Merge pull request #3 from 920232796/master

fix device support for mps
update the support for SD2.0

modules/devices.py

@@ -38,8 +38,8 @@ def get_optimal_device():
     if torch.cuda.is_available():
         return torch.device(get_cuda_device_string())
 
-    # if has_mps():
-    #     return torch.device("mps")
+    if has_mps():
+        return torch.device("mps")
 
     return cpu
 

modules/sd_hijack.py

@@ -29,7 +29,7 @@ diffusionmodules_model_AttnBlock_forward = ldm.modules.diffusionmodules.model.At
 # new memory efficient cross attention blocks do not support hypernets and we already
 # have memory efficient cross attention anyway, so this disables SD2.0's memory efficient cross attention
 ldm.modules.attention.MemoryEfficientCrossAttention = ldm.modules.attention.CrossAttention
-# ldm.modules.attention.BasicTransformerBlock.ATTENTION_MODES["softmax-xformers"] = ldm.modules.attention.CrossAttention
+ldm.modules.attention.BasicTransformerBlock.ATTENTION_MODES["softmax-xformers"] = ldm.modules.attention.CrossAttention
 
 # silence new console spam from SD2
 ldm.modules.attention.print = lambda *args: None

modules/shared.py

@@ -110,7 +110,11 @@ restricted_opts = {
 from omegaconf import OmegaConf
 config = OmegaConf.load(f"{cmd_opts.config}")
 # XLMR-Large
-text_model_name = config.model.params.cond_stage_config.params.name
+try:
+    text_model_name = config.model.params.cond_stage_config.params.name
+
+except :
+    text_model_name = "stable_diffusion"
 
 cmd_opts.disable_extension_access = (cmd_opts.share or cmd_opts.listen or cmd_opts.server_name) and not cmd_opts.enable_insecure_extension_access
 

v2-inference.yaml

@@ -0,0 +1,67 @@
+model:
+  base_learning_rate: 1.0e-4
+  target: ldm.models.diffusion.ddpm.LatentDiffusion
+  params:
+    linear_start: 0.00085
+    linear_end: 0.0120
+    num_timesteps_cond: 1
+    log_every_t: 200
+    timesteps: 1000
+    first_stage_key: "jpg"
+    cond_stage_key: "txt"
+    image_size: 64
+    channels: 4
+    cond_stage_trainable: false
+    conditioning_key: crossattn
+    monitor: val/loss_simple_ema
+    scale_factor: 0.18215
+    use_ema: False # we set this to false because this is an inference only config
+
+    unet_config:
+      target: ldm.modules.diffusionmodules.openaimodel.UNetModel
+      params:
+        use_checkpoint: True
+        use_fp16: True
+        image_size: 32 # unused
+        in_channels: 4
+        out_channels: 4
+        model_channels: 320
+        attention_resolutions: [ 4, 2, 1 ]
+        num_res_blocks: 2
+        channel_mult: [ 1, 2, 4, 4 ]
+        num_head_channels: 64 # need to fix for flash-attn
+        use_spatial_transformer: True
+        use_linear_in_transformer: True
+        transformer_depth: 1
+        context_dim: 1024
+        legacy: False
+
+    first_stage_config:
+      target: ldm.models.autoencoder.AutoencoderKL
+      params:
+        embed_dim: 4
+        monitor: val/rec_loss
+        ddconfig:
+          #attn_type: "vanilla-xformers"
+          double_z: true
+          z_channels: 4
+          resolution: 256
+          in_channels: 3
+          out_ch: 3
+          ch: 128
+          ch_mult:
+          - 1
+          - 2
+          - 4
+          - 4
+          num_res_blocks: 2
+          attn_resolutions: []
+          dropout: 0.0
+        lossconfig:
+          target: torch.nn.Identity
+
+    cond_stage_config:
+      target: ldm.modules.encoders.modules.FrozenOpenCLIPEmbedder
+      params:
+        freeze: True
+        layer: "penultimate"