| { |
| "output_dir": "xdhshshsgv", |
| "dataset_name": "HuggingFaceFW/finewiki", |
| "dataset_config": "sr", |
| "train_split": "train", |
| "validation_split": "validation", |
| "dataset_streaming": true, |
| "dataset_streaming_train_take": 256, |
| "dataset_streaming_validation_take": 256, |
| "dataset_streaming_candidates": [ |
| "HuggingFaceFW/finewiki" |
| ], |
| "dataset_streaming_overrides": { |
| "HuggingFaceFW/finewiki": "sr" |
| }, |
| "dataset_streaming_autodiscover": false, |
| "auto_train_each_dataset": false, |
| "quick_test_mode": true, |
| "quick_test_train_steps": 64, |
| "quick_test_max_training_cycles": 2, |
| "quick_test_max_total_steps": 512, |
| "quick_test_sample_attempts": 6, |
| "quick_test_sample_max_new_tokens": 160, |
| "quick_test_chat_max_new_tokens": 200, |
| "quick_test_dataset_take": 256, |
| "quick_test_alignment_examples": 512, |
| "quick_test_alignment_generations": 4, |
| "quick_test_streaming_chunk_size": 2048, |
| "quick_test_streaming_prefetch": 2, |
| "quick_test_streaming_processor_workers": 2, |
| "continuous_training": true, |
| "vocab_size": 104, |
| "sequence_length": 256, |
| "num_hidden_layers": 1, |
| "hidden_size": 128, |
| "intermediate_size": 512, |
| "num_attention_heads": 4, |
| "learning_rate": 0.00025, |
| "weight_decay": 0.01, |
| "train_batch_size": 4, |
| "eval_batch_size": 4, |
| "gradient_accumulation_steps": 1, |
| "train_steps": 64, |
| "warmup_ratio": 0.12, |
| "logging_steps": 4, |
| "eval_steps": 16, |
| "eval_max_batches": 8, |
| "seed": 42, |
| "ascii_only_tokenizer": true, |
| "sample_prompt": null, |
| "sample_max_new_tokens": 160, |
| "sample_chunk_size": 0, |
| "sample_temperature": 0.38, |
| "sample_temperature_floor": 0.18, |
| "sample_temperature_ceiling": 0.75, |
| "sample_top_p": 0.62, |
| "sample_top_p_floor": 0.35, |
| "sample_top_p_ceiling": 0.95, |
| "sample_repetition_penalty": 1.1, |
| "sample_attempts": 6, |
| "sample_temperature_decay": 0.6, |
| "sample_top_p_decay": 0.68, |
| "retry_sample_attempt_growth": 3, |
| "retry_sample_attempt_cap": 6, |
| "retry_sample_max_new_tokens_growth": 18, |
| "retry_sample_max_new_tokens_cap": 160, |
| "retry_chat_preview_max_new_tokens_growth": 20, |
| "retry_chat_preview_max_new_tokens_cap": 200, |
| "retry_sample_temperature_step": -0.04, |
| "retry_sample_top_p_boost": 0.05, |
| "retry_min_sample_temperature_decay": 0.5, |
| "retry_min_sample_top_p_decay": 0.5, |
| "retry_temperature_step_multiplier": 1.15, |
| "retry_top_p_boost_multiplier": 1.08, |
| "retry_cycle_growth_factor": 1.4, |
| "retry_cycle_growth_cap": 9.0, |
| "retry_learning_rate_decay": 0.85, |
| "retry_min_learning_rate": 5e-05, |
| "retry_gradient_accumulation_growth": 1, |
| "retry_gradient_accumulation_cap": 12, |
| "retry_relaxation_factor_decay": 0.9, |
| "retry_relaxation_perplexity_boost": 0.15, |
| "retry_sample_chunk_growth": 8, |
| "retry_sample_chunk_cap": 0, |
| "retry_chat_preview_chunk_growth": 10, |
| "retry_chat_preview_chunk_cap": 0, |
| "retry_max_training_cycles_growth": 1, |
| "retry_max_training_cycles_cap": 2.0, |
| "retry_max_total_steps_growth": 96, |
| "retry_max_total_steps_cap": 512, |
| "retry_dataset_step_multiplier_boost": 0.4, |
| "retry_dataset_step_multiplier_cap": 6.0, |
| "retry_dataset_example_fraction_decay": 0.88, |
| "retry_dataset_example_fraction_floor": 0.18, |
| "retry_dataset_token_ratio_decay": 0.85, |
| "retry_dataset_token_ratio_floor": 0.18, |
| "retry_dataset_scan_multiplier_growth": 1, |
