Spaces:
Running
Running
Jean Louis
commited on
Commit
·
988a4fc
1
Parent(s):
50993ec
Updated HTML files
Browse files- index.html +16 -0
index.html
CHANGED
@@ -48,6 +48,22 @@ what makes <em>you</em> tick. Let’s dive in together! 🎤💻✨</p>
|
|
48 |
|
49 |
<h4>Install NVIDIA Canary-1B-Flash fully free software Large Language Model (LLM) for speech recognition</h4>
|
50 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
51 |
<h4>Run NVIDIA Canary-1B-Flash as server</h4>
|
52 |
|
53 |
<h4>Prepare Shell scripts</h4>
|
|
|
48 |
|
49 |
<h4>Install NVIDIA Canary-1B-Flash fully free software Large Language Model (LLM) for speech recognition</h4>
|
50 |
|
51 |
+
<p>The Canary-1B-Flash model is a cutting-edge multilingual multi-tasking
|
52 |
+
model based on the Canary architecture, designed to achieve
|
53 |
+
state-of-the-art performance in various speech benchmarks. It has 883
|
54 |
+
million parameters and delivers high inference speeds, exceeding 1000
|
55 |
+
RTFx on the OpenASR Leaderboard datasets. Canary-1B-Flash supports
|
56 |
+
automatic speech-to-text recognition (ASR) in English, German, French,
|
57 |
+
and Spanish. Additionally, it facilitates translation between these
|
58 |
+
languages, with options for output with or without punctuation and
|
59 |
+
capitalization. The model includes experimental features for
|
60 |
+
generating word-level and segment-level timestamps, making it
|
61 |
+
versatile for applications requiring precise temporal
|
62 |
+
information. Canary-1B-Flash operates using a FastConformer Encoder
|
63 |
+
and a Transformer Decoder, combined with a concatenated tokenizer that
|
64 |
+
leverages SentencePiece for scalability across languages. This model
|
65 |
+
is available under the CC-BY-4.0 license.</p>
|
66 |
+
|
67 |
<h4>Run NVIDIA Canary-1B-Flash as server</h4>
|
68 |
|
69 |
<h4>Prepare Shell scripts</h4>
|