Pomni commited on
Commit
8ef3146
·
verified ·
1 Parent(s): aafa238

Complete quant list with correct sizes

Browse files
Files changed (1) hide show
  1. README.md +12 -12
README.md CHANGED
@@ -22,18 +22,18 @@ If you are looking for a program to run this model with, then I would recommend
22
  Not all of the links will work right now, as I am still in the middle of quantizing this model. Clicking on a link will download the corresponding quant instantly.
23
  | Link | Quant | Size | Notes
24
  |:-----|:-----|--------:|:------|
25
- | [GGML](https://huggingface.co/Pomni/distil-tiny.en-ggml-allquants/resolve/main/ggml-tiny.en-f32.bin) | F32 | ? MB | Likely overkill. |
26
- | [GGML](https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml-tiny.en.bin) | F16 | 77.7 MB | Performs better than Q8_0 for noisy audio and music. |
27
- | [GGML](https://huggingface.co/Pomni/whisper-tiny.en-ggml-allquants/resolve/main/ggml-tiny.en-q8_0.bin) | Q8_0 | ? MB | Sweet spot; superficial quality loss at nearly double the speed. |
28
- | [GGML](https://huggingface.co/Pomni/whisper-tiny.en-ggml-allquants/resolve/main/ggml-tiny.en-q6_k.bin) | Q6_K | ? MB | |
29
- | [GGML](https://huggingface.co/Pomni/whisper-tiny.en-ggml-allquants/resolve/main/ggml-tiny.en-q5_k.bin) | Q5_K | ? MB | |
30
- | [GGML](https://huggingface.co/Pomni/whisper-tiny.en-ggml-allquants/resolve/main/ggml-tiny.en-q5_1.bin) | Q5_1 | ? MB | |
31
- | [GGML](https://huggingface.co/Pomni/whisper-tiny.en-ggml-allquants/resolve/main/ggml-tiny.en-q5_0.bin) | Q5_0 | ? MB | Last "good" quant; anything below loses quality rapidly. |
32
- | [GGML](https://huggingface.co/Pomni/whisper-tiny.en-ggml-allquants/resolve/main/ggml-tiny.en-q4_k.bin) | Q4_K | ? MB | *Might* not have lost too much quality, but I'm not sure. |
33
- | [GGML](https://huggingface.co/Pomni/whisper-tiny.en-ggml-allquants/resolve/main/ggml-tiny.en-q4_1.bin) | Q4_1 | ? MB | |
34
- | [GGML](https://huggingface.co/Pomni/whisper-tiny.en-ggml-allquants/resolve/main/ggml-tiny.en-q4_0.bin) | Q4_0 | ? MB | |
35
- | [GGML](https://huggingface.co/Pomni/whisper-tiny.en-ggml-allquants/resolve/main/ggml-tiny.en-q3_k.bin) | Q3_K | ? MB | |
36
- | [GGML](https://huggingface.co/Pomni/whisper-tiny.en-ggml-allquants/resolve/main/ggml-tiny.en-q2_k.bin) | Q2_K | ? MB | Completely non-sensical outputs. |
37
 
38
  The F16 quant was taken from [ggerganov/whisper.cpp/ggml-tiny.en.bin](https://huggingface.co/ggerganov/whisper.cpp/blob/main/ggml-tiny.en.bin).
39
  ## Questions you may have
 
22
  Not all of the links will work right now, as I am still in the middle of quantizing this model. Clicking on a link will download the corresponding quant instantly.
23
  | Link | Quant | Size | Notes
24
  |:-----|:-----|--------:|:------|
25
+ | [GGML](https://huggingface.co/Pomni/whisper-tiny.en-ggml-allquants/resolve/main/ggml-tiny.en-f32.bin) | F32 | 152 MB | Likely overkill. |
26
+ | [GGML](https://huggingface.co/Pomni/whisper-tiny.en-ggml-allquants/resolve/main/ggml-tiny.en-f16.bin) | F16 | 77.7 MB | Performs better than Q8_0 for noisy audio and music. |
27
+ | [GGML](https://huggingface.co/Pomni/whisper-tiny.en-ggml-allquants/resolve/main/ggml-tiny.en-q8_0.bin) | Q8_0 | 43.6 MB | Sweet spot; superficial quality loss at nearly double the speed. |
28
+ | [GGML](https://huggingface.co/Pomni/whisper-tiny.en-ggml-allquants/resolve/main/ggml-tiny.en-q6_k.bin) | Q6_K | 34.7 MB | |
29
+ | [GGML](https://huggingface.co/Pomni/whisper-tiny.en-ggml-allquants/resolve/main/ggml-tiny.en-q5_k.bin) | Q5_K | 29.9 MB | |
30
+ | [GGML](https://huggingface.co/Pomni/whisper-tiny.en-ggml-allquants/resolve/main/ggml-tiny.en-q5_1.bin) | Q5_1 | 32.2 MB | |
31
+ | [GGML](https://huggingface.co/Pomni/whisper-tiny.en-ggml-allquants/resolve/main/ggml-tiny.en-q5_0.bin) | Q5_0 | 29.9 MB | Last "good" quant; anything below loses quality rapidly. |
32
+ | [GGML](https://huggingface.co/Pomni/whisper-tiny.en-ggml-allquants/resolve/main/ggml-tiny.en-q4_k.bin) | Q4_K | 25.3 MB | *Might* not have lost too much quality, but I'm not sure. |
33
+ | [GGML](https://huggingface.co/Pomni/whisper-tiny.en-ggml-allquants/resolve/main/ggml-tiny.en-q4_1.bin) | Q4_1 | 27.6 MB | |
34
+ | [GGML](https://huggingface.co/Pomni/whisper-tiny.en-ggml-allquants/resolve/main/ggml-tiny.en-q4_0.bin) | Q4_0 | 25.3 MB | |
35
+ | [GGML](https://huggingface.co/Pomni/whisper-tiny.en-ggml-allquants/resolve/main/ggml-tiny.en-q3_k.bin) | Q3_K | 20.5 MB | |
36
+ | [GGML](https://huggingface.co/Pomni/whisper-tiny.en-ggml-allquants/resolve/main/ggml-tiny.en-q2_k.bin) | Q2_K | 16.8 MB | Completely non-sensical outputs. |
37
 
38
  The F16 quant was taken from [ggerganov/whisper.cpp/ggml-tiny.en.bin](https://huggingface.co/ggerganov/whisper.cpp/blob/main/ggml-tiny.en.bin).
39
  ## Questions you may have