| "retry_dataset_scan_multiplier_cap": 6, |
| "chat_preview": false, |
| "chat_preview_system_prompt": null, |
| "chat_preview_user_prompt": null, |
| "chat_preview_max_new_tokens": 200, |
| "chat_preview_chunk_size": 0, |
| "chat_preview_fallback_message": "Lo siento, aún no puedo generar una respuesta coherente.", |
| "coherence_target_perplexity": 24.0, |
| "coherence_min_chars": 64, |
| "coherence_min_words": 11, |
| "coherence_unique_token_ratio": 0.48, |
| "coherence_max_word_repetition": 2, |
| "max_training_cycles": 2, |
| "cycle_step_growth": 2.6, |
| "coherence_alpha_token_ratio": 0.8, |
| "coherence_min_stopwords": 4, |
| "coherence_min_alpha_fraction": 0.72, |
| "coherence_min_clean_tokens": 11, |
| "coherence_min_clean_ratio": 0.78, |
| "coherence_min_dictionary_hits": 5, |
| "coherence_min_known_word_ratio": 0.4, |
| "coherence_max_unknown_ratio": 0.4, |
| "coherence_min_known_word_hits": 6, |
| "coherence_max_symbol_fraction": 0.08, |
| "coherence_max_char_run": 3, |
| "coherence_min_sentence_endings": 1, |
| "coherence_min_vowel_tokens": 6, |
| "coherence_min_vowel_ratio": 0.34, |
| "coherence_max_repeat_fraction": 0.3, |
| "coherence_max_repeat_tokens": 4, |
| "coherence_required_keywords": [ |
| "inteligencia", |
| "artificial", |
| "objetivo", |
| "principal", |
| "sistemas", |
| "datos", |
| "soluciones" |
| ], |
| "coherence_min_required_keywords": 3, |
| "coherence_relaxation_factor": 0.85, |
| "coherence_relaxation_max_passes": 6, |
| "coherence_relaxation_perplexity_growth": 0.38, |
| "dataset_min_chars": 60, |
| "dataset_min_words": 8, |
| "dataset_max_symbol_fraction": 0.12, |
| "dataset_max_char_run": 4, |
| "dataset_corrupt_keep_fraction": 0.01, |
| "dataset_min_vowel_tokens": 4, |
| "dataset_min_vowel_ratio": 0.28, |
| "dataset_max_examples_per_split": null, |
| "dataset_forbidden_substrings": [ |
| "<|", |
| "|>", |
| "assistant_", |
| "_assistant", |
| "<assistant", |
| "<user", |
| "</s>", |
| "<s>" |
| ], |
| "dataset_chat_system_prompt": "Eres un asistente que debe reproducir literalmente el texto proporcionado en partes ordenadas.", |
| "dataset_chat_user_prompt": "Entrega la parte {part} de {total} exactamente como aparece en la fuente.", |
| "dataset_chat_max_segment_chars": 0, |
| "dataset_chat_format_use_processes": true, |
| "dataset_chat_format_prefetch": null, |
| "dataset_catalogue_max_entries": null, |
| "coherence_known_words_top_k": 24, |
| "coherence_known_words_min_freq": 3, |
| "max_total_steps": 512, |
| "sentencepiece_input_sentence_size": 512, |
| "sentencepiece_max_sentence_length": 360, |
| "sentencepiece_shuffle_input_sentence": true, |
| "dataset_streaming_scan_multiplier": 1, |
| "dataset_streaming_scan_buffer": 128, |
| "dataset_streaming_loader_workers": 2, |
| "dataset_streaming_processor_workers": 2, |
| "dataset_streaming_scan_chunk_size": 2048, |
| "dataset_streaming_max_chunk_size": 2048, |
| "dataset_streaming_max_prefetch_chunks": 2, |
| "dataset_streaming_map_chunksize": null, |
| "dataset_streaming_progress_instant": true, |
| "dataset_streaming_prefetch_chunks": 2, |
| "dataset_streaming_progress_interval": 1.0, |
| "web_scraper_enabled": true, |
| "web_scraper_start_urls": [ |
| "https://example.com/" |
| ], |
| "web_scraper_max_depth": 2, |
| "web_scraper_max_pages": 200, |
| "web_scraper_concurrency": 8, |
| "web_scraper_timeout": 10.0, |
| "web_scraper_user_agent": "Mozilla/5.0 (compatible; TinyLLaMAWebScraper/1.0; +https://huggingface.co)", |
