Upload transformer component

transformer/config.json

@@ -1,6 +1,6 @@
 {
   "_class_name": "Flux2Transformer2DModel",
-  "_diffusers_version": "0.38.0",
+  "_diffusers_version": "0.37.1",
   "_name_or_path": "/root/.cache/huggingface/hub/models--black-forest-labs--FLUX.2-klein-9B/snapshots/92196c8e11f7b6cf2b7493e037d8c5345c559216/transformer",
   "attention_head_dim": 128,
   "axes_dims_rope": [
@@ -22,29 +22,12 @@
   "quantization_config": {
     "add_skip_keys": true,
     "dequantize_fp32": true,
-    "dynamic_loss_threshold": 0.01,
+    "dynamic_loss_threshold": 0.1,
     "group_size": 0,
     "is_integer": true,
     "is_training": false,
     "modules_dtype_dict": {
-      "int4": [
-        "transformer_blocks.0.attn.to_q.weight",
-        "transformer_blocks.0.attn.to_k.weight"
-      ],
-      "int5": [
-        "single_transformer_blocks.0.attn.to_qkv_mlp_proj.weight",
-        "single_transformer_blocks.1.attn.to_out.weight",
-        "single_transformer_blocks.2.attn.to_out.weight",
-        "single_transformer_blocks.4.attn.to_qkv_mlp_proj.weight",
-        "single_transformer_blocks.12.attn.to_qkv_mlp_proj.weight",
-        "single_transformer_blocks.17.attn.to_qkv_mlp_proj.weight",
-        "single_transformer_blocks.19.attn.to_qkv_mlp_proj.weight",
-        "single_transformer_blocks.20.attn.to_qkv_mlp_proj.weight",
-        "single_transformer_blocks.21.attn.to_qkv_mlp_proj.weight",
-        "single_transformer_blocks.22.attn.to_qkv_mlp_proj.weight",
-        "single_transformer_blocks.23.attn.to_qkv_mlp_proj.weight"
-      ],
-      "uint4": [
+      "int3": [
         "transformer_blocks.0.attn.to_v.weight",
         "transformer_blocks.0.attn.to_out.0.weight",
         "transformer_blocks.0.attn.add_q_proj.weight",
@@ -139,11 +122,15 @@
         "transformer_blocks.7.ff.linear_out.weight",
         "transformer_blocks.7.ff_context.linear_in.weight",
         "transformer_blocks.7.ff_context.linear_out.weight",
+        "single_transformer_blocks.0.attn.to_qkv_mlp_proj.weight",
         "single_transformer_blocks.0.attn.to_out.weight",
         "single_transformer_blocks.1.attn.to_qkv_mlp_proj.weight",
+        "single_transformer_blocks.1.attn.to_out.weight",
         "single_transformer_blocks.2.attn.to_qkv_mlp_proj.weight",
+        "single_transformer_blocks.2.attn.to_out.weight",
         "single_transformer_blocks.3.attn.to_qkv_mlp_proj.weight",
         "single_transformer_blocks.3.attn.to_out.weight",
+        "single_transformer_blocks.4.attn.to_qkv_mlp_proj.weight",
         "single_transformer_blocks.4.attn.to_out.weight",
         "single_transformer_blocks.5.attn.to_qkv_mlp_proj.weight",
         "single_transformer_blocks.5.attn.to_out.weight",
@@ -159,6 +146,7 @@
         "single_transformer_blocks.10.attn.to_out.weight",
         "single_transformer_blocks.11.attn.to_qkv_mlp_proj.weight",
         "single_transformer_blocks.11.attn.to_out.weight",
+        "single_transformer_blocks.12.attn.to_qkv_mlp_proj.weight",
         "single_transformer_blocks.12.attn.to_out.weight",
         "single_transformer_blocks.13.attn.to_qkv_mlp_proj.weight",
         "single_transformer_blocks.13.attn.to_out.weight",
@@ -168,26 +156,36 @@
         "single_transformer_blocks.15.attn.to_out.weight",
         "single_transformer_blocks.16.attn.to_qkv_mlp_proj.weight",
         "single_transformer_blocks.16.attn.to_out.weight",
+        "single_transformer_blocks.17.attn.to_qkv_mlp_proj.weight",
         "single_transformer_blocks.17.attn.to_out.weight",
         "single_transformer_blocks.18.attn.to_qkv_mlp_proj.weight",
         "single_transformer_blocks.18.attn.to_out.weight",
+        "single_transformer_blocks.19.attn.to_qkv_mlp_proj.weight",
         "single_transformer_blocks.19.attn.to_out.weight",
+        "single_transformer_blocks.20.attn.to_qkv_mlp_proj.weight",
         "single_transformer_blocks.20.attn.to_out.weight",
+        "single_transformer_blocks.21.attn.to_qkv_mlp_proj.weight",
         "single_transformer_blocks.21.attn.to_out.weight",
+        "single_transformer_blocks.22.attn.to_qkv_mlp_proj.weight",
         "single_transformer_blocks.22.attn.to_out.weight",
+        "single_transformer_blocks.23.attn.to_qkv_mlp_proj.weight",
         "single_transformer_blocks.23.attn.to_out.weight"
+      ],
+      "uint2": [
+        "transformer_blocks.0.attn.to_q.weight",
+        "transformer_blocks.0.attn.to_k.weight"
       ]
     },
     "modules_quant_config": {},
     "modules_to_not_convert": [
+      "double_stream_modulation_txt",
+      "time_guidance_embed",
+      "x_embedder",
       ".proj_out",
       "single_stream_modulation",
-      "double_stream_modulation_img",
-      "norm_out",
-      "x_embedder",
       "context_embedder",
-      "time_guidance_embed",
-      "double_stream_modulation_txt"
+      "norm_out",
+      "double_stream_modulation_img"
     ],
     "non_blocking": false,
     "quant_conv": false,
@@ -198,7 +196,7 @@
     "return_device": "cpu",
     "sdnq_version": "0.1.8",
     "svd_rank": 32,
-    "svd_steps": 8,
+    "svd_steps": 16,
     "use_dynamic_quantization": true,
     "use_grad_ckpt": true,
     "use_quantized_matmul": false,

transformer/diffusion_pytorch_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e01ac8976ff915fb864ec7395cd0bd3b2d60a659c0ac933558a69b2dfb97a42e
+size 4701916360