stability_sdk 0.3.2 → 0.3.3

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 4bfbab3c7b0c366f01681d5adf54193d9473741c8616d3417628011f87e60bea
-  data.tar.gz: f30748cc3ddefc1250b286f695cf039d8297970ef8757720620b9cec9d60e0cc
+  metadata.gz: 1bb93f79d9c09847cc8ee382f92cdf019137dbaa8b4c762bd1e24de392551a0f
+  data.tar.gz: a06818c77b6d490dec40047cf08b6b17e5ca72e09d8319499781dab3ef77a820
 SHA512:
-  metadata.gz: 7f1eab7a664df5a09cd7785f9636e7e7542b87ed3ce475ceb5dd40005ec6ddc2e246346e258225d2e6f04b3472f28a316c91a32b3aa3e069acf031df5db20b82
-  data.tar.gz: c74dffd4dcbffb5d3211d6bc4762d76384badef93c85e33432c93bf7c3b4c1f0eab2b934c8af0f4065f791e1fd3204408cdc50a5c16037c3cc26b44ef6bc449e
+  metadata.gz: 7ce00a0dcd937cdf0dbeb83b54d1437b5d26912b7139a99d125171f1364d3a03a377bef54698a9371b7b5dd44d6ad432bccfd01de9b02c3b7def1b01fec726f5
+  data.tar.gz: 84aa5aa534045145c9a4d6160b42e5cc2cdc5b41559e0c7122739f4720a00b36d757da57bab9cbef1b7936250e1a427b91014fa05b316ea45a33ade836c70261

data/README.md

@@ -50,13 +50,13 @@ Options:
     -H, --height=VAL                 height of image in pixel. default 512
     -W, --width=VAL                  width of image in pixel. default 512
     -C, --cfg_scale=VAL              CFG scale factor. default 7.0
-    -A, --sampler=VAL                ddim, plms, k_euler, k_euler_ancestral, k_heun, k_dpm_2, k_dpm_2_ancestral, k_lms. default k_lms
-    -s, --steps=VAL                  number of steps. default 50
+    -A, --sampler=VAL                ddim, plms, k_euler, k_euler_ancestral, k_heun, k_dpm_2, k_dpm_2_ancestral, k_lms, k_dpmpp_2s_ancestral, k_dpmpp_2m, k_dpmpp_sde. default auto-select
+    -s, --steps=VAL                  number of steps. default auto-select
     -S, --seed=VAL                   random seed to use in integer
     -p, --prefix=VAL                 output prefixes for artifacts. default `generation`
         --no-store                   do not write out artifacts
-    -n, --num_samples=VAL            number of samples to generate
-    -e, --engine=VAL                 engine to use for inference. default `stable-diffusion-v1`
+    -n, --num_samples=VAL            number of samples to generate. default 1
+    -e, --engine=VAL                 engine to use for inference. default `stable-diffusion-v1-5`
     -i, --init_image=VAL             path to init image
     -m, --mask_image=VAL             path to mask image
         --start_schedule=VAL         start schedule for init image (must be greater than 0, 1 is full strength text prompt, no trace of image). default 1.0

