Add Neuron-optimized files for CompVis/stable-diffusion-v1-4

🤖 Neuron Export Bot: Adding AWS Neuron-optimized model files.

Original model: [CompVis/stable-diffusion-v1-4](https://huggingface.co/CompVis/stable-diffusion-v1-4)
Task: text-to-image
Generated by: [badaoui](https://huggingface.co/badaoui)
Generated using: [Optimum Neuron Compiler Space](https://huggingface.co/spaces/optimum/neuron-export)

These files have been pre-compiled for AWS Neuron devices (Inferentia/Trainium) and should provide improved inference performance.

.gitattributes

@@ -33,3 +33,7 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+text_encoder/model.neuron filter=lfs diff=lfs merge=lfs -text
+unet/model.neuron filter=lfs diff=lfs merge=lfs -text
+vae_encoder/model.neuron filter=lfs diff=lfs merge=lfs -text
+vae_decoder/model.neuron filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,38 @@
+---
+tags:
+- neuron
+- optimized
+- aws-neuron
+- text-to-image
+base_model: CompVis/stable-diffusion-v1-4
+---
+
+# Neuron-Optimized CompVis/stable-diffusion-v1-4
+
+This repository contains AWS Neuron-optimized files for [CompVis/stable-diffusion-v1-4](https://huggingface.co/CompVis/stable-diffusion-v1-4).
+
+## Model Details
+
+- **Base Model**: [CompVis/stable-diffusion-v1-4](https://huggingface.co/CompVis/stable-diffusion-v1-4)
+- **Task**: text-to-image
+- **Optimization**: AWS Neuron compilation
+- **Generated by**: [badaoui](https://huggingface.co/badaoui)
+- **Generated using**: [Optimum Neuron Compiler Space](https://huggingface.co/spaces/optimum/neuron-export)
+
+## Usage
+
+This model has been optimized for AWS Neuron devices (Inferentia/Trainium). To use it:
+
+```python
+from optimum.neuron import NeuronStableDiffusionPipeline
+
+model = NeuronStableDiffusionPipeline.from_pretrained("badaoui/CompVis-stable-diffusion-v1-4-neuron")
+```
+
+## Performance
+
+These files are pre-compiled for AWS Neuron devices and should provide improved inference performance compared to the original model when deployed on Inferentia or Trainium instances.
+
+## Original Model
+
+For the original model, training details, and more information, please visit: [CompVis/stable-diffusion-v1-4](https://huggingface.co/CompVis/stable-diffusion-v1-4)

feature_extractor/preprocessor_config.json

@@ -0,0 +1,27 @@
+{
+  "crop_size": {
+    "height": 224,
+    "width": 224
+  },
+  "do_center_crop": true,
+  "do_convert_rgb": true,
+  "do_normalize": true,
+  "do_rescale": true,
+  "do_resize": true,
+  "image_mean": [
+    0.48145466,
+    0.4578275,
+    0.40821073
+  ],
+  "image_processor_type": "CLIPImageProcessor",
+  "image_std": [
+    0.26862954,
+    0.26130258,
+    0.27577711
+  ],
+  "resample": 3,
+  "rescale_factor": 0.00392156862745098,
+  "size": {
+    "shortest_edge": 224
+  }
+}

model_index.json

@@ -0,0 +1,38 @@
+{
+  "_class_name": "StableDiffusionPipeline",
+  "_diffusers_version": "0.34.0",
+  "_name_or_path": "CompVis/stable-diffusion-v1-4",
+  "feature_extractor": [
+    "transformers",
+    "CLIPImageProcessor"
+  ],
+  "image_encoder": [
+    null,
+    null
+  ],
+  "requires_safety_checker": true,
+  "safety_checker": [
+    "stable_diffusion",
+    "StableDiffusionSafetyChecker"
+  ],
+  "scheduler": [
+    "diffusers",
+    "PNDMScheduler"
+  ],
+  "text_encoder": [
+    "transformers",
+    "CLIPTextModel"
+  ],
+  "tokenizer": [
+    "transformers",
+    "CLIPTokenizer"
+  ],
+  "unet": [
+    "diffusers",
+    "UNet2DConditionModel"
+  ],
+  "vae": [
+    "diffusers",
+    "AutoencoderKL"
+  ]
+}

