Hugging Face's logo

amd
/
sdxl-base-amdnpu

Text-to-Image
ONNX
Safetensors
English
stable-diffusion
sdxl
Model card Files
xet
 Community
1

1.8.0

#1
by huisu - opened
base: refs/heads/main
←
from: refs/pr/1
Discussion Files changed
+179415
-178598
Files changed (16) hide show
  1. .gitattributes +4 -0
  2. unet/config.json +69 -69
  3. unet/dd/cache/NhwcConv_0-unetconv_inConv_meta.json +0 -0
  4. unet/dd/{dd_metastate_SD15_Unet_NhwcConv_0-unetconv_inConv.super → dd_metastate_SD15_Unet_NhwcConv_0-unetconv_inConv.ddbundle} +2 -2
  5. unet/dd/dd_metastate_SD15_Unet_NhwcConv_0-unetconv_inConv.fconst +2 -2
  6. unet/dd/dd_metastate_SD15_Unet_NhwcConv_0-unetconv_inConv.state +2 -2
  7. unet/dd/onnx_report.txt +0 -76
  8. unet/dd/replaced.onnx +2 -2
  9. vae_decoder/config.json +31 -31
  10. vae_decoder/dd/cache/NhwcConv_0-post_quant_convConv_meta.json +0 -0
  11. vae_decoder/dd/dd_metastate_Sd15_Decoder_NhwcConv_0-post_quant_convConv.ctrlpkt +0 -3
  12. unet/dd/dd_metastate_SD15_Unet_NhwcConv_0-unetconv_inConv.ctrlpkt → vae_decoder/dd/dd_metastate_Sd15_Decoder_NhwcConv_0-post_quant_convConv.ddbundle +2 -2
  13. vae_decoder/dd/dd_metastate_Sd15_Decoder_NhwcConv_0-post_quant_convConv.state +2 -2
  14. vae_decoder/dd/dd_metastate_Sd15_Decoder_NhwcConv_0-post_quant_convConv.super +0 -3
  15. vae_decoder/dd/onnx_report.txt +0 -35
  16. vae_decoder/dd/replaced.onnx +2 -2
.gitattributes CHANGED
@@ -41,3 +41,7 @@ vae_decoder/dd/dd_metastate_Sd15_Decoder_NhwcConv_0-post_quant_convConv.ctrlpkt
41
  vae_decoder/dd/dd_metastate_Sd15_Decoder_NhwcConv_0-post_quant_convConv.fconst filter=lfs diff=lfs merge=lfs -text
42
  vae_decoder/dd/dd_metastate_Sd15_Decoder_NhwcConv_0-post_quant_convConv.state filter=lfs diff=lfs merge=lfs -text
43
  vae_decoder/dd/dd_metastate_Sd15_Decoder_NhwcConv_0-post_quant_convConv.super filter=lfs diff=lfs merge=lfs -text
 
 
 
 
 
