Upload a better quant

model_index.json

@@ -1,6 +1,6 @@
 {
   "_class_name": "FluxPipeline",
-  "_diffusers_version": "0.36.0.dev0",
+  "_diffusers_version": "0.35.1",
   "_name_or_path": "black-forest-labs/FLUX.1-dev",
   "feature_extractor": [
     null,

scheduler/scheduler_config.json

@@ -1,6 +1,6 @@
 {
   "_class_name": "FlowMatchEulerDiscreteScheduler",
-  "_diffusers_version": "0.36.0.dev0",
+  "_diffusers_version": "0.35.1",
   "base_image_seq_len": 256,
   "base_shift": 0.5,
   "invert_sigmas": false,

text_encoder_2/config.json

@@ -23,21 +23,21 @@
   "output_past": true,
   "pad_token_id": 0,
   "quantization_config": {
-    "add_skip_keys": false,
+    "add_skip_keys": true,
     "dequantize_fp32": false,
     "group_size": 0,
     "is_integer": true,
     "modules_dtype_dict": {},
     "modules_to_not_convert": [
-      "wo",
       "wo"
     ],
     "non_blocking": false,
     "quant_conv": false,
     "quant_method": "sdnq",
-    "quantization_device": null,
-    "return_device": null,
+    "quantization_device": "xpu",
+    "return_device": "cpu",
     "svd_rank": 32,
+    "svd_steps": 8,
     "use_quantized_matmul": false,
     "use_quantized_matmul_conv": false,
     "use_svd": true,

text_encoder_2/model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:cfc8e0a4a7d2b29c410fe8a25059e0c154d0acca214d8571f0af671dc1c8fe96
+oid sha256:0301d1a2d789a1b4e6570566ee114343b790aeeb36ba0c8a59e866b262385510
 size 6309787408

text_encoder_2/quantization_config.json

@@ -1,19 +1,19 @@
 {
-  "add_skip_keys": false,
+  "add_skip_keys": true,
   "dequantize_fp32": false,
   "group_size": 0,
   "is_integer": true,
   "modules_dtype_dict": {},
   "modules_to_not_convert": [
-    "wo",
     "wo"
   ],
   "non_blocking": false,
   "quant_conv": false,
   "quant_method": "sdnq",
-  "quantization_device": null,
-  "return_device": null,
+  "quantization_device": "xpu",
+  "return_device": "cpu",
   "svd_rank": 32,
+  "svd_steps": 8,
   "use_quantized_matmul": false,
   "use_quantized_matmul_conv": false,
   "use_svd": true,

tokenizer_2/tokenizer.json

@@ -1,7 +1,14 @@
 {
   "version": "1.0",
   "truncation": null,
-  "padding": null,
+  "padding": {
+    "strategy": "BatchLongest",
+    "direction": "Right",
+    "pad_to_multiple_of": null,
+    "pad_id": 0,
+    "pad_type_id": 0,
+    "pad_token": "<pad>"
+  },
   "added_tokens": [
     {
       "id": 0,

transformer/config.json

@@ -1,6 +1,6 @@
 {
   "_class_name": "FluxTransformer2DModel",
-  "_diffusers_version": "0.36.0.dev0",
+  "_diffusers_version": "0.35.1",
   "_name_or_path": "black-forest-labs/FLUX.1-dev",
   "attention_head_dim": 128,
   "axes_dims_rope": [
@@ -18,21 +18,27 @@
   "patch_size": 1,
   "pooled_projection_dim": 768,
   "quantization_config": {
-    "add_skip_keys": false,
+    "add_skip_keys": true,
     "dequantize_fp32": false,
     "group_size": 0,
     "is_integer": true,
     "modules_dtype_dict": {},
     "modules_to_not_convert": [
+      ".proj_out",
       "pos_embed",
-      "norm"
+      ".x_embedder",
+      ".norm_out",
+      "single_transformer_blocks.0.norm.linear.weight",
+      ".context_embedder",
+      ".time_text_embed"
     ],
     "non_blocking": false,
     "quant_conv": false,
     "quant_method": "sdnq",
-    "quantization_device": null,
-    "return_device": null,
+    "quantization_device": "xpu",
+    "return_device": "cpu",
     "svd_rank": 32,
+    "svd_steps": 8,
     "use_quantized_matmul": false,
     "use_quantized_matmul_conv": false,
     "use_svd": true,

transformer/diffusion_pytorch_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9b1de96674e1330a1c9a30429a099d84e9eac2db08a8692290207d11d95bb098
-size 8224750072
+oid sha256:77d0d617e815df520a5aee5c8e77fd3e973e54b5a47d54cdeb9c1db8c73b8598
+size 6806132768

transformer/quantization_config.json

@@ -1,19 +1,25 @@
 {
-  "add_skip_keys": false,
+  "add_skip_keys": true,
   "dequantize_fp32": false,
   "group_size": 0,
   "is_integer": true,
   "modules_dtype_dict": {},
   "modules_to_not_convert": [
+    ".proj_out",
     "pos_embed",
-    "norm"
+    ".x_embedder",
+    ".norm_out",
+    "single_transformer_blocks.0.norm.linear.weight",
+    ".context_embedder",
+    ".time_text_embed"
   ],
   "non_blocking": false,
   "quant_conv": false,
   "quant_method": "sdnq",
-  "quantization_device": null,
-  "return_device": null,
+  "quantization_device": "xpu",
+  "return_device": "cpu",
   "svd_rank": 32,
+  "svd_steps": 8,
   "use_quantized_matmul": false,
   "use_quantized_matmul_conv": false,
   "use_svd": true,

vae/config.json

@@ -1,7 +1,7 @@
 {
   "_class_name": "AutoencoderKL",
-  "_diffusers_version": "0.36.0.dev0",
-  "_name_or_path": "/mnt/DataSSD/AI/models/sd-webui/Diffusers/models--black-forest-labs--FLUX.1-dev/snapshots/3de623fc3c33e44ffbe2bad470d0f45bccf2eb21/vae",
+  "_diffusers_version": "0.35.1",
+  "_name_or_path": "/home/disty/.cache/huggingface/hub/models--black-forest-labs--FLUX.1-dev/snapshots/3de623fc3c33e44ffbe2bad470d0f45bccf2eb21/vae",
   "act_fn": "silu",
   "block_out_channels": [
     128,