onnx-ruby 0.1.0 → 0.1.1

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 809c3c981b4ec890ed7169969b64f556dffb4a44a0c44e6126ec3f2ba28f2f07
-  data.tar.gz: 91233f1ee8c921c5cf4a2352f98e0f322aac91b27b0db01c55212c0ae8f9a750
+  metadata.gz: ec86be6d12455b53b31aeb0b680db0239f0fc6d2b0d93a57efcacf847728af43
+  data.tar.gz: e228293936cffd177c42055a5dcf34536ab1c30670797d231bc3a9885eba0a8e
 SHA512:
-  metadata.gz: b1ad99c83090be72d099836e84992d1a108b876272fd152fab1de4b43569930b1bdf985ca79e5fc8b27e3a4a8da65b33d823a560b9dd6cd4d7df384c942adb0a
-  data.tar.gz: f9a9c2b4278f7c20dffd890ece27cecf24004127944826b65bcef7a31a383c5f3799cecb01ca29e2ffd984f92698cb1ecf50e3681df2147b104c24839b9021df
+  metadata.gz: 609dc877ac507261a573b4284d249e779f07d6e45661ee894e29dd1c65e906cd3d89349f13d1931ab69c8a8f71765fc0c698e95bd59a114c9cef7db3a86ee39f
+  data.tar.gz: 86cd66b410257e644fad7a4e2148912fba30a03a1aa39e9112c5c087903fc311f8d7626416d52ccb5810380bf6316ac88ba2fb4f4a856791e0315e548f5e947b

data/CLAUDE.md

@@ -314,7 +314,7 @@ torch.onnx.export(model, dummy, "test/models/simple.onnx",
 
 ## Publishing
 
-- RubyGems.org: `gem push onnx-ruby-*.gem`
+- RubyGems.org: `GEM_HOST_API_KEY=rubygems_5d46e91ceb51fb455e98a7f491a2321bb6879f9be35d6842 gem push onnx-ruby-*.gem`
 - gem.coop: `GEM_HOST_API_KEY=hjncPswY8PbGDfLPw4RMj928 gem push onnx-ruby-*.gem --host https://beta.gem.coop/@johannesdwicahyo`
 
 ## Notes from zvec-ruby Experience

data/lib/onnx_ruby/classifier.rb

@@ -2,6 +2,8 @@
 
 module OnnxRuby
   class Classifier
+    include TokenizerSupport
+
     attr_reader :session, :labels
 
     def initialize(model_path, tokenizer: nil, labels: nil, **session_opts)
@@ -31,23 +33,9 @@ module OnnxRuby
 
     private
 
-    def resolve_tokenizer(tokenizer)
-      return nil if tokenizer.nil?
-
-      if tokenizer.respond_to?(:encode)
-        tokenizer
-      else
-        begin
-          require "tokenizers"
-          Tokenizers::Tokenizer.from_pretrained(tokenizer.to_s)
-        rescue LoadError
-          raise Error, "tokenizer-ruby gem is required for text tokenization. " \
-                       "Install with: gem install tokenizers"
-        end
-      end
-    end
-
     def prepare_inputs(inputs)
+      return [] if inputs.empty?
+
       if inputs.first.is_a?(String)
         raise Error, "tokenizer is required for text inputs" unless @tokenizer
 

data/lib/onnx_ruby/configuration.rb

@@ -2,8 +2,7 @@
 
 module OnnxRuby
   class Configuration
-    attr_accessor :models_path, :default_providers, :default_log_level,
-                  :pool_size, :pool_timeout
+    attr_accessor :models_path, :default_providers, :default_log_level
 
     def initialize
       @models_path = "app/models/onnx"
@@ -12,5 +11,25 @@ module OnnxRuby
       @pool_size = 5
       @pool_timeout = 5
     end
+
+    def pool_size
+      @pool_size
+    end
+
+    def pool_size=(value)
+      raise ArgumentError, "pool_size must be a positive Integer" unless value.is_a?(Integer) && value > 0
+
+      @pool_size = value
+    end
+
+    def pool_timeout
+      @pool_timeout
+    end
+
+    def pool_timeout=(value)
+      raise ArgumentError, "pool_timeout must be a positive Numeric" unless value.is_a?(Numeric) && value > 0
+
+      @pool_timeout = value
+    end
   end
 end

data/lib/onnx_ruby/embedder.rb

@@ -2,6 +2,8 @@
 
 module OnnxRuby
   class Embedder
+    include TokenizerSupport
+
     attr_reader :session
 
     def initialize(model_path, tokenizer: nil, normalize: true, **session_opts)
@@ -26,6 +28,7 @@ module OnnxRuby
       result = @session.run(feed)
 
       raw = find_output(result, %w[embeddings sentence_embedding output last_hidden_state])
+      return [] if raw.nil? || raw.empty?
 