scheduler/scheduler_config.json

@@ -0,0 +1,15 @@
+{
+  "_class_name": "PNDMScheduler",
+  "_diffusers_version": "0.34.0",
+  "beta_end": 0.012,
+  "beta_schedule": "scaled_linear",
+  "beta_start": 0.00085,
+  "clip_sample": false,
+  "num_train_timesteps": 1000,
+  "prediction_type": "epsilon",
+  "set_alpha_to_one": false,
+  "skip_prk_steps": true,
+  "steps_offset": 1,
+  "timestep_spacing": "leading",
+  "trained_betas": null
+}

text_encoder/config.json

@@ -0,0 +1,50 @@
+{
+  "_attn_implementation_autoset": true,
+  "architectures": [
+    "CLIPTextModel"
+  ],
+  "attention_dropout": 0.0,
+  "bos_token_id": 0,
+  "dropout": 0.0,
+  "eos_token_id": 2,
+  "export_model_type": "clip-text-model",
+  "hidden_act": "quick_gelu",
+  "hidden_size": 768,
+  "initializer_factor": 1.0,
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "layer_norm_eps": 1e-05,
+  "max_position_embeddings": 77,
+  "model_type": "clip_text_model",
+  "neuron": {
+    "auto_cast": "matmul",
+    "auto_cast_type": "bf16",
+    "compiler_type": "neuronx-cc",
+    "compiler_version": "2.19.8089.0+8ab9f450",
+    "dynamic_batch_size": false,
+    "inline_weights_to_neff": true,
+    "input_names": [
+      "input_ids"
+    ],
+    "model_type": "clip-text-model",
+    "optlevel": "2",
+    "output_attentions": false,
+    "output_hidden_states": false,
+    "output_names": [
+      "last_hidden_state",
+      "pooler_output"
+    ],
+    "static_batch_size": 1,
+    "static_sequence_length": 77,
+    "task": "feature-extraction",
+    "tensor_parallel_size": 1
+  },
+  "num_attention_heads": 12,
+  "num_hidden_layers": 12,
+  "pad_token_id": 1,
+  "projection_dim": 512,
+  "torch_dtype": "bfloat16",
+  "torchscript": true,
+  "transformers_version": "4.51.3",
+  "vocab_size": 49408
+}

text_encoder/model.neuron

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dd18698495540c1ec851f8bdda03c6d0d8be0c6673c2c272516bb0402806a767
+size 307198263

tokenizer/special_tokens_map.json

@@ -0,0 +1,24 @@
+{
+  "bos_token": {
+    "content": "<|startoftext|>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": "<|endoftext|>",
+  "unk_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  }
+}

tokenizer/tokenizer_config.json

@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "49406": {
+      "content": "<|startoftext|>",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "49407": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "bos_token": "<|startoftext|>",
+  "clean_up_tokenization_spaces": false,
+  "do_lower_case": true,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": {},
+  "model_max_length": 77,
+  "pad_token": "<|endoftext|>",
+  "tokenizer_class": "CLIPTokenizer",
+  "unk_token": "<|endoftext|>"
+}

unet/config.json

@@ -0,0 +1,130 @@
+{
+  "_class_name": "UNet2DConditionModel",
+  "_commit_hash": null,
+  "_diffusers_version": "0.34.0",
+  "_use_default_values": [
+    "use_linear_projection",
+    "transformer_layers_per_block",
+    "resnet_out_scale_factor",
+    "time_embedding_act_fn",
+    "timestep_post_act",
+    "num_attention_heads",
+    "encoder_hid_dim_type",
+    "dual_cross_attention",
+    "num_class_embeds",
+    "reverse_transformer_layers_per_block",
+    "addition_time_embed_dim",
+    "time_embedding_dim",
+    "attention_type",
+    "dropout",
+    "resnet_skip_time_act",
+    "resnet_time_scale_shift",
+    "class_embeddings_concat",
+    "conv_in_kernel",
+    "conv_out_kernel",
+    "mid_block_type",
+    "addition_embed_type",
+    "only_cross_attention",
+    "encoder_hid_dim",
+    "projection_class_embeddings_input_dim",
+    "cross_attention_norm",
+    "time_embedding_type",
+    "time_cond_proj_dim",
+    "upcast_attention",
+    "addition_embed_type_num_heads",
+    "mid_block_only_cross_attention",
+    "class_embed_type"
+  ],
+  "act_fn": "silu",
+  "addition_embed_type": null,
+  "addition_embed_type_num_heads": 64,
+  "addition_time_embed_dim": null,
+  "attention_head_dim": 8,
+  "attention_type": "default",
+  "block_out_channels": [
+    320,
+    640,
+    1280,
+    1280
+  ],
+  "center_input_sample": false,
+  "class_embed_type": null,
+  "class_embeddings_concat": false,
+  "conv_in_kernel": 3,
+  "conv_out_kernel": 3,
+  "cross_attention_dim": 768,
+  "cross_attention_norm": null,
+  "down_block_types": [
+    "CrossAttnDownBlock2D",
+    "CrossAttnDownBlock2D",
+    "CrossAttnDownBlock2D",
+    "DownBlock2D"
+  ],
+  "downsample_padding": 1,
+  "dropout": 0.0,
+  "dual_cross_attention": false,
+  "encoder_hid_dim": null,
+  "encoder_hid_dim_type": null,
+  "flip_sin_to_cos": true,
+  "freq_shift": 0,
+  "in_channels": 4,
+  "layers_per_block": 2,
+  "mid_block_only_cross_attention": null,
+  "mid_block_scale_factor": 1,
+  "mid_block_type": "UNetMidBlock2DCrossAttn",
+  "neuron": {
+    "auto_cast": "matmul",
+    "auto_cast_type": "bf16",
+    "compiler_type": "neuronx-cc",
+    "compiler_version": "2.19.8089.0+8ab9f450",
+    "dynamic_batch_size": false,
+    "inline_weights_to_neff": true,
+    "input_names": [
+      "sample",
+      "timestep",
+      "encoder_hidden_states"
+    ],
+    "model_type": "unet",
+    "optlevel": "2",
+    "output_attentions": false,
+    "output_hidden_states": false,
+    "output_names": [
+      "sample"
+    ],
+    "static_batch_size": 1,
+    "static_height": 128,
+    "static_num_channels": 4,
+    "static_sequence_length": 77,
+    "static_vae_scale_factor": 8,
+    "static_width": 128,
+    "task": "semantic-segmentation",
+    "tensor_parallel_size": 1
+  },
+  "norm_eps": 1e-05,
+  "norm_num_groups": 32,
+  "num_attention_heads": null,
+  "num_class_embeds": null,
+  "only_cross_attention": false,
+  "out_channels": 4,
+  "projection_class_embeddings_input_dim": null,
+  "resnet_out_scale_factor": 1.0,
+  "resnet_skip_time_act": false,
+  "resnet_time_scale_shift": "default",
+  "reverse_transformer_layers_per_block": null,
+  "sample_size": 64,
+  "time_cond_proj_dim": null,
+  "time_embedding_act_fn": null,
+  "time_embedding_dim": null,
+  "time_embedding_type": "positional",
+  "timestep_post_act": null,
+  "transformer_layers_per_block": 1,
+  "transformers_version": null,
+  "up_block_types": [
+    "UpBlock2D",
+    "CrossAttnUpBlock2D",
+    "CrossAttnUpBlock2D",
+    "CrossAttnUpBlock2D"
+  ],
+  "upcast_attention": false,
+  "use_linear_projection": false
+}

unet/model.neuron

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:609aed6600c12d5840fe6ea57e213358339e568860b06d8280402426e681c303
+size 2853990707

