| [WARNING|2024-11-12 05:01:57] logging.py:162 >> We recommend enable `upcast_layernorm` in quantized training. | |
| [INFO|2024-11-12 05:01:57] parser.py:355 >> Process rank: 0, device: cuda:0, n_gpu: 1, distributed training: False, compute dtype: torch.bfloat16 | |
| [INFO|2024-11-12 05:01:57] configuration_utils.py:733 >> loading configuration file config.json from cache at /root/.cache/huggingface/hub/models--google--gemma-2-9b-it/snapshots/11c9b309abf73637e4b6f9a3fa1e92e615547819/config.json | |
| [INFO|2024-11-12 05:01:57] configuration_utils.py:800 >> Model config Gemma2Config { | |
| "_name_or_path": "google/gemma-2-9b-it", | |
| "architectures": [ | |
| "Gemma2ForCausalLM" | |
| ], | |
| "attention_bias": false, | |
| "attention_dropout": 0.0, | |
| "attn_logit_softcapping": 50.0, | |
| "bos_token_id": 2, | |
| "cache_implementation": "hybrid", | |
| "eos_token_id": 1, | |
| "final_logit_softcapping": 30.0, | |
| "head_dim": 256, | |
| "hidden_act": "gelu_pytorch_tanh", | |
| "hidden_activation": "gelu_pytorch_tanh", | |
| "hidden_size": 3584, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 14336, | |
| "max_position_embeddings": 8192, | |
| "model_type": "gemma2", | |
| "num_attention_heads": 16, | |
| "num_hidden_layers": 42, | |
| "num_key_value_heads": 8, | |
| "pad_token_id": 0, | |
| "query_pre_attn_scalar": 256, | |
| "rms_norm_eps": 1e-06, | |
| "rope_theta": 10000.0, | |
| "sliding_window": 4096, | |
| "sliding_window_size": 4096, | |
| "torch_dtype": "bfloat16", | |
| "transformers_version": "4.44.2", | |
| "use_cache": true, | |
| "vocab_size": 256000 | |
| } | |
| [INFO|2024-11-12 05:02:00] tokenization_utils_base.py:2269 >> loading file tokenizer.model from cache at /root/.cache/huggingface/hub/models--google--gemma-2-9b-it/snapshots/11c9b309abf73637e4b6f9a3fa1e92e615547819/tokenizer.model | |
| [INFO|2024-11-12 05:02:00] tokenization_utils_base.py:2269 >> loading file tokenizer.json from cache at /root/.cache/huggingface/hub/models--google--gemma-2-9b-it/snapshots/11c9b309abf73637e4b6f9a3fa1e92e615547819/tokenizer.json | |
| [INFO|2024-11-12 05:02:00] tokenization_utils_base.py:2269 >> loading file added_tokens.json from cache at None | |
| [INFO|2024-11-12 05:02:00] tokenization_utils_base.py:2269 >> loading file special_tokens_map.json from cache at /root/.cache/huggingface/hub/models--google--gemma-2-9b-it/snapshots/11c9b309abf73637e4b6f9a3fa1e92e615547819/special_tokens_map.json | |
| [INFO|2024-11-12 05:02:00] tokenization_utils_base.py:2269 >> loading file tokenizer_config.json from cache at /root/.cache/huggingface/hub/models--google--gemma-2-9b-it/snapshots/11c9b309abf73637e4b6f9a3fa1e92e615547819/tokenizer_config.json | |
| [INFO|2024-11-12 05:02:02] configuration_utils.py:733 >> loading configuration file config.json from cache at /root/.cache/huggingface/hub/models--google--gemma-2-9b-it/snapshots/11c9b309abf73637e4b6f9a3fa1e92e615547819/config.json | |
| [INFO|2024-11-12 05:02:02] configuration_utils.py:800 >> Model config Gemma2Config { | |
