32m_full9669_gold69_fleurs_benchmark_20260614T225848Z.log — benchmark log.

Provenance

Run directory: /workspace/train32_runs/fa32m_streaming_bpe1024_full9669_20260614T130932Z
Source artifact: /workspace/train32_runs/fa32m_streaming_bpe1024_full9669_20260614T130932Z/fa32m_streaming_bpe1024_final.nemo
Base model: nvidia/stt_en_fastconformer_hybrid_medium_streaming_80ms
Tokenizer dir: /workspace/train32_bootstrap/tokenizers/fa_bpe_1024
SHA256 (fa32m_streaming_bpe1024_final.nemo): 034fb2afa19da13db8a120970a7f8d3e696987014cc62684ce50a1382d332448
Size: 127170560 bytes
Uploaded/updated: 2026-06-14T23:00:02Z

Manifest stats

json

{
  "train": {
    "rows": 5578135,
    "hours": 8862.916194048803
  },
  "val": {
    "rows": 56155,
    "hours": 89.6981422569458
  },
  "total": {
    "rows": 5634290,
    "hours": 8952.614336305749
  }
}

Loading

This is a NeMo .nemo artifact. Example:

python

import nemo.collections.asr as nemo_asr
model = nemo_asr.models.ASRModel.restore_from("fa32m_streaming_bpe1024_final.nemo")

Reza2kn/visualears-fastconformer-fa32m-streaming-bpe1024

VisualEars FastConformer FA32M Streaming BPE1024

Benchmark

Files

Provenance

Manifest stats

Loading

Reza2kn/
visualears-fastconformer-fa32m-streaming-bpe1024