vae_decoder/config.json

@@ -0,0 +1,73 @@
+{
+  "_class_name": "AutoencoderKL",
+  "_commit_hash": null,
+  "_diffusers_version": "0.34.0",
+  "_use_default_values": [
+    "latents_mean",
+    "use_post_quant_conv",
+    "norm_num_groups",
+    "use_quant_conv",
+    "mid_block_add_attention",
+    "shift_factor",
+    "latents_std",
+    "force_upcast"
+  ],
+  "act_fn": "silu",
+  "block_out_channels": [
+    128,
+    256,
+    512,
+    512
+  ],
+  "down_block_types": [
+    "DownEncoderBlock2D",
+    "DownEncoderBlock2D",
+    "DownEncoderBlock2D",
+    "DownEncoderBlock2D"
+  ],
+  "force_upcast": true,
+  "in_channels": 3,
+  "latent_channels": 4,
+  "latents_mean": null,
+  "latents_std": null,
+  "layers_per_block": 2,
+  "mid_block_add_attention": true,
+  "neuron": {
+    "auto_cast": "matmul",
+    "auto_cast_type": "bf16",
+    "compiler_type": "neuronx-cc",
+    "compiler_version": "2.19.8089.0+8ab9f450",
+    "dynamic_batch_size": false,
+    "inline_weights_to_neff": true,
+    "input_names": [
+      "latent_sample"
+    ],
+    "model_type": "vae-decoder",
+    "optlevel": "2",
+    "output_attentions": false,
+    "output_hidden_states": false,
+    "output_names": [
+      "sample"
+    ],
+    "static_batch_size": 1,
+    "static_height": 128,
+    "static_num_channels": 4,
+    "static_width": 128,
+    "task": "semantic-segmentation",
+    "tensor_parallel_size": 1
+  },
+  "norm_num_groups": 32,
+  "out_channels": 3,
+  "sample_size": 512,
+  "scaling_factor": 0.18215,
+  "shift_factor": null,
+  "transformers_version": null,
+  "up_block_types": [
+    "UpDecoderBlock2D",
+    "UpDecoderBlock2D",
+    "UpDecoderBlock2D",
+    "UpDecoderBlock2D"
+  ],
+  "use_post_quant_conv": true,
+  "use_quant_conv": true
+}

vae_decoder/model.neuron

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:46e461b9cd7437ebfb2c04a4b3f1d823b180f6e7faaa32f9701d38427d92035d
+size 370909619

vae_encoder/config.json

@@ -0,0 +1,73 @@
+{
+  "_class_name": "AutoencoderKL",
+  "_commit_hash": null,
+  "_diffusers_version": "0.34.0",
+  "_use_default_values": [
+    "latents_mean",
+    "use_post_quant_conv",
+    "norm_num_groups",
+    "use_quant_conv",
+    "mid_block_add_attention",
+    "shift_factor",
+    "latents_std",
+    "force_upcast"
+  ],
+  "act_fn": "silu",
+  "block_out_channels": [
+    128,
+    256,
+    512,
+    512
+  ],
+  "down_block_types": [
+    "DownEncoderBlock2D",
+    "DownEncoderBlock2D",
+    "DownEncoderBlock2D",
+    "DownEncoderBlock2D"
+  ],
+  "force_upcast": true,
+  "in_channels": 3,
+  "latent_channels": 4,
+  "latents_mean": null,
+  "latents_std": null,
+  "layers_per_block": 2,
+  "mid_block_add_attention": true,
+  "neuron": {
+    "auto_cast": "matmul",
+    "auto_cast_type": "bf16",
+    "compiler_type": "neuronx-cc",
+    "compiler_version": "2.19.8089.0+8ab9f450",
+    "dynamic_batch_size": false,
+    "inline_weights_to_neff": true,
+    "input_names": [
+      "sample"
+    ],
+    "model_type": "vae-encoder",
+    "optlevel": "2",
+    "output_attentions": false,
+    "output_hidden_states": false,
+    "output_names": [
+      "latent_parameters"
+    ],
+    "static_batch_size": 1,
+    "static_height": 1024,
+    "static_num_channels": 3,
+    "static_width": 1024,
+    "task": "semantic-segmentation",
+    "tensor_parallel_size": 1
+  },
+  "norm_num_groups": 32,
+  "out_channels": 3,
+  "sample_size": 512,
+  "scaling_factor": 0.18215,
+  "shift_factor": null,
+  "transformers_version": null,
+  "up_block_types": [
+    "UpDecoderBlock2D",
+    "UpDecoderBlock2D",
+    "UpDecoderBlock2D",
+    "UpDecoderBlock2D"
+  ],
+  "use_post_quant_conv": true,
+  "use_quant_conv": true
+}

vae_encoder/model.neuron

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4196c2f2b11b15c0ce566cd2bf19b2be5ad4b99d14f84de9b140279e006e185f
+size 222985073