41
  vae_decoder/dd/dd_metastate_Sd15_Decoder_NhwcConv_0-post_quant_convConv.fconst filter=lfs diff=lfs merge=lfs -text
42
  vae_decoder/dd/dd_metastate_Sd15_Decoder_NhwcConv_0-post_quant_convConv.state filter=lfs diff=lfs merge=lfs -text
43
  vae_decoder/dd/dd_metastate_Sd15_Decoder_NhwcConv_0-post_quant_convConv.super filter=lfs diff=lfs merge=lfs -text
44
+ *.onnx.data filter=lfs diff=lfs merge=lfs -text
45
+ *.state filter=lfs diff=lfs merge=lfs -text
46
+ *.fconst filter=lfs diff=lfs merge=lfs -text
47
+ *.ddbundle filter=lfs diff=lfs merge=lfs -text
unet/config.json CHANGED
@@ -1,69 +1,69 @@
1
- {
2
- "_class_name": "UNet2DConditionModel",
3
- "_diffusers_version": "0.19.0.dev0",
4
- "act_fn": "silu",
5
- "addition_embed_type": "text_time",
6
- "addition_embed_type_num_heads": 64,
7
- "addition_time_embed_dim": 256,
8
- "attention_head_dim": [
9
- 5,
10
- 10,
11
- 20
12
- ],
13
- "block_out_channels": [
14
- 320,
15
- 640,
16
- 1280
17
- ],
18
- "center_input_sample": false,
19
- "class_embed_type": null,
20
- "class_embeddings_concat": false,
21
- "conv_in_kernel": 3,
22
- "conv_out_kernel": 3,
23
- "cross_attention_dim": 2048,
24
- "cross_attention_norm": null,
25
- "down_block_types": [
26
- "DownBlock2D",
27
- "CrossAttnDownBlock2D",
28
- "CrossAttnDownBlock2D"
29
- ],
30
- "downsample_padding": 1,
31
- "dual_cross_attention": false,
32
- "encoder_hid_dim": null,
33
- "encoder_hid_dim_type": null,
34
- "flip_sin_to_cos": true,
35
- "freq_shift": 0,
36
- "in_channels": 4,
37
- "layers_per_block": 2,
38
- "mid_block_only_cross_attention": null,
39
- "mid_block_scale_factor": 1,
40
- "mid_block_type": "UNetMidBlock2DCrossAttn",
41
- "norm_eps": 1e-05,
42
- "norm_num_groups": 32,
43
- "num_attention_heads": null,
44
- "num_class_embeds": null,
45
- "only_cross_attention": false,
46
- "out_channels": 4,
47
- "projection_class_embeddings_input_dim": 2816,
48
- "resnet_out_scale_factor": 1.0,
49
- "resnet_skip_time_act": false,
50
- "resnet_time_scale_shift": "default",
51
- "sample_size": 128,
52
- "time_cond_proj_dim": null,
53
- "time_embedding_act_fn": null,
54
- "time_embedding_dim": null,
55
- "time_embedding_type": "positional",
56
- "timestep_post_act": null,
57
- "transformer_layers_per_block": [
58
- 1,
59
- 2,
60
- 10
61
- ],
62
- "up_block_types": [
63
- "CrossAttnUpBlock2D",
64
- "CrossAttnUpBlock2D",
65
- "UpBlock2D"
66
- ],
67
- "upcast_attention": null,
68
- "use_linear_projection": true
69
- }
 
1
+ {
2
+ "_class_name": "UNet2DConditionModel",
3
+ "_diffusers_version": "0.19.0.dev0",
4
+ "act_fn": "silu",
5
+ "addition_embed_type": "text_time",
6
+ "addition_embed_type_num_heads": 64,
7
+ "addition_time_embed_dim": 256,
8
+ "attention_head_dim": [
9
+ 5,
10
+ 10,
11
+ 20
12
+ ],
13
+ "block_out_channels": [
14
+ 320,
15
+ 640,
16
+ 1280
17
+ ],
18
+ "center_input_sample": false,
19
+ "class_embed_type": null,
20
+ "class_embeddings_concat": false,
21
+ "conv_in_kernel": 3,
22
+ "conv_out_kernel": 3,
23
+ "cross_attention_dim": 2048,
24
+ "cross_attention_norm": null,
25
+ "down_block_types": [
26
+ "DownBlock2D",
27
+ "CrossAttnDownBlock2D",
28
+ "CrossAttnDownBlock2D"
29
+ ],
30
+ "downsample_padding": 1,
31
+ "dual_cross_attention": false,
32
+ "encoder_hid_dim": null,
33
+ "encoder_hid_dim_type": null,
34
+ "flip_sin_to_cos": true,
35
+ "freq_shift": 0,
36
+ "in_channels": 4,
37
+ "layers_per_block": 2,
38
+ "mid_block_only_cross_attention": null,
39
+ "mid_block_scale_factor": 1,
40
+ "mid_block_type": "UNetMidBlock2DCrossAttn",
41
+ "norm_eps": 1e-05,
42
+ "norm_num_groups": 32,
43
+ "num_attention_heads": null,
44
+ "num_class_embeds": null,
45
+ "only_cross_attention": false,
46
+ "out_channels": 4,
47
+ "projection_class_embeddings_input_dim": 2816,
48
+ "resnet_out_scale_factor": 1.0,
49
+ "resnet_skip_time_act": false,
50
+ "resnet_time_scale_shift": "default",
51
+ "sample_size": 128,
52
+ "time_cond_proj_dim": null,
53
+ "time_embedding_act_fn": null,
54
+ "time_embedding_dim": null,
55
+ "time_embedding_type": "positional",
56
+ "timestep_post_act": null,
57
+ "transformer_layers_per_block": [
58
+ 1,
59
+ 2,
60
+ 10
61
+ ],
62
+ "up_block_types": [
63
+ "CrossAttnUpBlock2D",
64
+ "CrossAttnUpBlock2D",
65
+ "UpBlock2D"
66
+ ],
67
+ "upcast_attention": null,
68
+ "use_linear_projection": true
69
+ }
unet/dd/cache/NhwcConv_0-unetconv_inConv_meta.json CHANGED
The diff for this file is too large to render. See raw diff
 
