nanogpt 0.1.0 → 0.1.1

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: b10e62747a63be10b519fdd8eb1eaabfc63214b01ab553d4230af865a0a8f327
-  data.tar.gz: 69db126e3d02c897045e543981223d1d29d72f7625bd381e3ba4badee88b6d20
+  metadata.gz: 17c3921412aa2d70e303ba68f979f5ccf9bb3e9b815ddfe3fbfbfcc62b7a3be6
+  data.tar.gz: 1b0b97a5513a341ae93bd046a86df8237f3941d3cbe11dec480a6cb59c8ea597
 SHA512:
-  metadata.gz: 97f5846373224b889f22b80302a05271c7ab0276037e519c19c58cc0954aa982a4d3ac9992766aa73f6d1eec1dfe2b96b8a750414dd02818529e37789671cd87
-  data.tar.gz: accd2112564f004cc763dbaac418f35ce24426d4d86e53935a32bdf6411659463ed6cb8e9415146136ee78fa79748e39a1e506b5c44848a7b6e503dfee86e310
+  metadata.gz: b3b7d07a8016bb7817f20bb22476d1d85bdaad42600234f8cf6a763316b6b128e473bccf17eef5135486c5b67051c3d1f10278214a28ec72cd26da18a4f68c8c
+  data.tar.gz: 90fe296ae42f058d9a918ba630781edca983ef22314797ef44a82e15c257bf06903c10b0442611e78a9ff654c08978720e1ece4bb3bb13b52f5d3cffe2331122

data/exe/nanogpt

@@ -6,7 +6,7 @@ $stdout.sync = true
 require "nano_gpt"
 
 class NanoGPTCLI
-  COMMANDS = %w[train sample bench version help].freeze
+  COMMANDS = %w[prepare train sample bench version help].freeze
 
   def initialize(args)
     @command = args.shift
@@ -15,6 +15,8 @@ class NanoGPTCLI
 
   def run
     case @command
+    when "prepare"
+      prepare
     when "train"
       train
     when "sample"
@@ -35,6 +37,37 @@ class NanoGPTCLI
 
   private
 
+  def prepare
+    dataset = @args.first
+
+    # Find available datasets
+    data_dir = File.join(File.dirname(__FILE__), "..", "data")
+    available = Dir.glob(File.join(data_dir, "*", "prepare.rb")).map do |path|
+      File.basename(File.dirname(path))
+    end.sort
+
+    if dataset.nil?
+      puts "Usage: nanogpt prepare <dataset>"
+      puts ""
+      puts "Available datasets:"
+      available.each { |d| puts "  #{d}" }
+      exit 1
+    end
+
+    prepare_script = File.join(data_dir, dataset, "prepare.rb")
+
+    unless File.exist?(prepare_script)
+      puts "Error: Unknown dataset '#{dataset}'"
+      puts ""
+      puts "Available datasets:"
+      available.each { |d| puts "  #{d}" }
+      exit 1
+    end
+
+    puts "Preparing dataset: #{dataset}"
+    load prepare_script
+  end
+
   def train
     config = NanoGPT::TrainConfig.load(@args)
 
@@ -318,6 +351,7 @@ class NanoGPTCLI
       Usage: nanogpt <command> [options]
 
       Commands:
+        prepare   Download and prepare a dataset
         train     Train a GPT model
         sample    Generate text from a trained model
         bench     Run performance benchmarks
@@ -325,9 +359,9 @@ class NanoGPTCLI
         help      Show this help message
 
       Examples:
-        nanogpt train --config=config/train_shakespeare_char.json
-        nanogpt train --dataset=shakespeare_char --max_iters=1000
-        nanogpt sample --out_dir=out-shakespeare-char --num_samples=3
+        nanogpt prepare shakespeare_char
+        nanogpt train --dataset=shakespeare_char --device=mps
+        nanogpt sample --dataset=shakespeare_char --num_samples=3
         nanogpt bench --batch_size=8 --block_size=512
 
       For more information, visit: https://github.com/khasinski/nanogpt-rb

data/lib/nano_gpt/version.rb

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

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: nanogpt
 version: !ruby/object:Gem::Version
-  version: 0.1.0
+  version: 0.1.1
 platform: ruby
 authors:
 - Chris Hasiński