| "_name_or_path": "google/gemma-2-9b-it", | |
| "architectures": [ | |
| "Gemma2ForCausalLM" | |
| ], | |
| "attention_bias": false, | |
| "attention_dropout": 0.0, | |
| "attn_logit_softcapping": 50.0, | |
| "bos_token_id": 2, | |
| "cache_implementation": "hybrid", | |
| "eos_token_id": 1, | |
| "final_logit_softcapping": 30.0, | |
| "head_dim": 256, | |
| "hidden_act": "gelu_pytorch_tanh", | |
| "hidden_activation": "gelu_pytorch_tanh", | |
| "hidden_size": 3584, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 14336, | |
| "max_position_embeddings": 8192, | |
| "model_type": "gemma2", | |
| "num_attention_heads": 16, | |
| "num_hidden_layers": 42, | |
| "num_key_value_heads": 8, | |
| "pad_token_id": 0, | |
| "query_pre_attn_scalar": 256, | |
| "rms_norm_eps": 1e-06, | |
| "rope_theta": 10000.0, | |
| "sliding_window": 4096, | |
| "sliding_window_size": 4096, | |
| "torch_dtype": "bfloat16", | |
| "transformers_version": "4.44.2", | |
| "use_cache": true, | |
| "vocab_size": 256000 | |
| } | |
| [INFO|2024-11-12 05:02:02] tokenization_utils_base.py:2269 >> loading file tokenizer.model from cache at /root/.cache/huggingface/hub/models--google--gemma-2-9b-it/snapshots/11c9b309abf73637e4b6f9a3fa1e92e615547819/tokenizer.model | |
| [INFO|2024-11-12 05:02:02] tokenization_utils_base.py:2269 >> loading file tokenizer.json from cache at /root/.cache/huggingface/hub/models--google--gemma-2-9b-it/snapshots/11c9b309abf73637e4b6f9a3fa1e92e615547819/tokenizer.json | |
| [INFO|2024-11-12 05:02:02] tokenization_utils_base.py:2269 >> loading file added_tokens.json from cache at None | |
| [INFO|2024-11-12 05:02:02] tokenization_utils_base.py:2269 >> loading file special_tokens_map.json from cache at /root/.cache/huggingface/hub/models--google--gemma-2-9b-it/snapshots/11c9b309abf73637e4b6f9a3fa1e92e615547819/special_tokens_map.json | |
| [INFO|2024-11-12 05:02:02] tokenization_utils_base.py:2269 >> loading file tokenizer_config.json from cache at /root/.cache/huggingface/hub/models--google--gemma-2-9b-it/snapshots/11c9b309abf73637e4b6f9a3fa1e92e615547819/tokenizer_config.json | |
| [INFO|2024-11-12 05:02:03] logging.py:157 >> Loading dataset treino_pt_rde.json... | |
| [INFO|2024-11-12 05:02:37] configuration_utils.py:733 >> loading configuration file config.json from cache at /root/.cache/huggingface/hub/models--google--gemma-2-9b-it/snapshots/11c9b309abf73637e4b6f9a3fa1e92e615547819/config.json | |
| [INFO|2024-11-12 05:02:37] configuration_utils.py:800 >> Model config Gemma2Config { | |
| "_name_or_path": "google/gemma-2-9b-it", | |
| "architectures": [ | |
| "Gemma2ForCausalLM" | |
| ], | |
| "attention_bias": false, | |
| "attention_dropout": 0.0, | |
| "attn_logit_softcapping": 50.0, | |
| "bos_token_id": 2, | |
| "cache_implementation": "hybrid", | |
| "eos_token_id": 1, | |
| "final_logit_softcapping": 30.0, | |
| "head_dim": 256, | |
| "hidden_act": "gelu_pytorch_tanh", | |
| "hidden_activation": "gelu_pytorch_tanh", | |
| "hidden_size": 3584, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 14336, | |
