informers 0.1.3 → 0.2.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 2baf3ed7ae9b6bf6a1347f0dc880ae3a48f26daa518112e37d6bf03927faed67
-  data.tar.gz: 03cd4f92aa6a062fc23ca712369a8cf1db5300bb53b1eb99ad8d71574a1a8ce6
+  metadata.gz: 22f7bcebf0670078b65fdf9cba4d2b937c853a3b10cf36e47f50781e2663225c
+  data.tar.gz: 940c96ec6b749b7e0b0c283456e40bfe9e6cbb3a58e8fa11f6367e87b05d8694
 SHA512:
-  metadata.gz: cfef17a6c7b9a574c43f3f45cc4f20bb36c1d764f6c68f47036f41a7af9a54aecf1a678eda1e3f3f7b0da26ff8131e22dc13d56e16e461366b67d8b6b0d77e97
-  data.tar.gz: 8c99136eb43350c118402e0ac076055d4ab563e6f185d06c9b826c5d592a3955f64b2ea5284d32583e4725229201514a30527401f454cde45944fb54f9dd0b97
+  metadata.gz: 4cd8b58aae6e885409e297bc1ba09aedd029bb3dc26a193251f33c2bf6c9f6a8da69cb3727f799296a8c6644b014afc715e783a1e19a1074982af531e40db57b
+  data.tar.gz: 6f63489d0b303e9a7de13df11d5074bd4cb2dfa44febee4061262d5c188eeb62a7c975e89567048f801fa183c8d56925275768fccc9a4b5a48255abeeb379345

data/CHANGELOG.md

@@ -1,3 +1,8 @@
+## 0.2.0 (2022-09-06)
+
+- Added support for `optimum` and `transformers.onnx` models
+- Dropped support for Ruby < 2.7
+
 ## 0.1.3 (2021-09-25)
 
 - Added text generation

data/README.md

@@ -8,8 +8,6 @@ Supports:
 - Question answering
 - Named-entity recognition
 - Text generation
-- Summarization - *in development*
-- Translation - *in development*
 
 [![Build Status](https://github.com/ankane/informers/workflows/build/badge.svg?branch=master)](https://github.com/ankane/informers/actions)
 
@@ -18,7 +16,7 @@ Supports:
 Add this line to your application’s Gemfile:
 
 ```ruby
-gem 'informers'
+gem "informers"
 ```
 
 ## Getting Started
@@ -26,7 +24,9 @@ gem 'informers'
 - [Sentiment analysis](#sentiment-analysis)
 - [Question answering](#question-answering)
 - [Named-entity recognition](#named-entity-recognition)
-- [Text Generation](#text-generation)
+- [Text generation](#text-generation)
+- [Feature extraction](#feature-extraction)
+- [Fill mask](#fill-mask)
 
 ### Sentiment Analysis
 
@@ -109,6 +109,24 @@ This returns
 As far as I am concerned, I will be the first to admit that I am not a fan of the idea of a "free market." I think that the idea of a free market is a bit of a stretch. I think that the idea
 ```
 
+### Feature Extraction
+
+First, export a [pretrained model](tools/export.md).
+
+```ruby
+model = Informers::FeatureExtraction.new("feature-extraction.onnx")
+model.predict("This is super cool")
+```
+
+### Fill Mask
+
+First, export a [pretrained model](tools/export.md).
+
+```ruby
+model = Informers::FillMask.new("fill-mask.onnx")
+model.predict("This is a great <mask>")
+```
+
 ## Models
 
 Task | Description | Contributor | License | Link
@@ -116,7 +134,7 @@ Task | Description | Contributor | License | Link
 Sentiment analysis | DistilBERT fine-tuned on SST-2 | Hugging Face | Apache-2.0 | [Link](https://huggingface.co/distilbert-base-uncased-finetuned-sst-2-english)
 Question answering | DistilBERT fine-tuned on SQuAD | Hugging Face | Apache-2.0 | [Link](https://huggingface.co/distilbert-base-cased-distilled-squad)
 Named-entity recognition | BERT fine-tuned on CoNLL03 | Bayerische Staatsbibliothek | In-progress | [Link](https://huggingface.co/dbmdz/bert-large-cased-finetuned-conll03-english)
-Text generation | GPT-2 | Hugging Face | [Custom](https://github.com/openai/gpt-2/blob/master/LICENSE) | [Link](https://huggingface.co/gpt2)
+Text generation | GPT-2 | OpenAI | [Custom](https://github.com/openai/gpt-2/blob/master/LICENSE) | [Link](https://huggingface.co/gpt2)
 
 Some models are [quantized](https://medium.com/microsoftazure/faster-and-smaller-quantized-nlp-with-hugging-face-and-onnx-runtime-ec5525473bb7) to make them faster and smaller.
 

data/lib/informers/feature_extraction.rb

@@ -51,7 +51,7 @@ module Informers
         attention_mask: attention_mask
       }
       output = @model.predict(input)
-      scores = output["output_0"]
+      scores = output["output_0"] || output["last_hidden_state"]
 
       singular ? scores.first : scores
     end

data/lib/informers/fill_mask.rb

@@ -74,7 +74,8 @@ module Informers
         raise "More than one mask_token (<mask>) is not supported" if v.size > 1
       end
 
-      outputs = @model.predict(input)["output_0"]
+      res = @model.predict(input)
+      outputs = res["output_0"] || res["logits"]
       batch_size = outputs.size
 
       results = []

data/lib/informers/ner.rb

@@ -38,12 +38,12 @@ module Informers
           attention_mask: [[1] * tokens.size],
           token_type_ids: [[0] * tokens.size]
         }
-        output = @model.predict(input)
+        res = @model.predict(input)
 
         # transform
-        entities = output["output_0"][0]
+        output = res["output_0"] || res["logits"]
         score =
-          entities.map do |e|
+          output[0].map do |e|
             values = e.map { |v| Math.exp(v) }
             sum = values.sum
             values.map { |v| v / sum }

data/lib/informers/question_answering.rb

@@ -67,8 +67,8 @@ module Informers
       }
       output = @model.predict(input)
 
-      start = output["output_0"]
-      stop = output["output_1"]
+      start = output["output_0"] || output["start_logits"]
+      stop = output["output_1"] || output["end_logits"]
 
       # transform
       answers = []

data/lib/informers/sentiment_analysis.rb

@@ -50,11 +50,12 @@ module Informers
         input_ids: input_ids,
         attention_mask: attention_mask
       }
-      output = @model.predict(input)
+      res = @model.predict(input)
+      output = res["output_0"] || res["logits"]
 
       # transform
       scores =
-        output["output_0"].map do |row|
+        output.map do |row|
           mapped = row.map { |v| Math.exp(v) }
           sum = mapped.sum
           mapped.map { |v| v / sum }

data/lib/informers/text_generation.rb

@@ -31,11 +31,21 @@ module Informers
       input = {
         input_ids: [tokens]
       }
+      if @model.inputs.any? { |i| i[:name] == "attention_mask" }
+        input[:attention_mask] = [[1] * tokens.size]
+      end
+
+      output_name =
+        if @model.outputs.any? { |o| o[:name] == "output_0" }
+          "output_0"
+        else
+          "logits"
+        end
 
       (max_length - tokens.size).times do |i|
-        output = @model.predict(input, output_type: :numo, output_names: ["output_0"])
+        output = @model.predict(input, output_type: :numo, output_names: [output_name])
         # passed to input_ids
-        tokens << output["output_0"][0, true, true][-1, true].max_index
+        tokens << output[output_name][0, true, true][-1, true].max_index
       end
 
       @decoder.ids_to_text(tokens)

data/lib/informers/version.rb

@@ -1,3 +1,3 @@
 module Informers
-  VERSION = "0.1.3"
+  VERSION = "0.2.0"
 end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: informers
 version: !ruby/object:Gem::Version
-  version: 0.1.3
+  version: 0.2.0
 platform: ruby
 authors:
 - Andrew Kane
 autorequire:
 bindir: bin
 cert_chain: []
-date: 2021-09-25 00:00:00.000000000 Z
+date: 2022-09-06 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: blingfire
@@ -53,7 +53,7 @@ dependencies:
       - !ruby/object:Gem::Version
         version: 0.5.1
 description:
-email: andrew@chartkick.com
+email: andrew@ankane.org
 executables: []
 extensions: []
 extra_rdoc_files: []
@@ -91,14 +91,14 @@ required_ruby_version: !ruby/object:Gem::Requirement
   requirements:
   - - ">="
     - !ruby/object:Gem::Version
-      version: '2.5'
+      version: '2.7'
 required_rubygems_version: !ruby/object:Gem::Requirement
   requirements:
   - - ">="
     - !ruby/object:Gem::Version
       version: '0'
 requirements: []
-rubygems_version: 3.2.22
+rubygems_version: 3.3.7
 signing_key:
 specification_version: 4
 summary: State-of-the-art natural language processing for Ruby