unet/dd/{dd_metastate_SD15_Unet_NhwcConv_0-unetconv_inConv.super → dd_metastate_SD15_Unet_NhwcConv_0-unetconv_inConv.ddbundle} RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:1248c187b5b209a54a2555d999a0a774ce336d9a22fb136550232f079df36dc9
3
- size 1046920
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:88962e2b56f1f691406068788316aa07a8b1de27cb56accc8289957dae7502f1
3
+ size 2343794277
unet/dd/dd_metastate_SD15_Unet_NhwcConv_0-unetconv_inConv.fconst CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:f300ba1ffc2c55b5a3f20f55e370bac288a196607d742f758f9f0116ef736b11
3
- size 2936554816
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:94087ca135d8b548baa9a669d6b64b116e667725ce1ab7c8bda24722c13e0254
3
+ size 2976580160
unet/dd/dd_metastate_SD15_Unet_NhwcConv_0-unetconv_inConv.state CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:1e137a3fa14050ef88348949f3fd1a48596fda8222ca520e13ab9ca2b39ff6bd
3
- size 46584922
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0bca72843f365b113570ba86f01e93631029c8fcddf35c3f339dfc9a196b9c40
3
+ size 7067
unet/dd/onnx_report.txt DELETED
@@ -1,76 +0,0 @@
1
- DynamicDispatch Offload - not offloaded
2
- +-------------------+-------+------------------------+------------------------+
3
- | Op Type | Count | Inputs | Outputs |
4
- +===================+=======+========================+========================+
5
- | Cast | 1 | [12,1] - INT64 | [12,1] - FLOAT |
6
- | CastAvx | 1 | [2,128,128,4] - FLOAT | [2,128,128,4] - |
7
- | | | | BFLOAT16 |
8
- | CastAvx | 1 | [2,77,2048] - FLOAT | [2,77,2048] - BFLOAT16 |
9
- | CastAvx | 1 | [2,320] - FLOAT | [2,320] - BFLOAT16 |
10
- | CastAvx | 1 | [2,2816] - FLOAT | [2,2816] - BFLOAT16 |
11
- | CastAvx | 1 | [2,128,128,4] - | [2,128,128,4] - FLOAT |
12
- | | | BFLOAT16 | |
13
- | Concat | 1 | [12,128] - FLOAT | [12,256] - FLOAT |
14
- | | | [12,128] - FLOAT | |
15
- | Concat | 1 | [2,160] - FLOAT | [2,320] - FLOAT |
16
- | | | [2,160] - FLOAT | |
17
- | Concat | 1 | [2,1280] - FLOAT | [2,2816] - FLOAT |
18
- | | | [2,1536] - FLOAT | |
19
- | Cos | 1 | [12,128] - FLOAT | [12,128] - FLOAT |
20
- | Cos | 1 | [2,160] - FLOAT | [2,160] - FLOAT |
21
- | Expand | 1 | [1] - FLOAT | [2] - FLOAT |
22
- | | | [1] - INT64 | |
23
- | Mul | 1 | [12,1] - FLOAT | [12,128] - FLOAT |
24
- | | | [1,128] - FLOAT | |
25
- | Mul | 1 | [2,1] - FLOAT | [2,160] - FLOAT |
26
- | | | [1,160] - FLOAT | |
27
- | Reshape | 1 | [2] - FLOAT | [2,1] - FLOAT |
28
- | | | [2] - INT64 | |
29
- | Reshape | 1 | [2,6] - INT64 | [12,1] - INT64 |
30
- | | | [2] - INT64 | |
31
- | Reshape | 1 | [12,256] - FLOAT | [2,1536] - FLOAT |
32
- | | | [2] - INT64 | |
33
- | Sin | 1 | [12,128] - FLOAT | [12,128] - FLOAT |
34
- | Sin | 1 | [2,160] - FLOAT | [2,160] - FLOAT |
35
- | Transpose | 1 | [2,4,128,128] - FLOAT | [2,128,128,4] - FLOAT |