| "max_position_embeddings": 8192, | |
| "model_type": "gemma2", | |
| "num_attention_heads": 16, | |
| "num_hidden_layers": 42, | |
| "num_key_value_heads": 8, | |
| "pad_token_id": 0, | |
| "query_pre_attn_scalar": 256, | |
| "rms_norm_eps": 1e-06, | |
| "rope_theta": 10000.0, | |
| "sliding_window": 4096, | |
| "sliding_window_size": 4096, | |
| "torch_dtype": "bfloat16", | |
| "transformers_version": "4.44.2", | |
| "use_cache": true, | |
| "vocab_size": 256000 | |
| } | |
| [WARNING|2024-11-12 05:02:37] logging.py:162 >> FlashAttention-2 is not installed, use eager attention. | |
| [INFO|2024-11-12 05:02:37] logging.py:157 >> Quantizing model to 4 bit with bitsandbytes. | |
| [INFO|2024-11-12 05:02:37] modeling_utils.py:3678 >> loading weights file model.safetensors from cache at /root/.cache/huggingface/hub/models--google--gemma-2-9b-it/snapshots/11c9b309abf73637e4b6f9a3fa1e92e615547819/model.safetensors.index.json | |
| [INFO|2024-11-12 05:09:58] modeling_utils.py:1606 >> Instantiating Gemma2ForCausalLM model under default dtype torch.bfloat16. | |
| [INFO|2024-11-12 05:09:58] configuration_utils.py:1038 >> Generate config GenerationConfig { | |
| "bos_token_id": 2, | |
| "cache_implementation": "hybrid", | |
| "eos_token_id": 1, | |
| "pad_token_id": 0 | |
| } | |
| [INFO|2024-11-12 05:11:28] modeling_utils.py:4507 >> All model checkpoint weights were used when initializing Gemma2ForCausalLM. | |
| [INFO|2024-11-12 05:11:28] modeling_utils.py:4515 >> All the weights of Gemma2ForCausalLM were initialized from the model checkpoint at google/gemma-2-9b-it. | |
| If your task is similar to the task the model of the checkpoint was trained on, you can already use Gemma2ForCausalLM for predictions without further training. | |
| [INFO|2024-11-12 05:11:28] configuration_utils.py:993 >> loading configuration file generation_config.json from cache at /root/.cache/huggingface/hub/models--google--gemma-2-9b-it/snapshots/11c9b309abf73637e4b6f9a3fa1e92e615547819/generation_config.json | |
| [INFO|2024-11-12 05:11:28] configuration_utils.py:1038 >> Generate config GenerationConfig { | |
| "bos_token_id": 2, | |
| "cache_implementation": "hybrid", | |
| "eos_token_id": 1, | |
| "pad_token_id": 0 | |
| } | |
| [INFO|2024-11-12 05:11:29] logging.py:157 >> Gradient checkpointing enabled. | |
| [INFO|2024-11-12 05:11:29] logging.py:157 >> Using vanilla attention implementation. | |
| [INFO|2024-11-12 05:11:29] logging.py:157 >> Upcasting trainable params to float32. | |
| [INFO|2024-11-12 05:11:29] logging.py:157 >> Fine-tuning method: LoRA | |
| [INFO|2024-11-12 05:11:29] logging.py:157 >> Found linear modules: up_proj,v_proj,k_proj,o_proj,down_proj,q_proj,gate_proj | |
| [INFO|2024-11-12 05:11:29] logging.py:157 >> trainable params: 27,009,024 || all params: 9,268,715,008 || trainable%: 0.2914 | |
| [INFO|2024-11-12 05:11:29] trainer.py:648 >> Using auto half precision backend | |
| [INFO|2024-11-12 05:11:30] trainer.py:2134 >> ***** Running training ***** | |
| [INFO|2024-11-12 05:11:30] trainer.py:2135 >> Num examples = 300 | |