data/exe/stability-client

@@ -21,8 +21,8 @@ opt.on("--api_key=VAL", "api key of DreamStudio account. You can also specify by
 opt.on("-H", "--height=VAL", Integer, "height of image in pixel. default 512") {|v| options[:height] = v }
 opt.on("-W", "--width=VAL", Integer, "width of image in pixel. default 512") {|v| options[:width] = v }
 opt.on("-C", "--cfg_scale=VAL", Float, "CFG scale factor. default 7.0") {|v| options[:cfg_scale] = v }
-opt.on("-A", "--sampler=VAL", String, "ddim, plms, k_euler, k_euler_ancestral, k_heun, k_dpm_2, k_dpm_2_ancestral, k_lms. default k_lms") {|v| options[:sampler] = v }
-opt.on("-s", "--steps=VAL", Integer, "number of steps. default 50") {|v| options[:steps] = v }
+opt.on("-A", "--sampler=VAL", String, "ddim, plms, k_euler, k_euler_ancestral, k_heun, k_dpm_2, k_dpm_2_ancestral, k_lms, k_dpmpp_2s_ancestral, k_dpmpp_2m, k_dpmpp_sde. default auto-select") {|v| options[:sampler] = v }
+opt.on("-s", "--steps=VAL", Integer, "number of steps. default auto-select") {|v| options[:steps] = v }
 opt.on("-S", "--seed=VAL", Integer, "random seed to use in integer") {|v| options[:seed] = v }
 opt.on("-p", "--prefix=VAL", String, "output prefixes for artifacts. default `generation`") {|v| options[:prefix] = v }
 opt.on("--no-store", "do not write out artifacts") {|v| options[:no_store] = v }
@@ -37,6 +37,7 @@ opt.on("--guidance_cuts=VAL", Integer, "Number of cuts to use for guidance. defa
 opt.on("--guidance_strength=VAL", Float, "Strength of the guidance. We recommend values in range [0.0,1.0]. A good default is 0.25. default nil") {|v| options[:guidance_strength] = v }
 opt.on("--guidance_prompt=VAL", String, "Prompt to use for guidance, defaults to `YOUR_PROMPT_TEXT` argument (above) if not specified.") {|v| options[:guidance_prompt] = v }
 opt.on("--guidance_models=VAL", Array, "Models to use for guidance. default nil") {|v| options[:guidance_models] = v }
+opt.on("-t", "--artifact_types=VAL", Array, "filter artifacts by type (ARTIFACT_IMAGE, ARTIFACT_TEXT, ARTIFACT_CLASSIFICATIONS, etc)") {|v| options[:artifact_types] = v }
 opt.on("-v", "--verbose") { logger.level = Logger::DEBUG }
 opt.parse!(ARGV)
 

data/lib/generation_pb.rb

@@ -3,6 +3,7 @@
 
 require 'google/protobuf'
 
+require 'google/protobuf/struct_pb'
 require 'tensors_pb'
 