36
- | Transpose | 1 | [2,128,128,4] - FLOAT | [2,4,128,128] - FLOAT |
37
- +-------------------+-------+------------------------+------------------------+
38
- | Not offloaded sum | 21 | | |
39
- +-------------------+-------+------------------------+------------------------+
40
- DynamicDispatch Offload - offloaded
41
- +-------------------+-----------------+-------------------+-------------------+
42
- | Op Type | Count | Inputs | Outputs |
43
- +===================+=================+===================+===================+
44
- | DynamicDispatch | 1 | [2,128,128,4] - | [2,128,128,4] - |
45
- | | | BFLOAT16 | BFLOAT16 |
46
- | | | [2,2816] - | |
47
- | | | BFLOAT16 | |
48
- | | | [2,77,2048] - | |
49
- | | | BFLOAT16 | |
50
- | | | [2,320] - | |
51
- | | | BFLOAT16 | |
52
- +-------------------+-----------------+-------------------+-------------------+
53
- | Offloaded sum | 1 | | |
54
- +-------------------+-----------------+-------------------+-------------------+
55
- | Offloaded Op | SDFlatMHA | | |
56
- | Types | SDCastBf2Bfp | | |
57
- | | SDGemm | | |
58
- | | SDGemm_bfp | | |
59
- | | SDConcat_bfp | | |
60
- | | SDLayerNorm_bfp | | |
61
- | | SDConv_bfp | | |
62
- | | SDSilu | | |
63
- | | SDCastBfp2Bf | | |
64
- | | SDAdd_bfp | | |
65
- | | SDMHA_bfp | | |
66
- | | SDMul_bfp | | |
67
- | | SDResize_bfp | | |
68
- | | SDAdd | | |
69
- | | SDConv | | |
70
- | | SDGroupNorm_bfp | | |
71
- +-------------------+-----------------+-------------------+-------------------+
72
- | Offloaded sum (dd | 1742 | | |
73
- | fusion) | | | |
74
- | Offload Ratio (dd | 98.81% | | |
75
- | fusion) | | | |
76
- +-------------------+-----------------+-------------------+-------------------+
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
unet/dd/replaced.onnx CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:d6552f886f2b2f7add90da12a82079663356a41d7792ff0fcbaf437304e05c6d
3
- size 1845731
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ac00ccf10f88bf6dfd68550cd203b8895a99c8b954b8ef8dde377aa6d7b3a7b6
3
+ size 1953134
vae_decoder/config.json CHANGED
@@ -1,31 +1,31 @@
1
- {
2
- "_class_name": "AutoencoderKL",
3
- "_diffusers_version": "0.19.0.dev0",
4
- "act_fn": "silu",
5
- "block_out_channels": [
6
- 128,
7
- 256,
8
- 512,
9
- 512
10
- ],
11
- "down_block_types": [
12
- "DownEncoderBlock2D",
13
- "DownEncoderBlock2D",
14
- "DownEncoderBlock2D",
15
- "DownEncoderBlock2D"
16
- ],
17
- "force_upcast": true,
18
- "in_channels": 3,
19
- "latent_channels": 4,
20
- "layers_per_block": 2,
21
- "norm_num_groups": 32,
22
- "out_channels": 3,
23
- "sample_size": 1024,
24
- "scaling_factor": 0.13025,
25
- "up_block_types": [
26
- "UpDecoderBlock2D",
27
- "UpDecoderBlock2D",
28
- "UpDecoderBlock2D",
29
- "UpDecoderBlock2D"
30
- ]
31
- }
 
1
+ {
2
+ "_class_name": "AutoencoderKL",
3
+ "_diffusers_version": "0.19.0.dev0",
4
+ "act_fn": "silu",
5
+ "block_out_channels": [
6
+ 128,
7
+ 256,
8
+ 512,
9
+ 512
10
+ ],
11
+ "down_block_types": [
12
+ "DownEncoderBlock2D",
13
+ "DownEncoderBlock2D",
14
+ "DownEncoderBlock2D",
15
+ "DownEncoderBlock2D"
16
+ ],
17
+ "force_upcast": true,
18
+ "in_channels": 3,
19
+ "latent_channels": 4,
20
+ "layers_per_block": 2,
21
+ "norm_num_groups": 32,
22
+ "out_channels": 3,
23
+ "sample_size": 1024,
24
+ "scaling_factor": 0.13025,
25
+ "up_block_types": [
26
+ "UpDecoderBlock2D",
27
+ "UpDecoderBlock2D",
28
+ "UpDecoderBlock2D",
29
+ "UpDecoderBlock2D"
30
+ ]
31
+ }
vae_decoder/dd/cache/NhwcConv_0-post_quant_convConv_meta.json CHANGED
The diff for this file is too large to render. See raw diff
 
vae_decoder/dd/dd_metastate_Sd15_Decoder_NhwcConv_0-post_quant_convConv.ctrlpkt DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:4650279f5948218b061bac546f1fea0a9353f2eb1dbf3cd14e3735c005266b2a
3
- size 77767176
 
 
 
 
unet/dd/dd_metastate_SD15_Unet_NhwcConv_0-unetconv_inConv.ctrlpkt → vae_decoder/dd/dd_metastate_Sd15_Decoder_NhwcConv_0-post_quant_convConv.ddbundle RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:c2879a4292c14dbf15ffec07f04db70d2338c7efe3cdc6f62330a940ed76438c
3
- size 167321796
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d950f0330501ba30ab62df6ddb1dfe829d1d351bc0029a218851dbb483eafca4
3
+ size 717607226
vae_decoder/dd/dd_metastate_Sd15_Decoder_NhwcConv_0-post_quant_convConv.state CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:40c5c84d5fec429fb4be7b916a8b8f20ef2c23f913c3d6b74c05d914aa0b5a15
3
- size 22813525
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e699e9bb2ff2de2a464b0fe57cf627edbc2ddfa6d6e93afb08beb213f3addd52
3
+ size 5420
vae_decoder/dd/dd_metastate_Sd15_Decoder_NhwcConv_0-post_quant_convConv.super DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:10334e04fe3040fe12d4b37e1f6d0c08f7ebf5d0ab4e7e5acc8589bc82c0d57a
3
- size 785568
 
 
 
 
vae_decoder/dd/onnx_report.txt DELETED
@@ -1,35 +0,0 @@
1
- DynamicDispatch Offload - not offloaded
2
- +-------------------+-------+------------------------+------------------------+
3
- | Op Type | Count | Inputs | Outputs |
4
- +===================+=======+========================+========================+
5
- | CastAvx | 1 | [1,128,128,4] - FLOAT | [1,128,128,4] - |
6
- | | | | BFLOAT16 |
7
- | CastAvx | 1 | [1,1024,1024,3] - | [1,1024,1024,3] - |
8
- | | | BFLOAT16 | FLOAT |
9
- | Transpose | 1 | [1,4,128,128] - FLOAT | [1,128,128,4] - FLOAT |
10
- | Transpose | 1 | [1,1024,1024,3] - | [1,3,1024,1024] - |
11
- | | | FLOAT | FLOAT |
12
- +-------------------+-------+------------------------+------------------------+
13
- | Not offloaded sum | 4 | | |
14
- +-------------------+-------+------------------------+------------------------+
15
- DynamicDispatch Offload - offloaded
16
- +--------------------+-------------+--------------------+---------------------+
17
- | Op Type | Count | Inputs | Outputs |
18
- +====================+=============+====================+=====================+
19
- | DynamicDispatch | 1 | [1,128,128,4] - | [1,1024,1024,3] - |
20
- | | | BFLOAT16 | BFLOAT16 |
21
- +--------------------+-------------+--------------------+---------------------+
22
- | Offloaded sum | 1 | | |
23
- +--------------------+-------------+--------------------+---------------------+
24
- | Offloaded Op Types | SDAdd | | |
25
- | | SDConv | | |
26
- | | SDMHA_VAE | | |
27
- | | SDGroupNorm | | |
28
- | | SDResize | | |
29
- | | SDGemm | | |
30
- +--------------------+-------------+--------------------+---------------------+
31
- | Offloaded sum (dd | 89 | | |
32
- | fusion) | | | |
33
- | Offload Ratio (dd | 95.7% | | |
34
- | fusion) | | | |
35
- +--------------------+-------------+--------------------+---------------------+
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
vae_decoder/dd/replaced.onnx CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:47b2dc68234073a32ecffc65ad06dbba3f1e0e0dd0bfa005e37d1fe18f928e37
3
- size 46031
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:188300511a830dd0f8f5db137e5524fa6bd233050f402a1cf57f70bf32ecdbbb
3
+ size 47015