Update README.md
Browse files
README.md
CHANGED
|
@@ -17,9 +17,7 @@ Made possible by Tsolfils brilliant node https://github.com/tsolful/ComfyUI-INT8
|
|
| 17 |
I spent a while going down a rabbit hole splicing int4/int8/bf16 combos in a modified version of comfy-quants, trying to figure out what works before finding Tsolfils node. I’m pleased I was heading in the right direction but his node is probably much better than what I was going to do.
|
| 18 |
|
| 19 |
Testing has been unscientific and done in fits and bursts so keep that caveat in mind.
|
| 20 |
-
|
| 21 |
It seems any amount of true w4a4 tensors on their own will taint ltx/eros/sulphur. You get swimming textures on any length/size generation. So that's a no go. But packing them in int8mm seems to fix that problem.
|
| 22 |
-
|
| 23 |
At first I thought the hard limit was a 50/50 combo of int8 and w4a4 int8 matmul, but once I stopped letting true w4a4 filter into the mix I was able to make a 20/80 mix and 100% w4a4 int8-matmul with no more texture swimming artifacts or vastly diminished at any rate.
|
| 24 |
|
| 25 |
Quality - 10Eros_v1.4_DMD_TFO_Int8-Int8mm_Q.safetensor sits below fp8 in my opinion but that's subjective and I have no LPIPS data to back that up.
|
|
|
|
| 17 |
I spent a while going down a rabbit hole splicing int4/int8/bf16 combos in a modified version of comfy-quants, trying to figure out what works before finding Tsolfils node. I’m pleased I was heading in the right direction but his node is probably much better than what I was going to do.
|
| 18 |
|
| 19 |
Testing has been unscientific and done in fits and bursts so keep that caveat in mind.
|
|
|
|
| 20 |
It seems any amount of true w4a4 tensors on their own will taint ltx/eros/sulphur. You get swimming textures on any length/size generation. So that's a no go. But packing them in int8mm seems to fix that problem.
|
|
|
|
| 21 |
At first I thought the hard limit was a 50/50 combo of int8 and w4a4 int8 matmul, but once I stopped letting true w4a4 filter into the mix I was able to make a 20/80 mix and 100% w4a4 int8-matmul with no more texture swimming artifacts or vastly diminished at any rate.
|
| 22 |
|
| 23 |
Quality - 10Eros_v1.4_DMD_TFO_Int8-Int8mm_Q.safetensor sits below fp8 in my opinion but that's subjective and I have no LPIPS data to back that up.
|