Card: speed on the demo Space (source 86b6e7cbb263)
Browse files
README.md
CHANGED
|
@@ -76,6 +76,8 @@ print(stt.transcribe("μμ±.wav")) # νμΌ κ²½λ‘, λλ stt.transcribe(λ°°
|
|
| 76 |
|
| 77 |
Apple M5 Proμμ 60λΆ μμ±μ Core AIλ‘ 5.50μ΄μ λ°μμΌμ΅λλ€(μ€μκ°μ 654λ°°, μ΅λ λ©λͺ¨λ¦¬ 3.4GB). κ°μ λ§₯μμ PyTorch MPSλ 21.7μ΄μμ΅λλ€. μ΄ κ°μ ꡬ쑰μ μΆλ ₯ ν¬κΈ°κ° κ°μ μ΄μ κ°μ€μΉλ‘ μ° κ°μ
λλ€.
|
| 78 |
|
|
|
|
|
|
|
| 79 |
## νκ³
|
| 80 |
|
| 81 |
* **λλ
체λ‘λ§ νκ°νμ΅λλ€.** λν, μ‘μ, μ¬λ¬ μ¬λμ΄ ν¨κ» λ§νλ μμ±μ νμ§μ μ¬μ§ μμμ΅λλ€.
|
|
@@ -113,7 +115,7 @@ https://huggingface.co/Haverbex/eg2-ko-reader/discussions/new
|
|
| 113 |
* **What it is**: a Korean syllable CTC reader (6-layer Transformer, 19.7M parameters) that reads the output of the frozen EmbeddingGemma 2 audio encoder (25 tokens per second, 512 numbers each). No autoregressive decoder and no language model.
|
| 114 |
* **Use**: `hf download Haverbex/eg2-ko-reader --local-dir eg2-ko-reader`, `pip install -r eg2-ko-reader/requirements.txt`, `python eg2-ko-reader/eg2_ko_reader.py audio.wav`. The download is about 1.6 GB and includes the encoder.
|
| 115 |
* **Accuracy** (character error rate without spaces, greedy): Zeroth-Korean test 7.26 %, FLEURS ko test 5.59 %. Read speech only; one training seed; no confidence interval for the test sets.
|
| 116 |
-
* **Speed**: 60 minutes of audio in 5.50 s with Core AI on an Apple M5 Pro (measured with earlier weights of the same architecture).
|
| 117 |
* **Limits**: Hangul syllables and spaces only (no punctuation, digits or Latin letters); not evaluated on conversation or noise; no streaming; no speaker labels.
|
| 118 |
* **Training data**: Zeroth-Korean and FLEURS ko (CC BY 4.0), and about 460 hours of YODAS2 Korean (YouTube CC BY video, CC BY 3.0) with machine-made transcripts. Not legally reviewed.
|
| 119 |
* **License**: the reader, the code and the Core AI bundle are under the EG2-KO Reader License 1.0: individuals may use them for any purpose; organizations may use them for non-commercial purposes and must ask for written permission before commercial use. This is source-available, not an OSI open-source license. `encoder/` is Apache-2.0.
|
|
|
|
| 76 |
|
| 77 |
Apple M5 Proμμ 60λΆ μμ±μ Core AIλ‘ 5.50μ΄μ λ°μμΌμ΅λλ€(μ€μκ°μ 654λ°°, μ΅λ λ©λͺ¨λ¦¬ 3.4GB). κ°μ λ§₯μμ PyTorch MPSλ 21.7μ΄μμ΅λλ€. μ΄ κ°μ ꡬ쑰μ μΆλ ₯ ν¬κΈ°κ° κ°μ μ΄μ κ°μ€μΉλ‘ μ° κ°μ
λλ€.
|
| 78 |
|
| 79 |
+
λ°λͺ¨ Space(Hugging Face λ¬΄λ£ CPU, `cpu-basic`)μμλ 112.1μ΄ μμ±μ 21.2μ΄μ, 4.7μ΄ μμ±μ 0.8μ΄μ λ°μμΌμ΅λλ€(μ€μκ°μ μ½ 5λ°°, 2026-11-17, ν λ²μ© μ° κ°).
|
| 80 |
+
|
| 81 |
## νκ³
|
| 82 |
|
| 83 |
* **λλ
체λ‘λ§ νκ°νμ΅λλ€.** λν, μ‘μ, μ¬λ¬ μ¬λμ΄ ν¨κ» λ§νλ μμ±μ νμ§μ μ¬μ§ μμμ΅λλ€.
|
|
|
|
| 115 |
* **What it is**: a Korean syllable CTC reader (6-layer Transformer, 19.7M parameters) that reads the output of the frozen EmbeddingGemma 2 audio encoder (25 tokens per second, 512 numbers each). No autoregressive decoder and no language model.
|
| 116 |
* **Use**: `hf download Haverbex/eg2-ko-reader --local-dir eg2-ko-reader`, `pip install -r eg2-ko-reader/requirements.txt`, `python eg2-ko-reader/eg2_ko_reader.py audio.wav`. The download is about 1.6 GB and includes the encoder.
|
| 117 |
* **Accuracy** (character error rate without spaces, greedy): Zeroth-Korean test 7.26 %, FLEURS ko test 5.59 %. Read speech only; one training seed; no confidence interval for the test sets.
|
| 118 |
+
* **Speed**: 60 minutes of audio in 5.50 s with Core AI on an Apple M5 Pro (measured with earlier weights of the same architecture). On the free-CPU demo Space, 112.1 s of audio took 21.2 s (about 5x realtime, measured once, 2026-11-17).
|
| 119 |
* **Limits**: Hangul syllables and spaces only (no punctuation, digits or Latin letters); not evaluated on conversation or noise; no streaming; no speaker labels.
|
| 120 |
* **Training data**: Zeroth-Korean and FLEURS ko (CC BY 4.0), and about 460 hours of YODAS2 Korean (YouTube CC BY video, CC BY 3.0) with machine-made transcripts. Not legally reviewed.
|
| 121 |
* **License**: the reader, the code and the Core AI bundle are under the EG2-KO Reader License 1.0: individuals may use them for any purpose; organizations may use them for non-commercial purposes and must ask for written permission before commercial use. This is source-available, not an OSI open-source license. `encoder/` is Apache-2.0.
|