 Google::Protobuf::DescriptorPool.generated_pool.build do
@@ -52,6 +53,7 @@ Google::Protobuf::DescriptorPool.generated_pool.build do
       proto3_optional :downsampling_factor, :uint64, 4
       proto3_optional :cfg_scale, :float, 5
       proto3_optional :init_noise_scale, :float, 6
+      proto3_optional :step_noise_scale, :float, 7
     end
     add_message "gooseai.ConditionerParameters" do
       proto3_optional :vector_adjust_prior, :string, 1
@@ -114,6 +116,7 @@ Google::Protobuf::DescriptorPool.generated_pool.build do
       repeated :parameters, :message, 7, "gooseai.StepParameter"
       proto3_optional :masked_area_init, :enum, 8, "gooseai.MaskedAreaInit"
       proto3_optional :weight_method, :enum, 9, "gooseai.WeightMethod"
+      proto3_optional :quantize, :bool, 10
     end
     add_message "gooseai.ClassifierConcept" do
       optional :concept, :string, 1
@@ -131,6 +134,49 @@ Google::Protobuf::DescriptorPool.generated_pool.build do
       repeated :exceeds, :message, 2, "gooseai.ClassifierCategory"
       proto3_optional :realized_action, :enum, 3, "gooseai.Action"
     end
+    add_message "gooseai.InterpolateParameters" do
+      repeated :ratios, :float, 1
+      proto3_optional :mode, :enum, 2, "gooseai.InterpolateMode"
+    end
+    add_message "gooseai.TransformBlend" do
+      optional :amount, :float, 1
+      optional :target, :message, 2, "gooseai.Artifact"
+    end
+    add_message "gooseai.TransformColorAdjust" do
+      proto3_optional :brightness, :float, 1
+      proto3_optional :contrast, :float, 2
+      proto3_optional :hue, :float, 3
+      proto3_optional :saturation, :float, 4
+      proto3_optional :lightness, :float, 5
+    end
+    add_message "gooseai.TransformColorMatch" do
+      optional :color_mode, :enum, 1, "gooseai.ColorMatchMode"
+      optional :image, :message, 2, "gooseai.Artifact"
+    end
+    add_message "gooseai.TransformDepthCalc" do
+      proto3_optional :blend_weight, :float, 1
+      proto3_optional :blur_radius, :uint32, 2
+      proto3_optional :reverse, :bool, 3
+    end
+    add_message "gooseai.TransformMatrix" do
+      repeated :data, :float, 1
+    end
+    add_message "gooseai.TransformResample" do
+      optional :border_mode, :enum, 1, "gooseai.BorderMode"
+      optional :transform, :message, 2, "gooseai.TransformMatrix"
+      proto3_optional :prev_transform, :message, 3, "gooseai.TransformMatrix"
+      proto3_optional :depth_warp, :float, 4
+      proto3_optional :export_mask, :bool, 5
+    end
+    add_message "gooseai.TransformParameters" do
+      oneof :transform do
+        optional :blend, :message, 1, "gooseai.TransformBlend"
+        optional :color_adjust, :message, 2, "gooseai.TransformColorAdjust"
+        optional :color_match, :message, 3, "gooseai.TransformColorMatch"
+        optional :depth_calc, :message, 4, "gooseai.TransformDepthCalc"
+        optional :resample, :message, 5, "gooseai.TransformResample"
+      end
+    end
     add_message "gooseai.AssetParameters" do
       optional :action, :enum, 1, "gooseai.AssetAction"
       optional :project_id, :string, 2
@@ -156,10 +202,13 @@ Google::Protobuf::DescriptorPool.generated_pool.build do
       optional :requested_type, :enum, 3, "gooseai.ArtifactType"
       repeated :prompt, :message, 4, "gooseai.Prompt"
       proto3_optional :conditioner, :message, 6, "gooseai.ConditionerParameters"
+      proto3_optional :extras, :message, 2047, "google.protobuf.Struct"
       oneof :params do
         optional :image, :message, 5, "gooseai.ImageParameters"
         optional :classifier, :message, 7, "gooseai.ClassifierParameters"
         optional :asset, :message, 8, "gooseai.AssetParameters"
+        optional :interpolate, :message, 11, "gooseai.InterpolateParameters"
+        optional :transform, :message, 12, "gooseai.TransformParameters"
       end
     end
     add_message "gooseai.OnStatus" do
@@ -194,6 +243,7 @@ Google::Protobuf::DescriptorPool.generated_pool.build do
