Update README.md
Browse files
README.md
CHANGED
|
@@ -54,6 +54,7 @@ Using [llama.cpp quantize cae9fb4](https://github.com/ggerganov/llama.cpp/commit
|
|
| 54 |
- Bravo IQ1_S worse than Alpha?
|
| 55 |
- [Latent loss](https://huggingface.co/Eviation/flux-imatrix/blob/main/images/latent_loss.png)
|
| 56 |
- [Per layer quantization cost](https://huggingface.co/Eviation/flux-imatrix/blob/main/images/casting_cost.png) from [chrisgoringe/casting_cost](https://github.com/chrisgoringe/mixed-gguf-converter/blob/main/costs/casting_cost.yaml)
|
|
|
|
| 57 |
|
| 58 |
### Per weight type quantization loss in increasing order:
|
| 59 |
- double_blocks.%d.txt_attn.proj.weight
|
|
|
|
| 54 |
- Bravo IQ1_S worse than Alpha?
|
| 55 |
- [Latent loss](https://huggingface.co/Eviation/flux-imatrix/blob/main/images/latent_loss.png)
|
| 56 |
- [Per layer quantization cost](https://huggingface.co/Eviation/flux-imatrix/blob/main/images/casting_cost.png) from [chrisgoringe/casting_cost](https://github.com/chrisgoringe/mixed-gguf-converter/blob/main/costs/casting_cost.yaml)
|
| 57 |
+
- [Ablation latent loss per weight type](https://huggingface.co/Eviation/flux-imatrix/blob/main/images/latent_loss_ablation.png)
|
| 58 |
|
| 59 |
### Per weight type quantization loss in increasing order:
|
| 60 |
- double_blocks.%d.txt_attn.proj.weight
|