segmind

/

SSD-1B

like 791

pip install git+https://github.com/huggingface/diffusers
pip install transformers accelerate safetensors
from diffusers import StableDiffusionXLPipeline
import torch
pipe = StableDiffusionXLPipeline.from_pretrained("segmind/SSD-1B", torch_dtype=torch.float16, use_safetensors=True, variant="fp16")
pipe.to("cuda")
# if using torch < 2.0
# pipe.enable_xformers_memory_efficient_attention()
prompt = "An astronaut riding a green horse" # Your prompt here
neg_prompt = "ugly, blurry, poor quality" # Negative prompt here
image = pipe(prompt=prompt, negative_prompt=neg_prompt).images[0]

The config attributes {'reverse_transformer_layers_per_block': [[4, 4, 10], [2, 1, 1], 1]} were passed to UNet2DConditionModel, but are not expected and will be ignored. Please verify your config.json configuration file.
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
Cell In[13], line 1
----> 1 pipe = StableDiffusionXLPipeline.from_pretrained("segmind/SSD-1B", torch_dtype=torch.float16, use_safetensors=True, variant="fp16")

File /opt/conda/lib/python3.10/site-packages/diffusers/pipelines/pipeline_utils.py:1105, in DiffusionPipeline.from_pretrained(cls, pretrained_model_name_or_path, **kwargs)
   1102     loaded_sub_model = passed_class_obj[name]
   1103 else:
   1104     # load sub model
-> 1105     loaded_sub_model = load_sub_model(
   1106         library_name=library_name,
   1107         class_name=class_name,
   1108         importable_classes=importable_classes,
   1109         pipelines=pipelines,
   1110         is_pipeline_module=is_pipeline_module,
   1111         pipeline_class=pipeline_class,
   1112         torch_dtype=torch_dtype,
   1113         provider=provider,
   1114         sess_options=sess_options,
   1115         device_map=device_map,
   1116         max_memory=max_memory,
   1117         offload_folder=offload_folder,
   1118         offload_state_dict=offload_state_dict,
   1119         model_variants=model_variants,
   1120         name=name,
   1121         from_flax=from_flax,
   1122         variant=variant,
   1123         low_cpu_mem_usage=low_cpu_mem_usage,
   1124         cached_folder=cached_folder,
   1125     )
   1126     logger.info(
   1127         f"Loaded {name} as {class_name} from `{name}` subfolder of {pretrained_model_name_or_path}."
   1128     )
   1130 init_kwargs[name] = loaded_sub_model  # UNet(...), # DiffusionSchedule(...)

File /opt/conda/lib/python3.10/site-packages/diffusers/pipelines/pipeline_utils.py:472, in load_sub_model(library_name, class_name, importable_classes, pipelines, is_pipeline_module, pipeline_class, torch_dtype, provider, sess_options, device_map, max_memory, offload_folder, offload_state_dict, model_variants, name, from_flax, variant, low_cpu_mem_usage, cached_folder)
    470 # check if the module is in a subdirectory
    471 if os.path.isdir(os.path.join(cached_folder, name)):
--> 472     loaded_sub_model = load_method(os.path.join(cached_folder, name), **loading_kwargs)
    473 else:
    474     # else load from the root directory
    475     loaded_sub_model = load_method(cached_folder, **loading_kwargs)

File /opt/conda/lib/python3.10/site-packages/diffusers/models/modeling_utils.py:636, in ModelMixin.from_pretrained(cls, pretrained_model_name_or_path, **kwargs)
    633 if low_cpu_mem_usage:
    634     # Instantiate model with empty weights
    635     with accelerate.init_empty_weights():
--> 636         model = cls.from_config(config, **unused_kwargs)
    638     # if device_map is None, load the state dict and move the params from meta device to the cpu
    639     if device_map is None:

File /opt/conda/lib/python3.10/site-packages/diffusers/configuration_utils.py:254, in ConfigMixin.from_config(cls, config, return_unused_kwargs, **kwargs)
    251         init_dict[deprecated_kwarg] = unused_kwargs.pop(deprecated_kwarg)
    253 # Return model and optionally state and/or unused_kwargs
--> 254 model = cls(**init_dict)
    256 # make sure to also save config parameters that might be used for compatible classes
    257 model.register_to_config(**hidden_dict)