| "web_scraper_text_min_chars": 160, |
| "web_scraper_retry_attempts": 2, |
| "web_scraper_max_links_per_page": 200, |
| "sklearn_newsgroups_enabled": true, |
| "sklearn_newsgroups_subset": "all", |
| "sklearn_newsgroups_categories": [], |
| "sklearn_newsgroups_remove_headers": true, |
| "sklearn_newsgroups_remove_quotes": true, |
| "sklearn_newsgroups_texts_per_second": 500, |
| "sklearn_newsgroups_batch_size": 500, |
| "sklearn_newsgroups_workers": 8, |
| "sklearn_newsgroups_max_texts": null, |
| "sklearn_newsgroups_shuffle": true, |
| "dataset_streaming_validation_shards": 64, |
| "dataset_allow_non_streaming_fallback": false, |
| "dataset_repo_fallback_enabled": true, |
| "dataset_repo_fallback_max_files": null, |
| "dataset_repo_fallback_extensions": [ |
| ".jsonl", |
| ".json", |
| ".txt", |
| ".csv", |
| ".md", |
| ".jsonl.gz", |
| ".json.gz", |
| ".txt.gz", |
| ".csv.gz", |
| ".parquet", |
| ".jsonl.zst" |
| ], |
| "dataset_repo_fallback_workers": null, |
| "dataset_repo_fallback_segment_chars": 2048, |
| "dataset_chat_format_workers": null, |
| "dataset_chat_format_batch_size": null, |
| "dataset_step_multiplier": 1.3, |
| "dataset_step_example_fraction": 0.92, |
| "dataset_step_token_ratio": 0.68, |
| "tokenizer_additional_special_tokens": [ |
| "<|im_start|>", |
| "<|im_end|>", |
| "<|system|>", |
| "<|user|>", |
| "<|assistant|>", |
| "<|tool|>", |
| "<|assistant_reasoning|>", |
| "<|assistant_reasoning_end|>" |
| ], |
| "tokenizer_chat_template": "\n{% for message in messages %}\n{% if message['role'] == 'system' %}\n<|im_start|>system\n{{ message['content'] }}<|im_end|>\n{% elif message['role'] == 'user' %}\n<|im_start|>user\n{{ message['content'] }}<|im_end|>\n{% elif message['role'] == 'assistant' %}\n<|im_start|>assistant\n{% if message.get('reasoning') %}\n<|assistant_reasoning|>\n{{ message['reasoning'] | trim }}\n<|assistant_reasoning_end|>\n{% endif %}\n{{ message['content'] }}<|im_end|>\n{% elif message['role'] == 'tool' %}\n<|im_start|>tool\n{{ message['content'] }}<|im_end|>\n{% endif %}\n{% endfor %}\n<|im_start|>assistant\n", |
| "allow_dataset_downloads": false, |
| "huggingface_username": "Ignaciohhhhggfgjfrffd", |
| "huggingface_repo_name": "tiny-llama-ultra-compact", |
| "huggingface_upload": true, |
| "huggingface_periodic_upload_interval": 600, |
| "auto_install_dependencies": true, |
| "runtime_dependency_torch_index_url": "https://download.pytorch.org/whl/cpu", |
| "runtime_dependencies": { |
| "torch": "torch", |
| "datasets": "datasets", |
| "transformers": "transformers", |
| "accelerate": "accelerate", |
| "peft": "peft", |
| "trl": "trl", |
| "sentencepiece": "sentencepiece", |
| "huggingface_hub": "huggingface_hub", |
| "requests": "requests", |
| "bs4": "beautifulsoup4", |
| "sklearn": "scikit-learn" |
| }, |
| "numeric_self_test_enabled": true, |
| "numeric_self_test_attempts": 1, |
| "numeric_self_test_relaxation_passes": 1, |
| "numeric_self_test_perplexity": 48.0, |
| "numeric_self_test_history_limit": 2, |
| "coherence_stopwords": [ |
| "la", |
| "el", |
| "de", |
| "que", |
| "en", |
| "y", |
| "es", |
| "para", |
| "con", |
| "no", |
| "lo", |
| "una", |
| "un", |
| "the", |
| "and", |
| "of", |
| "to", |
| "in", |
| "is" |
| ], |
| "coherence_dictionary": [ |
| "inteligencia", |
| "artificial", |
| "datos", |
| "sistema", |
| "aprendizaje", |
| "tecnologia", |
| "modelo", |
| "informacion", |
| "conocimiento", |
| "red", |
