Reza2kn/visualears-fastconformer-fa-depoisoned-phaseB-streaming-torchscript-fp16

VisualEars PhaseB Streaming TorchScript FP16

TorchScript FP16 streaming split for Reza2kn/visualears-fastconformer-fa-depoisoned-phaseB.

Files:

• streaming_encoder_ctc_fp16_streaming_ts.pt
• rnnt_predictor_fp16_streaming_ts.pt
• rnnt_joint_fp16_streaming_ts.pt
• streaming_native_fp16_manifest.json

Streaming contract:

• feature chunk: 1 x 80 x 32
• chunk/shift config: chunk_size=4, shift_size=3, left_chunks=64
• feature shift window: [17, 24]
• encoder output per chunk: 1 x 512 x 2
• channel cache: 17 x 1 x 256 x 512
• time cache: 17 x 1 x 512 x 4
• heads: RNNT and CTC

Validation: runtime smoke passed for encoder+CTC, RNNT predictor, and RNNT joint with live streaming cache inputs/outputs.