File /opt/conda/lib/python3.10/site-packages/diffusers/configuration_utils.py:636, in register_to_config.<locals>.inner_init(self, *args, **kwargs)
    634 new_kwargs = {**config_init_kwargs, **new_kwargs}
    635 getattr(self, "register_to_config")(**new_kwargs)
--> 636 init(self, *args, **init_kwargs)

File /opt/conda/lib/python3.10/site-packages/diffusers/models/unet_2d_condition.py:440, in UNet2DConditionModel.__init__(self, sample_size, in_channels, out_channels, center_input_sample, flip_sin_to_cos, freq_shift, down_block_types, mid_block_type, up_block_types, only_cross_attention, block_out_channels, layers_per_block, downsample_padding, mid_block_scale_factor, dropout, act_fn, norm_num_groups, norm_eps, cross_attention_dim, transformer_layers_per_block, encoder_hid_dim, encoder_hid_dim_type, attention_head_dim, num_attention_heads, dual_cross_attention, use_linear_projection, class_embed_type, addition_embed_type, addition_time_embed_dim, num_class_embeds, upcast_attention, resnet_time_scale_shift, resnet_skip_time_act, resnet_out_scale_factor, time_embedding_type, time_embedding_dim, time_embedding_act_fn, timestep_post_act, time_cond_proj_dim, conv_in_kernel, conv_out_kernel, projection_class_embeddings_input_dim, attention_type, class_embeddings_concat, mid_block_only_cross_attention, cross_attention_norm, addition_embed_type_num_heads)
    437     output_channel = block_out_channels[i]
    438     is_final_block = i == len(block_out_channels) - 1
--> 440     down_block = get_down_block(
    441         down_block_type,
    442         num_layers=layers_per_block[i],
    443         transformer_layers_per_block=transformer_layers_per_block[i],
    444         in_channels=input_channel,
    445         out_channels=output_channel,
    446         temb_channels=blocks_time_embed_dim,
    447         add_downsample=not is_final_block,
    448         resnet_eps=norm_eps,
    449         resnet_act_fn=act_fn,
    450         resnet_groups=norm_num_groups,
    451         cross_attention_dim=cross_attention_dim[i],
    452         num_attention_heads=num_attention_heads[i],
    453         downsample_padding=downsample_padding,
    454         dual_cross_attention=dual_cross_attention,
    455         use_linear_projection=use_linear_projection,
    456         only_cross_attention=only_cross_attention[i],
    457         upcast_attention=upcast_attention,
    458         resnet_time_scale_shift=resnet_time_scale_shift,
    459         attention_type=attention_type,
    460         resnet_skip_time_act=resnet_skip_time_act,
    461         resnet_out_scale_factor=resnet_out_scale_factor,
    462         cross_attention_norm=cross_attention_norm,
    463         attention_head_dim=attention_head_dim[i] if attention_head_dim[i] is not None else output_channel,
    464         dropout=dropout,
    465     )
    466     self.down_blocks.append(down_block)
    468 # mid

File /opt/conda/lib/python3.10/site-packages/diffusers/models/unet_2d_blocks.py:119, in get_down_block(down_block_type, num_layers, in_channels, out_channels, temb_channels, add_downsample, resnet_eps, resnet_act_fn, transformer_layers_per_block, num_attention_heads, resnet_groups, cross_attention_dim, downsample_padding, dual_cross_attention, use_linear_projection, only_cross_attention, upcast_attention, resnet_time_scale_shift, attention_type, resnet_skip_time_act, resnet_out_scale_factor, cross_attention_norm, attention_head_dim, downsample_type, dropout)
    117     if cross_attention_dim is None:
    118         raise ValueError("cross_attention_dim must be specified for CrossAttnDownBlock2D")
--> 119     return CrossAttnDownBlock2D(
    120         num_layers=num_layers,
    121         transformer_layers_per_block=transformer_layers_per_block,
    122         in_channels=in_channels,
    123         out_channels=out_channels,
    124         temb_channels=temb_channels,
    125         dropout=dropout,
    126         add_downsample=add_downsample,
    127         resnet_eps=resnet_eps,
    128         resnet_act_fn=resnet_act_fn,
    129         resnet_groups=resnet_groups,
    130         downsample_padding=downsample_padding,
    131         cross_attention_dim=cross_attention_dim,
    132         num_attention_heads=num_attention_heads,
    133         dual_cross_attention=dual_cross_attention,
    134         use_linear_projection=use_linear_projection,
    135         only_cross_attention=only_cross_attention,
    136         upcast_attention=upcast_attention,
    137         resnet_time_scale_shift=resnet_time_scale_shift,
    138         attention_type=attention_type,
    139     )
    140 elif down_block_type == "SimpleCrossAttnDownBlock2D":
    141     if cross_attention_dim is None:

File /opt/conda/lib/python3.10/site-packages/diffusers/models/unet_2d_blocks.py:1001, in CrossAttnDownBlock2D.__init__(self, in_channels, out_channels, temb_channels, dropout, num_layers, transformer_layers_per_block, resnet_eps, resnet_time_scale_shift, resnet_act_fn, resnet_groups, resnet_pre_norm, num_attention_heads, cross_attention_dim, output_scale_factor, downsample_padding, add_downsample, dual_cross_attention, use_linear_projection, only_cross_attention, upcast_attention, attention_type)
    985 resnets.append(
    986     ResnetBlock2D(
    987         in_channels=in_channels,
   (...)
    997     )
    998 )
    999 if not dual_cross_attention:
   1000     attentions.append(
-> 1001         Transformer2DModel(
   1002             num_attention_heads,
   1003             out_channels // num_attention_heads,
   1004             in_channels=out_channels,
   1005             num_layers=transformer_layers_per_block,
   1006             cross_attention_dim=cross_attention_dim,
   1007             norm_num_groups=resnet_groups,
   1008             use_linear_projection=use_linear_projection,
   1009             only_cross_attention=only_cross_attention,
   1010             upcast_attention=upcast_attention,
   1011             attention_type=attention_type,
   1012         )
   1013     )
   1014 else:
   1015     attentions.append(
   1016         DualTransformer2DModel(
   1017             num_attention_heads,
   (...)
   1023         )
   1024     )

File /opt/conda/lib/python3.10/site-packages/diffusers/configuration_utils.py:636, in register_to_config.<locals>.inner_init(self, *args, **kwargs)
    634 new_kwargs = {**config_init_kwargs, **new_kwargs}
    635 getattr(self, "register_to_config")(**new_kwargs)
--> 636 init(self, *args, **init_kwargs)

File /opt/conda/lib/python3.10/site-packages/diffusers/models/transformer_2d.py:191, in Transformer2DModel.__init__(self, num_attention_heads, attention_head_dim, in_channels, out_channels, num_layers, dropout, norm_num_groups, cross_attention_dim, attention_bias, sample_size, num_vector_embeds, patch_size, activation_fn, num_embeds_ada_norm, use_linear_projection, only_cross_attention, double_self_attention, upcast_attention, norm_type, norm_elementwise_affine, attention_type)
    164     self.pos_embed = PatchEmbed(
    165         height=sample_size,
    166         width=sample_size,
   (...)
    169         embed_dim=inner_dim,
    170     )
    172 # 3. Define transformers blocks
    173 self.transformer_blocks = nn.ModuleList(
    174     [
    175         BasicTransformerBlock(
    176             inner_dim,
    177             num_attention_heads,
    178             attention_head_dim,
    179             dropout=dropout,
    180             cross_attention_dim=cross_attention_dim,
    181             activation_fn=activation_fn,
    182             num_embeds_ada_norm=num_embeds_ada_norm,
    183             attention_bias=attention_bias,
    184             only_cross_attention=only_cross_attention,
    185             double_self_attention=double_self_attention,
    186             upcast_attention=upcast_attention,
    187             norm_type=norm_type,
    188             norm_elementwise_affine=norm_elementwise_affine,
    189             attention_type=attention_type,
    190         )
--> 191         for d in range(num_layers)
    192     ]
    193 )
    195 # 4. Define output layers
    196 self.out_channels = in_channels if out_channels is None else out_channels

TypeError: 'list' object cannot be interpreted as an integer