| "texto", |
| "la", |
| "el", |
| "los", |
| "las", |
| "una", |
| "un", |
| "the", |
| "and", |
| "of", |
| "to", |
| "in", |
| "is", |
| "data", |
| "learning", |
| "system", |
| "machine", |
| "model", |
| "science", |
| "technology", |
| "respuesta", |
| "coherente", |
| "pregunta", |
| "objetivo", |
| "principal", |
| "ayuda", |
| "soluciones", |
| "responsables", |
| "decisiones", |
| "automatizan", |
| "transparentes", |
| "transparente", |
| "colaboran", |
| "problemas", |
| "confiables" |
| ], |
| "coherence_forbidden_substrings": [ |
| "�", |
| "�", |
| "<unk>", |
| "<s>", |
| "</s>", |
| "<|", |
| "|>", |
| "__", |
| "assistant_", |
| "<assistant", |
| "<user" |
| ], |
| "repair_support_terms": [ |
| "datos", |
| "modelo", |
| "sistema", |
| "informacion", |
| "aprendizaje", |
| "tecnologia", |
| "respuesta", |
| "respuestas", |
| "claridad", |
| "coherencia", |
| "coherentes", |
| "naturales", |
| "natural", |
| "organiza", |
| "usa", |
| "presenta", |
| "objetivo", |
| "principales", |
| "beneficios", |
| "aplicaciones", |
| "impacto", |
| "soluciones", |
| "problemas", |
| "decisiones", |
| "responsables", |
| "algoritmos" |
| ], |
| "repair_forbidden_fragments": [ |
| "entrega la parte", |
| "debe reproducir literalmente", |
| "<|im_start|>system", |
| "<|im_start|>user" |
| ], |
| "repair_keyword_answer": "La inteligencia artificial tiene como objetivo principal desarrollar sistemas que aprenden de los datos para ofrecer soluciones responsables y confiables. Estos sistemas analizan información, automatizan decisiones y colaboran con las personas para resolver problemas complejos de forma transparente.", |
| "repair_min_alpha_fraction": 0.68, |
| "repair_min_vowel_ratio": 0.3, |
| "repair_max_char_run": 3, |
| "enable_alignment_pipeline": false, |
| "alignment_pipeline": { |
| "enabled": false, |
| "base_model_dir": null, |
| "stage1_output_dir": "outputs/Qwen-1.5B-Mega-SFT-Stage1", |
| "stage2_output_dir": "outputs/Qwen-1.5B-Reasoning-GRPO-Stage2", |
| "stage3_output_dir": "outputs/Qwen-1.5B-Final-Aligned-Model", |
| "stage1_examples_per_dataset": 512, |
| "stage1_translation_examples": 512, |
| "stage1_num_proc": 4, |
| "stage1_streaming": true, |
| "stage1_streaming_buffer": 4096, |
| "stage1_dataset_seed": 42, |
| "stage2_max_examples": 512, |
| "stage2_num_generations": 4, |
| "stage2_max_prompt_length": 256, |
| "stage2_max_completion_length": 786, |
| "stage3_examples_per_dataset": 512, |
| "stage3_dataset_seed": 42, |
| "report_to": "wandb", |
| "upload_repo_suffix": "alignment", |
| "stage1_run_name": "Qwen-1.5B-Mega-SFT-Stage1", |
| "stage2_run_name": "Qwen-1.5B-Reasoning-GRPO-Stage2", |
| "stage3_run_name": "Qwen-1.5B-Final-Aligned-DPO-Stage3", |
| "stage4_output_dir": "outputs/Qwen-1.5B-Reflective-Awareness-Stage4", |
| "stage5_output_dir": "outputs/Qwen-1.5B-Elite-Reasoning-Stage5", |
| "stage4_examples_per_dataset": 512, |
| "stage4_dataset_seed": 52, |
| "stage4_run_name": "Qwen-1.5B-Reflective-Awareness-Stage4", |
| "stage5_max_examples": 512, |
| "stage5_num_generations": 4, |
| "stage5_max_prompt_length": 320, |
| "stage5_max_completion_length": 896, |
| "stage5_run_name": "Qwen-1.5B-Elite-Reasoning-Stage5", |
| "stage6_output_dir": "outputs/Qwen-1.5B-Bilingual-Translator-Stage6", |
| "stage6_examples_per_dataset": 8000, |
| "stage6_dataset_seed": 64, |
| "stage6_run_name": "Qwen-1.5B-Bilingual-Translator-Stage6" |
| }, |
| "exported_at": "2025-10-30T21:43:14.906209Z" |
| } |