       value :ARTIFACT_MASK, 7
       value :ARTIFACT_LATENT, 8
       value :ARTIFACT_TENSOR, 9
+      value :ARTIFACT_DEPTH, 10
     end
     add_enum "gooseai.MaskedAreaInit" do
       value :MASKED_AREA_INIT_ZERO, 0
@@ -215,6 +265,7 @@ Google::Protobuf::DescriptorPool.generated_pool.build do
       value :SAMPLER_K_LMS, 7
       value :SAMPLER_K_DPMPP_2S_ANCESTRAL, 8
       value :SAMPLER_K_DPMPP_2M, 9
+      value :SAMPLER_K_DPMPP_SDE, 10
     end
     add_enum "gooseai.Upscaler" do
       value :UPSCALER_RGB, 0
@@ -248,6 +299,23 @@ Google::Protobuf::DescriptorPool.generated_pool.build do
       value :CLSFR_MODE_ZEROSHOT, 0
       value :CLSFR_MODE_MULTICLASS, 1
     end
+    add_enum "gooseai.InterpolateMode" do
+      value :INTERPOLATE_LINEAR, 0
+      value :INTERPOLATE_RIFE, 1
+      value :INTERPOLATE_VAE_LINEAR, 2
+      value :INTERPOLATE_VAE_SLERP, 3
+    end
+    add_enum "gooseai.BorderMode" do
+      value :BORDER_REFLECT, 0
+      value :BORDER_REPLICATE, 1
+      value :BORDER_WRAP, 2
+      value :BORDER_ZERO, 3
+    end
+    add_enum "gooseai.ColorMatchMode" do
+      value :COLOR_MATCH_HSV, 0
+      value :COLOR_MATCH_LAB, 1
+      value :COLOR_MATCH_RGB, 2
+    end
     add_enum "gooseai.AssetAction" do
       value :ASSET_PUT, 0
       value :ASSET_GET, 1
@@ -288,6 +356,14 @@ module Gooseai
   ClassifierConcept = ::Google::Protobuf::DescriptorPool.generated_pool.lookup("gooseai.ClassifierConcept").msgclass
   ClassifierCategory = ::Google::Protobuf::DescriptorPool.generated_pool.lookup("gooseai.ClassifierCategory").msgclass
   ClassifierParameters = ::Google::Protobuf::DescriptorPool.generated_pool.lookup("gooseai.ClassifierParameters").msgclass
+  InterpolateParameters = ::Google::Protobuf::DescriptorPool.generated_pool.lookup("gooseai.InterpolateParameters").msgclass
+  TransformBlend = ::Google::Protobuf::DescriptorPool.generated_pool.lookup("gooseai.TransformBlend").msgclass
+  TransformColorAdjust = ::Google::Protobuf::DescriptorPool.generated_pool.lookup("gooseai.TransformColorAdjust").msgclass
+  TransformColorMatch = ::Google::Protobuf::DescriptorPool.generated_pool.lookup("gooseai.TransformColorMatch").msgclass
+  TransformDepthCalc = ::Google::Protobuf::DescriptorPool.generated_pool.lookup("gooseai.TransformDepthCalc").msgclass
+  TransformMatrix = ::Google::Protobuf::DescriptorPool.generated_pool.lookup("gooseai.TransformMatrix").msgclass
+  TransformResample = ::Google::Protobuf::DescriptorPool.generated_pool.lookup("gooseai.TransformResample").msgclass
+  TransformParameters = ::Google::Protobuf::DescriptorPool.generated_pool.lookup("gooseai.TransformParameters").msgclass
   AssetParameters = ::Google::Protobuf::DescriptorPool.generated_pool.lookup("gooseai.AssetParameters").msgclass
   AnswerMeta = ::Google::Protobuf::DescriptorPool.generated_pool.lookup("gooseai.AnswerMeta").msgclass
   Answer = ::Google::Protobuf::DescriptorPool.generated_pool.lookup("gooseai.Answer").msgclass
@@ -305,6 +381,9 @@ module Gooseai
   ModelArchitecture = ::Google::Protobuf::DescriptorPool.generated_pool.lookup("gooseai.ModelArchitecture").enummodule
   Action = ::Google::Protobuf::DescriptorPool.generated_pool.lookup("gooseai.Action").enummodule
   ClassifierMode = ::Google::Protobuf::DescriptorPool.generated_pool.lookup("gooseai.ClassifierMode").enummodule
+  InterpolateMode = ::Google::Protobuf::DescriptorPool.generated_pool.lookup("gooseai.InterpolateMode").enummodule
+  BorderMode = ::Google::Protobuf::DescriptorPool.generated_pool.lookup("gooseai.BorderMode").enummodule
+  ColorMatchMode = ::Google::Protobuf::DescriptorPool.generated_pool.lookup("gooseai.ColorMatchMode").enummodule
   AssetAction = ::Google::Protobuf::DescriptorPool.generated_pool.lookup("gooseai.AssetAction").enummodule
   AssetUse = ::Google::Protobuf::DescriptorPool.generated_pool.lookup("gooseai.AssetUse").enummodule
   StageAction = ::Google::Protobuf::DescriptorPool.generated_pool.lookup("gooseai.StageAction").enummodule

