Hugging Face's logo

fireworks-ai
/
FLUX.1-schnell-fp8-flumina

Safetensors
Model card Files Community
Lucas Nestler commited on
Commit
c541d34
·
unverified ·
1 Parent(s): d45a331

add h100

Browse files
Files changed (1) hide show
  1. README.md +8 -0
README.md CHANGED
@@ -11,26 +11,34 @@ Note:
11
  - The "compile blocks & extras" refers to the option within this repo setting the config values `"compile_blocks" true` & `"compile_extras": true`. ❌ means both were set to false, ✅ means both were set to true.
12
  - All generations which including a ❌ or ✅ are using this repo.
13
 
 
14
  | Resolution | Device | Test | Average it/s |
15
  | ---------- | ---------- | -------------------------- | ------------ |
16
  | 1024x1024 | RTX4090 | bfl codebase fp8 wo quant | 1.7 |
17
  | 1024x1024 | RTX4090 | ❌ compile blocks & extras | 2.55 |
18
  | 1024x1024 | RTX4090 | ✅ compile blocks & extras | 3.51 |
 
19
  | 1024x1024 | RTX6000ADA | bfl codebase | 1.74 |
20
  | 1024x1024 | RTX6000ADA | ❌ compile blocks & extras | 2.08 |
21
  | 1024x1024 | RTX6000ADA | ✅ compile blocks & extras | 2.8 |
 
 
22
  | 768x768 | RTX4090 | bfl codebase fp8 wo quant | 2.32 |
23
  | 768x768 | RTX4090 | ❌ compile blocks & extras | 4.47 |
24
  | 768x768 | RTX4090 | ✅ compile blocks & extras | 6.2 |
25
  | 768x768 | RTX6000ADA | bfl codebase | 3.01 |
26
  | 768x768 | RTX6000ADA | ❌ compile blocks & extras | 3.43 |
27
  | 768x768 | RTX6000ADA | ✅ compile blocks & extras | 4.46 |
 
 
28
  | 1024x720 | RTX4090 | bfl codebase fp8 wo quant | 3.01 |
29
  | 1024x720 | RTX4090 | ❌ compile blocks & extras | 3.6 |
30
  | 1024x720 | RTX4090 | ✅ compile blocks & extras | 4.96 |
31
  | 1024x720 | RTX6000ADA | bfl codebase | 2.37 |
32
  | 1024x720 | RTX6000ADA | ❌ compile blocks & extras | 2.87 |
33
  | 1024x720 | RTX6000ADA | ✅ compile blocks & extras | 3.78 |
 
 
34
 
35
  ## Table of Contents
36
 
 
11
  - The "compile blocks & extras" refers to the option within this repo setting the config values `"compile_blocks" true` & `"compile_extras": true`. ❌ means both were set to false, ✅ means both were set to true.
12
  - All generations which including a ❌ or ✅ are using this repo.
13
 
14
+
15
  | Resolution | Device | Test | Average it/s |
16
  | ---------- | ---------- | -------------------------- | ------------ |
17
  | 1024x1024 | RTX4090 | bfl codebase fp8 wo quant | 1.7 |
18
  | 1024x1024 | RTX4090 | ❌ compile blocks & extras | 2.55 |
19
  | 1024x1024 | RTX4090 | ✅ compile blocks & extras | 3.51 |
20
+ | 1024x1024 | RTX4090 | ✅ compile blocks & extras | 3.51 |
21
  | 1024x1024 | RTX6000ADA | bfl codebase | 1.74 |
22
  | 1024x1024 | RTX6000ADA | ❌ compile blocks & extras | 2.08 |
23
  | 1024x1024 | RTX6000ADA | ✅ compile blocks & extras | 2.8 |
24
+ | 1024x1024 | H100 | ❌ compile blocks & extras | 6.1 |
25
+ | 1024x1024 | H100 | ✅ compile blocks & extras | 11.5 |
26
  | 768x768 | RTX4090 | bfl codebase fp8 wo quant | 2.32 |
27
  | 768x768 | RTX4090 | ❌ compile blocks & extras | 4.47 |
28
  | 768x768 | RTX4090 | ✅ compile blocks & extras | 6.2 |
29
  | 768x768 | RTX6000ADA | bfl codebase | 3.01 |
30
  | 768x768 | RTX6000ADA | ❌ compile blocks & extras | 3.43 |
31
  | 768x768 | RTX6000ADA | ✅ compile blocks & extras | 4.46 |
32
+ | 768x768 | H100 | ❌ compile blocks & extras | 10.3 |
33
+ | 768x768 | H100 | ✅ compile blocks & extras | 20.8 |
34
  | 1024x720 | RTX4090 | bfl codebase fp8 wo quant | 3.01 |
35
  | 1024x720 | RTX4090 | ❌ compile blocks & extras | 3.6 |
36
  | 1024x720 | RTX4090 | ✅ compile blocks & extras | 4.96 |
37
  | 1024x720 | RTX6000ADA | bfl codebase | 2.37 |
38
  | 1024x720 | RTX6000ADA | ❌ compile blocks & extras | 2.87 |
39
  | 1024x720 | RTX6000ADA | ✅ compile blocks & extras | 3.78 |
40
+ | 1024x720 | H100 | ❌ compile blocks & extras | 8.2 |
41
+ | 1024x720 | H100 | ✅ compile blocks & extras | 15.7 |
42
 
43
  ## Table of Contents
44