ModelVerifiedExternal

Timestamps Asr Canary Ctc Mlx Q8

Quantized MLX port of the timestampsasrmodel bundled inside nvidia/canary-1b-v2. It is the dedicated CTC forced-alignment head paired with canary; vocabulary is byte-identical to canary's sentencepiece (16 384 pieces).

Sourcehuggingface.co/TechHara/timestamps-asr-canary-ctc-mlx-q8/tree/07ab32c41100e1aab092a904713d13c885c92f95

KindOtherBase modelnvidia/canary-1b-v2VersionRevision 07ab32cPublisher@TechHaraCgrade

Model source

Kind: Other
File size: 702 MB
Base model: nvidia/canary-1b-v2
Version: vRevision 07ab32c
Updated: Oct 2, 2026
Publisher: @TechHara
Downloads: 32
Parameters: 1B
Source: Hugging Face

Source description

canary-1b-v2 — timestamps_asr_model — MLX q8

Quantized MLX port of the timestamps_asr_model bundled inside nvidia/canary-1b-v2. It is the dedicated CTC forced-alignment head paired with canary; vocabulary is byte-identical to canary's sentencepiece (16 384 pieces).


Architecture	24-layer Conformer encoder + 1×1 CTC head
Params	626 M
Mel features	128
d_model	1024
CTC vocab	16 384 + blank
Subsampling	8× (dw_striding)
Self-attention	full rel-pos
Quantization	bits=8, group_size=64 (MLX default)
Source	extracted from `canary-1b-v2.nemo`

Files

File	Size
`config.json`	preprocessor + encoder + CTC config (with inline vocab)
`model.safetensors`	q8 weights + bf16 scales/biases
`tokenizer.model`	canary's sentencepiece (1:1 copy)

License

CC-BY-4.0 — inherited from nvidia/canary-1b-v2.

Sources

1 source

Hugging FacePrimaryVersion 07ab32c41100e1aab092a904713d13c885c92f95

huggingface.co/TechHara/timestamps-asr-canary-ctc-mlx-q8/tree/07ab32c41100e1aab092a904713d13c885c92f95

Verified Oct 2

Model artifacts

1 artifact

model.safetensors

safetensors · 702 MB · SHA-256 07adc9f5f828…59b9 · Hugging Face

Download

Source excerpts

3 excerpts

ReadmeJul 11huggingface.co/TechHara/timestamps-asr-canary-ctc-mlx-q8

--- language: en license: cc-by-4.0 library_name: mlx tags: - mlx - asr - forced-alignment - ctc - quantized - q8 base_model: nvidia/canary-1b-v2 --- # canary-1b-v2 — timestamps_asr_model — MLX q8 Quantized MLX port of the `timestamps_asr_model` bundled inside [nvidia/canary-1b-v2](https://huggingface.co/nvidia/canary-1b-v2). It is the dedicated CTC forced-alignment head paired with canary; vocabulary is byte-identical to canary's sentencepiece (16 384 pieces). | | | |---|---| | Architecture | 24-layer Conformer encoder + 1×1 CTC head | | Params | 626 M | | Mel features | 128 | | d_model | 1024 | | CTC vocab | 16 384 + blank | | Subsampling | 8× (dw_striding) | | Self-attention | full rel-pos | | Quantization | bits=8, group_size=64 (MLX default) | | Source | extracted from `canary-1b-v2.nemo` | ## Files | File | Size | |---|---| | `config.json` | preprocessor + encoder + CTC config (with inline vocab) | | `model.safetensors` | q8 weights + bf16 scales/biases | | `tokenizer.model` | canary's sentencepiece (1:1 copy) | ## License CC-BY-4.0 — inherited from `nvidia/canary-1b-v2`.

NoteOct 2huggingface.co/TechHara/timestamps-asr-canary-ctc-mlx-q8

Source context: 32 downloads · 0 likes · Library mlx · Repo TechHara/timestamps-asr-canary-ctc-mlx-q8

NoteJul 17huggingface.co/TechHara/timestamps-asr-canary-ctc-mlx-q8

Source context: 36 downloads · 0 likes · Library mlx · Repo TechHara/timestamps-asr-canary-ctc-mlx-q8