data/lib/stability_sdk/cli.rb

@@ -4,12 +4,22 @@ module StabilitySDK
       logger.debug "received answer: #{answer}"
       return if options[:no_store]
 
+      allowed_artifact_types = []
+      if options.has_key?(:artifact_types)
+        allowed_artifact_types = options[:artifact_types].map(&:to_sym)
+      end
+
       answer.artifacts.each_with_index do |artifact, idx|
         filename_base = "#{options[:prefix] || "generation"}-#{answer.request_id}-#{answer.answer_id}-#{idx}"
 
         filename = ""
         contents = ""
 
+        if allowed_artifact_types.size > 0 && !allowed_artifact_types.include?(artifact.type)
+          logger.info "skipping #{artifact.type} due to artifact type filter"
+          next
+        end
+
         case artifact.type
         when :ARTIFACT_IMAGE
           ext = MIME::Types[artifact.mime].first.preferred_extension

data/lib/stability_sdk/client.rb

@@ -8,10 +8,8 @@ module StabilitySDK
     DEFAULT_IMAGE_WIDTH = 512
     DEFAULT_IMAGE_HEIGHT = 512
     DEFAULT_SAMPLE_SIZE = 1
-    DEFAULT_STEPS = 50
     DEFAULT_ENGINE_ID = "stable-diffusion-v1-5"
     DEFAULT_CFG_SCALE = 7.0
-    DEFAULT_SAMPLER_ALGORITHM = Gooseai::DiffusionSampler::SAMPLER_K_LMS
     DEFAULT_START_SCHEDULE = 1.0
     DEFAULT_END_SCHEDULE = 0.01
 
@@ -24,6 +22,9 @@ module StabilitySDK
       "k_dpm_2": Gooseai::DiffusionSampler::SAMPLER_K_DPM_2,
       "k_dpm_2_ancestral": Gooseai::DiffusionSampler::SAMPLER_K_DPM_2_ANCESTRAL,
       "k_lms": Gooseai::DiffusionSampler::SAMPLER_K_LMS,
+      "k_dpmpp_2s_ancestral": Gooseai::DiffusionSampler::SAMPLER_K_DPMPP_2S_ANCESTRAL,
+      "k_dpmpp_2m": Gooseai::DiffusionSampler::SAMPLER_K_DPMPP_2M,
+      "k_dpmpp_sde": Gooseai::DiffusionSampler::SAMPLER_K_DPMPP_SDE,
     }
 
     def initialize(options={})
@@ -57,10 +58,10 @@ module StabilitySDK
       end
 
       samples = options.has_key?(:num_samples) ? options[:num_samples] : DEFAULT_SAMPLE_SIZE
-      steps = options.has_key?(:steps) ? options[:steps] : DEFAULT_STEPS
+      steps = options.has_key?(:steps) ? options[:steps] : nil
       seed = options.has_key?(:seed) ? [options[:seed]] : [rand(4294967295)]
       transform = Gooseai::TransformType.new(
-        diffusion: options.has_key?(:sampler) ? SAMPLER_ALGORITHMS[options[:sampler].to_sym] : DEFAULT_SAMPLER_ALGORITHM,
+        diffusion: options.has_key?(:sampler) ? SAMPLER_ALGORITHMS[options[:sampler].to_sym] : nil,
       )
       step_parameter = Gooseai::StepParameter.new(
         scaled_step: 0,

data/lib/stability_sdk/version.rb

@@ -1,3 +1,3 @@
 module StabilitySDK
-  VERSION = "0.3.2"
+  VERSION = "0.3.3"
 end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: stability_sdk
 version: !ruby/object:Gem::Version
-  version: 0.3.2
+  version: 0.3.3
 platform: ruby
 authors:
 - Kosei Moriyama
 autorequire: 
 bindir: exe
 cert_chain: []
-date: 2022-12-02 00:00:00.000000000 Z
+date: 2023-02-10 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: grpc
@@ -140,7 +140,7 @@ required_rubygems_version: !ruby/object:Gem::Requirement
     - !ruby/object:Gem::Version
       version: '0'
 requirements: []
-rubygems_version: 3.1.6
+rubygems_version: 3.1.4
 signing_key: 
 specification_version: 4
 summary: Ruby client for interacting with stability.ai APIs (e.g. stable diffusion