       # If output is 3D (batch, seq_len, dim) — do mean pooling
       embeddings = if raw.first.is_a?(Array) && raw.first.first.is_a?(Array)
@@ -39,22 +42,6 @@ module OnnxRuby
 
     private
 
-    def resolve_tokenizer(tokenizer)
-      return nil if tokenizer.nil?
-
-      if tokenizer.respond_to?(:encode) || tokenizer.respond_to?(:encode_batch)
-        tokenizer
-      else
-        begin
-          require "tokenizers"
-          Tokenizers::Tokenizer.from_pretrained(tokenizer.to_s)
-        rescue LoadError
-          raise Error, "tokenizer-ruby gem is required for text tokenization. " \
-                       "Install with: gem install tokenizers"
-        end
-      end
-    end
-
     def prepare_inputs(inputs)
       if inputs.first.is_a?(String)
         raise Error, "tokenizer is required for text inputs" unless @tokenizer
@@ -100,6 +87,8 @@ module OnnxRuby
 
     def build_feed(ids, masks)
       input_names = @session.inputs.map { |i| i[:name] }
+      raise OnnxRuby::Error, "Model has no input names" if input_names.empty?
+
       feed = {}
       feed[input_names.find { |n| n.include?("input_id") } || input_names[0]] = ids
       mask_name = input_names.find { |n| n.include?("mask") || n.include?("attention") }
@@ -119,6 +108,8 @@ module OnnxRuby
     # Mean pooling over token embeddings, masked by attention_mask
     def mean_pool(hidden_states, masks)
       hidden_states.each_with_index.map do |tokens, batch_idx|
+        return [] if tokens.nil? || tokens.empty? || tokens.first.nil?
+
         mask = masks && masks[batch_idx]
         dim = tokens.first.length
         sum = Array.new(dim, 0.0)

data/lib/onnx_ruby/lazy_session.rb

@@ -28,8 +28,6 @@ module OnnxRuby
     private
 
     def load_session
-      return @session if @session
-
       @mutex.synchronize do
         @session ||= Session.new(@model_path, **@opts)
       end

data/lib/onnx_ruby/reranker.rb

@@ -2,6 +2,8 @@
 
 module OnnxRuby
   class Reranker
+    include TokenizerSupport
+
     attr_reader :session
 
     def initialize(model_path, tokenizer: nil, **session_opts)
@@ -37,22 +39,6 @@ module OnnxRuby
 
     private
 
-    def resolve_tokenizer(tokenizer)
-      return nil if tokenizer.nil?
-
-      if tokenizer.respond_to?(:encode)
-        tokenizer
-      else
-        begin
-          require "tokenizers"
-          Tokenizers::Tokenizer.from_pretrained(tokenizer.to_s)
-        rescue LoadError
-          raise Error, "tokenizer-ruby gem is required for text tokenization. " \
-                       "Install with: gem install tokenizers"
-        end
-      end
-    end
-
     def score_pairs(pairs)
       if @tokenizer.respond_to?(:encode_batch)
         encodings = @tokenizer.encode_batch(pairs)

data/lib/onnx_ruby/tokenizer_support.rb

@@ -0,0 +1,22 @@
+# frozen_string_literal: true
+
+module OnnxRuby
+  module TokenizerSupport
+    private
+
+    def resolve_tokenizer(tokenizer)
+      case tokenizer
+      when String
+        require "tokenizers"
+        Tokenizers::Tokenizer.from_pretrained(tokenizer.to_s)
+      when nil
+        nil
+      else
+        tokenizer
+      end
+    rescue LoadError
+      raise OnnxRuby::Error,
+        "tokenizers gem required for text inputs. Install: gem install tokenizers"
+    end
+  end
+end

data/lib/onnx_ruby/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module OnnxRuby
-  VERSION = "0.1.0"
+  VERSION = "0.1.1"
 end

data/lib/onnx_ruby.rb

@@ -12,6 +12,7 @@ end
 require_relative "onnx_ruby/onnx_ruby_ext"
 require_relative "onnx_ruby/tensor"
 require_relative "onnx_ruby/session"
+require_relative "onnx_ruby/tokenizer_support"
 require_relative "onnx_ruby/embedder"
 require_relative "onnx_ruby/classifier"
 require_relative "onnx_ruby/reranker"

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: onnx-ruby
 version: !ruby/object:Gem::Version
-  version: 0.1.0
+  version: 0.1.1
 platform: ruby
 authors:
 - Johannes Dwi Cahyo
@@ -96,6 +96,7 @@ files:
 - lib/onnx_ruby/session.rb
 - lib/onnx_ruby/session_pool.rb
 - lib/onnx_ruby/tensor.rb
+- lib/onnx_ruby/tokenizer_support.rb
 - lib/onnx_ruby/version.rb
 - onnx-ruby.gemspec
 homepage: https://github.com/johannesdwicahyo/onnx-ruby