| [INFO|2024-11-12 05:11:30] trainer.py:2136 >> Num Epochs = 3 | |
| [INFO|2024-11-12 05:11:30] trainer.py:2137 >> Instantaneous batch size per device = 2 | |
| [INFO|2024-11-12 05:11:30] trainer.py:2140 >> Total train batch size (w. parallel, distributed & accumulation) = 8 | |
| [INFO|2024-11-12 05:11:30] trainer.py:2141 >> Gradient Accumulation steps = 4 | |
| [INFO|2024-11-12 05:11:30] trainer.py:2142 >> Total optimization steps = 111 | |
| [INFO|2024-11-12 05:11:30] trainer.py:2143 >> Number of trainable parameters = 27,009,024 | |
| [INFO|2024-11-12 05:22:57] logging.py:157 >> {'loss': 1.9494, 'learning_rate': 4.9005e-05, 'epoch': 0.27, 'throughput': 33.83} | |
| [INFO|2024-11-12 05:34:08] logging.py:157 >> {'loss': 0.3681, 'learning_rate': 4.6101e-05, 'epoch': 0.53, 'throughput': 33.79} | |
| [INFO|2024-11-12 05:45:21] logging.py:157 >> {'loss': 0.4155, 'learning_rate': 4.1517e-05, 'epoch': 0.80, 'throughput': 33.72} | |
| [INFO|2024-11-12 05:56:21] logging.py:157 >> {'loss': 0.3174, 'learning_rate': 3.5619e-05, 'epoch': 1.07, 'throughput': 33.71} | |
| [INFO|2024-11-12 06:07:30] logging.py:157 >> {'loss': 0.2294, 'learning_rate': 2.8876e-05, 'epoch': 1.33, 'throughput': 33.75} | |
| [INFO|2024-11-12 06:18:44] logging.py:157 >> {'loss': 0.2442, 'learning_rate': 2.1825e-05, 'epoch': 1.60, 'throughput': 33.76} | |
| [INFO|2024-11-12 06:29:58] logging.py:157 >> {'loss': 0.3141, 'learning_rate': 1.5026e-05, 'epoch': 1.87, 'throughput': 33.77} | |
| [INFO|2024-11-12 06:41:10] logging.py:157 >> {'loss': 0.1918, 'learning_rate': 9.0208e-06, 'epoch': 2.13, 'throughput': 33.79} | |
| [INFO|2024-11-12 06:52:18] logging.py:157 >> {'loss': 0.1719, 'learning_rate': 4.2873e-06, 'epoch': 2.40, 'throughput': 33.78} | |
| [INFO|2024-11-12 07:03:38] logging.py:157 >> {'loss': 0.1387, 'learning_rate': 1.2018e-06, 'epoch': 2.67, 'throughput': 33.80} | |
| [INFO|2024-11-12 07:14:44] logging.py:157 >> {'loss': 0.1394, 'learning_rate': 1.0012e-08, 'epoch': 2.93, 'throughput': 33.79} | |
| [INFO|2024-11-12 07:15:51] trainer.py:3503 >> Saving model checkpoint to saves/Gemma-2-9B-Instruct/lora/train_gemma2/checkpoint-111 | |
| [INFO|2024-11-12 07:15:51] configuration_utils.py:733 >> loading configuration file config.json from cache at /root/.cache/huggingface/hub/models--google--gemma-2-9b-it/snapshots/11c9b309abf73637e4b6f9a3fa1e92e615547819/config.json | |
| [INFO|2024-11-12 07:15:51] configuration_utils.py:800 >> Model config Gemma2Config { | |
| "architectures": [ | |
| "Gemma2ForCausalLM" | |
| ], | |
| "attention_bias": false, | |
| "attention_dropout": 0.0, | |
| "attn_logit_softcapping": 50.0, | |
| "bos_token_id": 2, | |
| "cache_implementation": "hybrid", | |
| "eos_token_id": 1, | |
| "final_logit_softcapping": 30.0, | |
| "head_dim": 256, | |
| "hidden_act": "gelu_pytorch_tanh", | |
| "hidden_activation": "gelu_pytorch_tanh", | |
| "hidden_size": 3584, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 14336, | |
| "max_position_embeddings": 8192, | |
| "model_type": "gemma2", | |
