tokenizers 0.2.1 → 0.2.2

data/Cargo.toml

@@ -3,8 +3,3 @@ members = ["ext/tokenizers"]
 
 [profile.release]
 strip = true
-
-[patch.crates-io]
-number_prefix = { git = "https://github.com/ankane/rust-number-prefix", branch = "license-file" }
-rb-sys-env = { git = "https://github.com/oxidize-rb/rb-sys" }
-tokenizers = { git = "https://github.com/huggingface/tokenizers" }

data/README.md

@@ -24,8 +24,8 @@ Encode
 
 ```ruby
 encoded = tokenizer.encode("I can feel the magic, can you?")
-encoded.ids
 encoded.tokens
+encoded.ids
 ```
 
 Decode

data/ext/tokenizers/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "tokenizers"
-version = "0.2.1"
+version = "0.2.2"
 license = "Apache-2.0"
 authors = ["Andrew Kane <andrew@ankane.org>"]
 edition = "2021"
@@ -13,6 +13,7 @@ crate-type = ["cdylib"]
 magnus = "0.4"
 
 [dependencies.tokenizers]
-version = "0.13.2"
+version = "0.13.2" # also update in from_pretrained.rb
+git = "https://github.com/huggingface/tokenizers"
 default-features = false
-features = ["progressbar", "http", "onig", "esaxx_fast"]
+features = ["progressbar", "onig", "esaxx_fast"]

data/ext/tokenizers/src/lib.rs

@@ -27,10 +27,7 @@ fn module() -> RModule {
 #[magnus::init]
 fn init() -> RbResult<()> {
     let module = module();
-    module.define_singleton_method(
-        "_from_pretrained",
-        function!(RbTokenizer::from_pretrained, 3),
-    )?;
+    module.define_singleton_method("from_file", function!(RbTokenizer::from_file, 1))?;
 
     let class = module.define_class("BPE", Default::default())?;
     class.define_singleton_method("new", function!(RbBPE::new, 2))?;

data/ext/tokenizers/src/tokenizer.rs

@@ -1,5 +1,5 @@
-use magnus::Module;
 use std::cell::RefCell;
+use std::path::PathBuf;
 use tk::tokenizer::Tokenizer;
 use tk::AddedToken;
 
@@ -8,7 +8,7 @@ use super::encoding::RbEncoding;
 use super::models::RbBPE;
 use super::normalizers::RbBertNormalizer;
 use super::pre_tokenizers::RbBertPreTokenizer;
-use super::{module, RbError, RbResult};
+use super::{RbError, RbResult};
 
 #[magnus::wrap(class = "Tokenizers::Tokenizer")]
 pub struct RbTokenizer {
@@ -22,22 +22,8 @@ impl RbTokenizer {
         }
     }
 
-    pub fn from_pretrained(
-        identifier: String,
-        revision: String,
-        auth_token: Option<String>,
-    ) -> RbResult<Self> {
-        let version = module().const_get("VERSION").unwrap();
-        let params = tk::FromPretrainedParameters {
-            revision,
-            auth_token,
-            user_agent: [("bindings", "Ruby".to_string()), ("version", version)]
-                .iter()
-                .map(|(k, v)| (k.to_string(), v.to_string()))
-                .collect(),
-        };
-
-        Tokenizer::from_pretrained(identifier, Some(params))
+    pub fn from_file(path: PathBuf) -> RbResult<Self> {
+        Tokenizer::from_file(path)
             .map(|v| RbTokenizer {
                 tokenizer: RefCell::new(v),
             })

