Update README.md
Browse files
README.md
CHANGED
|
@@ -42,25 +42,12 @@ It does not generate audio; it produces text based on audio input.
|
|
| 42 |
# 📦 What This Quantized Version Enables
|
| 43 |
|
| 44 |
This NVFP4 quantized version reduces memory requirements significantly:
|
| 45 |
-
|
| 46 |
Size: ~22 GB (down from ~67 GB)
|
| 47 |
-
|
| 48 |
Should fit comfortably on a single RTX 5090
|
| 49 |
|
| 50 |
-
Fully compatible with vLLM (including streaming text output)
|
| 51 |
-
|
| 52 |
Preserves most reasoning performance from the BF16 release
|
| 53 |
-
|
| 54 |
Because of this, anyone with a high-end consumer GPU can experiment with advanced audio reasoning locally.
|
| 55 |
|
| 56 |
-
🖥 Supported Audio Behavior
|
| 57 |
-
|
| 58 |
-
The model supports:
|
| 59 |
-
|
| 60 |
-
✔ Streaming text output through vLLM
|
| 61 |
-
✔ Reading uploaded audio files (WAV/MP3/etc) via ffmpeg
|
| 62 |
-
✘ It does not synthesize audio
|
| 63 |
-
✘ It does not require pre-burned waveforms — any user-provided audio file works
|
| 64 |
|
| 65 |
Check the original model card for more information about this model.
|
| 66 |
|
|
|
|
| 42 |
# 📦 What This Quantized Version Enables
|
| 43 |
|
| 44 |
This NVFP4 quantized version reduces memory requirements significantly:
|
|
|
|
| 45 |
Size: ~22 GB (down from ~67 GB)
|
|
|
|
| 46 |
Should fit comfortably on a single RTX 5090
|
| 47 |
|
|
|
|
|
|
|
| 48 |
Preserves most reasoning performance from the BF16 release
|
|
|
|
| 49 |
Because of this, anyone with a high-end consumer GPU can experiment with advanced audio reasoning locally.
|
| 50 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 51 |
|
| 52 |
Check the original model card for more information about this model.
|
| 53 |
|