| "num_attention_heads": 16, | |
| "num_hidden_layers": 42, | |
| "num_key_value_heads": 8, | |
| "pad_token_id": 0, | |
| "query_pre_attn_scalar": 256, | |
| "rms_norm_eps": 1e-06, | |
| "rope_theta": 10000.0, | |
| "sliding_window": 4096, | |
| "sliding_window_size": 4096, | |
| "torch_dtype": "bfloat16", | |
| "transformers_version": "4.44.2", | |
| "use_cache": true, | |
| "vocab_size": 256000 | |
| } | |
| [INFO|2024-11-12 07:15:52] tokenization_utils_base.py:2684 >> tokenizer config file saved in saves/Gemma-2-9B-Instruct/lora/train_gemma2/checkpoint-111/tokenizer_config.json | |
| [INFO|2024-11-12 07:15:52] tokenization_utils_base.py:2693 >> Special tokens file saved in saves/Gemma-2-9B-Instruct/lora/train_gemma2/checkpoint-111/special_tokens_map.json | |
| [INFO|2024-11-12 07:15:58] trainer.py:2394 >> | |
| Training completed. Do not forget to share your model on huggingface.co/models =) | |
| [INFO|2024-11-12 07:15:58] trainer.py:3503 >> Saving model checkpoint to saves/Gemma-2-9B-Instruct/lora/train_gemma2 | |
| [INFO|2024-11-12 07:15:58] configuration_utils.py:733 >> loading configuration file config.json from cache at /root/.cache/huggingface/hub/models--google--gemma-2-9b-it/snapshots/11c9b309abf73637e4b6f9a3fa1e92e615547819/config.json | |
| [INFO|2024-11-12 07:15:58] configuration_utils.py:800 >> Model config Gemma2Config { | |
| "architectures": [ | |
| "Gemma2ForCausalLM" | |
| ], | |
| "attention_bias": false, | |
| "attention_dropout": 0.0, | |
| "attn_logit_softcapping": 50.0, | |
| "bos_token_id": 2, | |
| "cache_implementation": "hybrid", | |
| "eos_token_id": 1, | |
| "final_logit_softcapping": 30.0, | |
| "head_dim": 256, | |
| "hidden_act": "gelu_pytorch_tanh", | |
| "hidden_activation": "gelu_pytorch_tanh", | |
| "hidden_size": 3584, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 14336, | |
| "max_position_embeddings": 8192, | |
| "model_type": "gemma2", | |
| "num_attention_heads": 16, | |
| "num_hidden_layers": 42, | |
| "num_key_value_heads": 8, | |
| "pad_token_id": 0, | |
| "query_pre_attn_scalar": 256, | |
| "rms_norm_eps": 1e-06, | |
| "rope_theta": 10000.0, | |
| "sliding_window": 4096, | |
| "sliding_window_size": 4096, | |
| "torch_dtype": "bfloat16", | |
| "transformers_version": "4.44.2", | |
| "use_cache": true, | |
| "vocab_size": 256000 | |
| } | |
| [INFO|2024-11-12 07:15:59] tokenization_utils_base.py:2684 >> tokenizer config file saved in saves/Gemma-2-9B-Instruct/lora/train_gemma2/tokenizer_config.json | |
| [INFO|2024-11-12 07:15:59] tokenization_utils_base.py:2693 >> Special tokens file saved in saves/Gemma-2-9B-Instruct/lora/train_gemma2/special_tokens_map.json | |
| [WARNING|2024-11-12 07:16:00] logging.py:162 >> No metric eval_loss to plot. | |
| [WARNING|2024-11-12 07:16:00] logging.py:162 >> No metric eval_accuracy to plot. | |
| [INFO|2024-11-12 07:16:00] modelcard.py:449 >> Dropping the following result as it does not have all the necessary fields: | |
| {'task': {'name': 'Causal Language Modeling', 'type': 'text-generation'}} | |