data/lib/tokenizers/from_pretrained.rb

@@ -0,0 +1,119 @@
+module Tokenizers
+  module FromPretrained
+    # for user agent
+    TOKENIZERS_VERSION = "0.13.2"
+
+    # use Ruby for downloads
+    # this avoids the need to vendor OpenSSL on Linux
+    # and reduces the extension size by about half
+    def from_pretrained(identifier, revision: "main", auth_token: nil)
+      require "cgi"
+      require "digest"
+      require "fileutils"
+      require "json"
+      require "open-uri"
+
+      cache_dir = ensure_cache_dir
+
+      # string options are headers
+      options = {
+        open_timeout: 3,
+        read_timeout: 30,
+        "User-Agent" => "tokenizers/#{TOKENIZERS_VERSION}; bindings/Ruby; version/#{VERSION}"
+      }
+      if auth_token
+        options["Authorization"] = "Bearer #{auth_token}"
+      end
+
+      url = "https://huggingface.co/%s/resolve/%s/tokenizer.json" % [identifier, revision].map { |v| CGI.escape(v) }
+
+      path =
+        begin
+          cached_path(cache_dir, url, options)
+        rescue OpenURI::HTTPError
+          raise Error, "Model \"#{identifier}\" on the Hub doesn't have a tokenizer"
+        end
+
+      from_file(path)
+    end
+
+    private
+
+    # use same storage format as Rust version
+    # https://github.com/epwalsh/rust-cached-path
+    def cached_path(cache_dir, url, options)
+      fsum = Digest::SHA256.hexdigest(url)
+      meta_paths = Dir[File.join(cache_dir, "#{fsum}.*.meta")]
+      meta = meta_paths.map { |f| JSON.load_file(f) }.max_by { |m| m["creation_time"] }
+      etag = meta["etag"] if meta
+
+      if etag
+        esum = Digest::SHA256.hexdigest(etag)
+        resource_path = File.join(cache_dir, "#{fsum}.#{esum}")
+        options["If-None-Match"] = etag if File.exist?(resource_path)
+      end
+
+      options[:content_length_proc] = -> (_) { puts "Downloading..." }
+
+      tempfile =
+        begin
+          URI.open(url, options)
+        rescue OpenURI::HTTPError => e
+          if e.message == "304 Not Modified"
+            return resource_path
+          else
+            raise e
+          end
+        end
+
+      etag = tempfile.meta["etag"]
+      esum = Digest::SHA256.hexdigest(etag)
+      resource_path = File.join(cache_dir, "#{fsum}.#{esum}")
+      meta_path = "#{resource_path}.meta"
+
+      meta = {
+        resource: url,
+        resource_path: resource_path,
+        meta_path: meta_path,
+        etag: etag,
+        expires: nil,
+        creation_time: Time.now.to_f
+      }
+
+      File.write("#{resource_path}.lock", "")
+      File.open(resource_path, "wb") { |f| IO.copy_stream(tempfile, f) }
+      File.write(meta_path, JSON.generate(meta))
+
+      resource_path
+    end
+
+    def cache_dir
+      if ENV["TOKENIZERS_CACHE"]
+        ENV["TOKENIZERS_CACHE"]
+      else
+        # use same directory as Rust version
+        # https://docs.rs/dirs/latest/dirs/fn.cache_dir.html
+        dir =
+          if Gem.win_platform?
+            ENV.fetch("LOCALAPPDATA")
+          elsif mac?
+            File.join(ENV.fetch("HOME"), "Library", "Caches")
+          else
+            ENV["XDG_CACHE_HOME"] || File.join(ENV.fetch("HOME"), ".cache")
+          end
+
+        File.join(dir, "huggingface", "tokenizers")
+      end
+    end
+
+    def ensure_cache_dir
+      dir = cache_dir
+      FileUtils.mkdir_p(dir)
+      dir
+    end
+
+    def mac?
+      RbConfig::CONFIG["host_os"] =~ /darwin/i
+    end
+  end
+end

data/lib/tokenizers/version.rb

@@ -1,3 +1,3 @@
 module Tokenizers
-  VERSION = "0.2.1"
+  VERSION = "0.2.2"
 end

data/lib/tokenizers.rb

@@ -7,12 +7,11 @@ end
 
 # modules
 require "tokenizers/char_bpe_tokenizer"
+require "tokenizers/from_pretrained"
 require "tokenizers/version"
 
 module Tokenizers
   class Error < StandardError; end
 
-  def self.from_pretrained(identifier, revision: "main", auth_token: nil)
-    _from_pretrained(identifier, revision, auth_token)
-  end
+  extend FromPretrained
 end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: tokenizers
 version: !ruby/object:Gem::Version
-  version: 0.2.1
+  version: 0.2.2
 platform: ruby
 authors:
 - Andrew Kane
 autorequire:
 bindir: bin
 cert_chain: []
-date: 2023-01-12 00:00:00.000000000 Z
+date: 2023-01-15 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: rb_sys
@@ -48,6 +48,7 @@ files:
 - ext/tokenizers/src/tokenizer.rs
 - lib/tokenizers.rb
 - lib/tokenizers/char_bpe_tokenizer.rb
+- lib/tokenizers/from_pretrained.rb
 - lib/tokenizers/version.rb
 homepage: https://github.com/ankane/tokenizers-ruby
 licenses: