Pomni commited on
Commit
aba251d
·
verified ·
1 Parent(s): f42ee36

Update quant list with correct sizes

Browse files
Files changed (1) hide show
  1. README.md +10 -10
README.md CHANGED
@@ -32,16 +32,16 @@ Clicking on a link will download the corresponding quant instantly.
32
  |:-----|:-----|--------:|:------|
33
  | [GGML](https://huggingface.co/Pomni/distil-small.en-ggml-allquants/resolve/main/ggml-distil-small.en-f32.bin) | F32 | 665 MB | Likely overkill. |
34
  | [GGML](https://huggingface.co/Pomni/distil-small.en-ggml-allquants/resolve/main/ggml-distil-small.en-f16.bin) | F16 | 336 MB | Performs better than Q8_0 for noisy audio and music. |
35
- | [GGML](https://huggingface.co/Pomni/distil-small.en-ggml-allquants/resolve/main/ggml-distil-small.en-q8_0.bin) | Q8_0 | 188 MB | Sweet spot; superficial quality loss at nearly double the speed. |
36
- | [GGML](https://huggingface.co/Pomni/distil-small.en-ggml-allquants/resolve/main/ggml-distil-small.en-q6_k.bin) | Q6_K | 148 MB | |
37
- | [GGML](https://huggingface.co/Pomni/distil-small.en-ggml-allquants/resolve/main/ggml-distil-small.en-q5_k.bin) | Q5_K | 127 MB | |
38
- | [GGML](https://huggingface.co/Pomni/distil-small.en-ggml-allquants/resolve/main/ggml-distil-small.en-q5_1.bin) | Q5_1 | 137 MB | |
39
- | [GGML](https://huggingface.co/Pomni/distil-small.en-ggml-allquants/resolve/main/ggml-distil-small.en-q5_0.bin) | Q5_0 | 127 MB | Last "good" quant; anything below loses quality rapidly. |
40
- | [GGML](https://huggingface.co/Pomni/distil-small.en-ggml-allquants/resolve/main/ggml-distil-small.en-q4_k.bin) | Q4_K | 106 MB | *Might* not have lost too much quality, but I'm not sure. |
41
- | [GGML](https://huggingface.co/Pomni/distil-small.en-ggml-allquants/resolve/main/ggml-distil-small.en-q4_1.bin) | Q4_1 | 117 MB | |
42
- | [GGML](https://huggingface.co/Pomni/distil-small.en-ggml-allquants/resolve/main/ggml-distil-small.en-q4_0.bin) | Q4_0 | 106 MB | |
43
- | [GGML](https://huggingface.co/Pomni/distil-small.en-ggml-allquants/resolve/main/ggml-distil-small.en-q3_k.bin) | Q3_K | 84.9 MB | |
44
- | [GGML](https://huggingface.co/Pomni/distil-small.en-ggml-allquants/resolve/main/ggml-distil-small.en-q2_k.bin) | Q2_K | 68.4 MB | Completely non-sensical outputs. |
45
 
46
  The F32 quant was taken from [distil-whisper/distil-small.en/ggml-distil-small.en.fp32.bin](https://huggingface.co/distil-whisper/distil-small.en/blob/main/ggml-distil-small.en.fp32.bin), and the F16 quant was taken from [distil-whisper/distil-small.en/ggml-distil-small.en.bin](https://huggingface.co/distil-whisper/distil-small.en/blob/main/ggml-distil-small.en.bin).
47
  ## Questions you may have
 
32
  |:-----|:-----|--------:|:------|
33
  | [GGML](https://huggingface.co/Pomni/distil-small.en-ggml-allquants/resolve/main/ggml-distil-small.en-f32.bin) | F32 | 665 MB | Likely overkill. |
34
  | [GGML](https://huggingface.co/Pomni/distil-small.en-ggml-allquants/resolve/main/ggml-distil-small.en-f16.bin) | F16 | 336 MB | Performs better than Q8_0 for noisy audio and music. |
35
+ | [GGML](https://huggingface.co/Pomni/distil-small.en-ggml-allquants/resolve/main/ggml-distil-small.en-q8_0.bin) | Q8_0 | 184 MB | Sweet spot; superficial quality loss at nearly double the speed. |
36
+ | [GGML](https://huggingface.co/Pomni/distil-small.en-ggml-allquants/resolve/main/ggml-distil-small.en-q6_k.bin) | Q6_K | 144 MB | |
37
+ | [GGML](https://huggingface.co/Pomni/distil-small.en-ggml-allquants/resolve/main/ggml-distil-small.en-q5_k.bin) | Q5_K | 123 MB | |
38
+ | [GGML](https://huggingface.co/Pomni/distil-small.en-ggml-allquants/resolve/main/ggml-distil-small.en-q5_1.bin) | Q5_1 | 133 MB | |
39
+ | [GGML](https://huggingface.co/Pomni/distil-small.en-ggml-allquants/resolve/main/ggml-distil-small.en-q5_0.bin) | Q5_0 | 123 MB | Last "good" quant; anything below loses quality rapidly. |
40
+ | [GGML](https://huggingface.co/Pomni/distil-small.en-ggml-allquants/resolve/main/ggml-distil-small.en-q4_k.bin) | Q4_K | 103 MB | *Might* not have lost too much quality, but I'm not sure. |
41
+ | [GGML](https://huggingface.co/Pomni/distil-small.en-ggml-allquants/resolve/main/ggml-distil-small.en-q4_1.bin) | Q4_1 | 113 MB | |
42
+ | [GGML](https://huggingface.co/Pomni/distil-small.en-ggml-allquants/resolve/main/ggml-distil-small.en-q4_0.bin) | Q4_0 | 103 MB | |
43
+ | [GGML](https://huggingface.co/Pomni/distil-small.en-ggml-allquants/resolve/main/ggml-distil-small.en-q3_k.bin) | Q3_K | 81 MB | |
44
+ | [GGML](https://huggingface.co/Pomni/distil-small.en-ggml-allquants/resolve/main/ggml-distil-small.en-q2_k.bin) | Q2_K | 64.5 MB | Completely non-sensical outputs. |
45
 
46
  The F32 quant was taken from [distil-whisper/distil-small.en/ggml-distil-small.en.fp32.bin](https://huggingface.co/distil-whisper/distil-small.en/blob/main/ggml-distil-small.en.fp32.bin), and the F16 quant was taken from [distil-whisper/distil-small.en/ggml-distil-small.en.bin](https://huggingface.co/distil-whisper/distil-small.en/blob/main/ggml-distil-small.en.bin).
47
  ## Questions you may have