[WARNING|2025-01-15 22:42:41] logging.py:162 >> `ddp_find_unused_parameters` needs to be set as False for LoRA in DDP training. [INFO|2025-01-15 22:42:41] parser.py:359 >> Process rank: 0, device: cuda:0, n_gpu: 1, distributed training: True, compute dtype: torch.bfloat16 [INFO|2025-01-15 22:42:42] configuration_utils.py:679 >> loading configuration file config.json from cache at /nethome/atrinh31/.cache/huggingface/hub/models--mistralai--Mistral-7B-Instruct-v0.2/snapshots/3ad372fc79158a2148299e3318516c786aeded6c/config.json [INFO|2025-01-15 22:42:42] configuration_utils.py:746 >> Model config MistralConfig { "_name_or_path": "mistralai/Mistral-7B-Instruct-v0.2", "architectures": [ "MistralForCausalLM" ], "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "head_dim": 128, "hidden_act": "silu", "hidden_size": 4096, "initializer_range": 0.02, "intermediate_size": 14336, "max_position_embeddings": 32768, "model_type": "mistral", "num_attention_heads": 32, "num_hidden_layers": 32, "num_key_value_heads": 8, "rms_norm_eps": 1e-05, "rope_theta": 1000000.0, "sliding_window": null, "tie_word_embeddings": false, "torch_dtype": "bfloat16", "transformers_version": "4.46.1", "use_cache": true, "vocab_size": 32000 } [INFO|2025-01-15 22:42:42] parser.py:359 >> Process rank: 3, device: cuda:3, n_gpu: 1, distributed training: True, compute dtype: torch.bfloat16 [INFO|2025-01-15 22:42:42] tokenization_utils_base.py:2211 >> loading file tokenizer.model from cache at /nethome/atrinh31/.cache/huggingface/hub/models--mistralai--Mistral-7B-Instruct-v0.2/snapshots/3ad372fc79158a2148299e3318516c786aeded6c/tokenizer.model [INFO|2025-01-15 22:42:42] parser.py:359 >> Process rank: 1, device: cuda:1, n_gpu: 1, distributed training: True, compute dtype: torch.bfloat16 [INFO|2025-01-15 22:42:42] parser.py:359 >> Process rank: 4, device: cuda:4, n_gpu: 1, distributed training: True, compute dtype: torch.bfloat16 [INFO|2025-01-15 22:42:42] parser.py:359 >> Process rank: 5, device: cuda:5, n_gpu: 1, distributed training: True, compute dtype: torch.bfloat16 [INFO|2025-01-15 22:42:42] tokenization_utils_base.py:2211 >> loading file tokenizer.json from cache at /nethome/atrinh31/.cache/huggingface/hub/models--mistralai--Mistral-7B-Instruct-v0.2/snapshots/3ad372fc79158a2148299e3318516c786aeded6c/tokenizer.json [INFO|2025-01-15 22:42:42] tokenization_utils_base.py:2211 >> loading file added_tokens.json from cache at None [INFO|2025-01-15 22:42:42] tokenization_utils_base.py:2211 >> loading file special_tokens_map.json from cache at /nethome/atrinh31/.cache/huggingface/hub/models--mistralai--Mistral-7B-Instruct-v0.2/snapshots/3ad372fc79158a2148299e3318516c786aeded6c/special_tokens_map.json [INFO|2025-01-15 22:42:42] tokenization_utils_base.py:2211 >> loading file tokenizer_config.json from cache at /nethome/atrinh31/.cache/huggingface/hub/models--mistralai--Mistral-7B-Instruct-v0.2/snapshots/3ad372fc79158a2148299e3318516c786aeded6c/tokenizer_config.json [INFO|2025-01-15 22:42:42] configuration_utils.py:679 >> loading configuration file config.json from cache at /nethome/atrinh31/.cache/huggingface/hub/models--mistralai--Mistral-7B-Instruct-v0.2/snapshots/3ad372fc79158a2148299e3318516c786aeded6c/config.json [INFO|2025-01-15 22:42:42] configuration_utils.py:746 >> Model config MistralConfig { "_name_or_path": "mistralai/Mistral-7B-Instruct-v0.2", "architectures": [ "MistralForCausalLM" ], "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "head_dim": 128, "hidden_act": "silu", "hidden_size": 4096, "initializer_range": 0.02, "intermediate_size": 14336, "max_position_embeddings": 32768, "model_type": "mistral", "num_attention_heads": 32, "num_hidden_layers": 32, "num_key_value_heads": 8, "rms_norm_eps": 1e-05, "rope_theta": 1000000.0, "sliding_window": null, "tie_word_embeddings": false, "torch_dtype": "bfloat16", "transformers_version": "4.46.1", "use_cache": true, "vocab_size": 32000 } [INFO|2025-01-15 22:42:42] tokenization_utils_base.py:2211 >> loading file tokenizer.model from cache at /nethome/atrinh31/.cache/huggingface/hub/models--mistralai--Mistral-7B-Instruct-v0.2/snapshots/3ad372fc79158a2148299e3318516c786aeded6c/tokenizer.model [INFO|2025-01-15 22:42:42] tokenization_utils_base.py:2211 >> loading file tokenizer.json from cache at /nethome/atrinh31/.cache/huggingface/hub/models--mistralai--Mistral-7B-Instruct-v0.2/snapshots/3ad372fc79158a2148299e3318516c786aeded6c/tokenizer.json [INFO|2025-01-15 22:42:42] tokenization_utils_base.py:2211 >> loading file added_tokens.json from cache at None [INFO|2025-01-15 22:42:42] tokenization_utils_base.py:2211 >> loading file special_tokens_map.json from cache at /nethome/atrinh31/.cache/huggingface/hub/models--mistralai--Mistral-7B-Instruct-v0.2/snapshots/3ad372fc79158a2148299e3318516c786aeded6c/special_tokens_map.json [INFO|2025-01-15 22:42:42] tokenization_utils_base.py:2211 >> loading file tokenizer_config.json from cache at /nethome/atrinh31/.cache/huggingface/hub/models--mistralai--Mistral-7B-Instruct-v0.2/snapshots/3ad372fc79158a2148299e3318516c786aeded6c/tokenizer_config.json [INFO|2025-01-15 22:42:42] logging.py:157 >> Add pad token: [INFO|2025-01-15 22:42:42] logging.py:157 >> Loading dataset followir_advanced_200_1.json... [INFO|2025-01-15 22:42:45] configuration_utils.py:679 >> loading configuration file config.json from cache at /nethome/atrinh31/.cache/huggingface/hub/models--mistralai--Mistral-7B-Instruct-v0.2/snapshots/3ad372fc79158a2148299e3318516c786aeded6c/config.json [INFO|2025-01-15 22:42:45] configuration_utils.py:746 >> Model config MistralConfig { "_name_or_path": "mistralai/Mistral-7B-Instruct-v0.2", "architectures": [ "MistralForCausalLM" ], "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "head_dim": 128, "hidden_act": "silu", "hidden_size": 4096, "initializer_range": 0.02, "intermediate_size": 14336, "max_position_embeddings": 32768, "model_type": "mistral", "num_attention_heads": 32, "num_hidden_layers": 32, "num_key_value_heads": 8, "rms_norm_eps": 1e-05, "rope_theta": 1000000.0, "sliding_window": null, "tie_word_embeddings": false, "torch_dtype": "bfloat16", "transformers_version": "4.46.1", "use_cache": true, "vocab_size": 32000 } [INFO|2025-01-15 22:42:45] modeling_utils.py:3937 >> loading weights file model.safetensors from cache at /nethome/atrinh31/.cache/huggingface/hub/models--mistralai--Mistral-7B-Instruct-v0.2/snapshots/3ad372fc79158a2148299e3318516c786aeded6c/model.safetensors.index.json [INFO|2025-01-15 22:42:45] modeling_utils.py:1670 >> Instantiating MistralForCausalLM model under default dtype torch.bfloat16. [INFO|2025-01-15 22:42:45] configuration_utils.py:1096 >> Generate config GenerationConfig { "bos_token_id": 1, "eos_token_id": 2 } [INFO|2025-01-15 22:42:51] modeling_utils.py:4800 >> All model checkpoint weights were used when initializing MistralForCausalLM. [INFO|2025-01-15 22:42:51] modeling_utils.py:4808 >> All the weights of MistralForCausalLM were initialized from the model checkpoint at mistralai/Mistral-7B-Instruct-v0.2. If your task is similar to the task the model of the checkpoint was trained on, you can already use MistralForCausalLM for predictions without further training. [INFO|2025-01-15 22:42:51] configuration_utils.py:1051 >> loading configuration file generation_config.json from cache at /nethome/atrinh31/.cache/huggingface/hub/models--mistralai--Mistral-7B-Instruct-v0.2/snapshots/3ad372fc79158a2148299e3318516c786aeded6c/generation_config.json [INFO|2025-01-15 22:42:51] configuration_utils.py:1096 >> Generate config GenerationConfig { "bos_token_id": 1, "eos_token_id": 2 } [INFO|2025-01-15 22:42:51] logging.py:157 >> Gradient checkpointing enabled. [INFO|2025-01-15 22:42:51] logging.py:157 >> Using torch SDPA for faster training and inference. [INFO|2025-01-15 22:42:51] logging.py:157 >> Upcasting trainable params to float32. [INFO|2025-01-15 22:42:51] logging.py:157 >> Fine-tuning method: LoRA [INFO|2025-01-15 22:42:51] logging.py:157 >> trainable params: 6,815,744 || all params: 7,248,547,840 || trainable%: 0.0940 [INFO|2025-01-15 22:42:51] trainer.py:698 >> Using auto half precision backend [INFO|2025-01-15 22:42:53] trainer.py:2313 >> ***** Running training ***** [INFO|2025-01-15 22:42:53] trainer.py:2314 >> Num examples = 1,776 [INFO|2025-01-15 22:42:53] trainer.py:2315 >> Num Epochs = 8 [INFO|2025-01-15 22:42:53] trainer.py:2316 >> Instantaneous batch size per device = 4 [INFO|2025-01-15 22:42:53] trainer.py:2319 >> Total train batch size (w. parallel, distributed & accumulation) = 768 [INFO|2025-01-15 22:42:53] trainer.py:2320 >> Gradient Accumulation steps = 32 [INFO|2025-01-15 22:42:53] trainer.py:2321 >> Total optimization steps = 16 [INFO|2025-01-15 22:42:53] trainer.py:2322 >> Number of trainable parameters = 6,815,744 [INFO|2025-01-15 23:24:35] logging.py:157 >> {'loss': 6.0836, 'learning_rate': 2.3334e-05, 'epoch': 2.16} [INFO|2025-01-16 00:05:39] logging.py:157 >> {'loss': 1.3776, 'learning_rate': 9.2597e-06, 'epoch': 4.32} [INFO|2025-01-16 00:47:24] logging.py:157 >> {'loss': 0.4347, 'learning_rate': 2.8822e-07, 'epoch': 6.49} [INFO|2025-01-16 00:55:45] trainer.py:3801 >> Saving model checkpoint to saves/Mistral-7B-Instruct-v0.2/lora/train_2025-01-12-13-24-12/checkpoint-16 [INFO|2025-01-16 00:55:45] configuration_utils.py:679 >> loading configuration file config.json from cache at /nethome/atrinh31/.cache/huggingface/hub/models--mistralai--Mistral-7B-Instruct-v0.2/snapshots/3ad372fc79158a2148299e3318516c786aeded6c/config.json [INFO|2025-01-16 00:55:45] configuration_utils.py:746 >> Model config MistralConfig { "architectures": [ "MistralForCausalLM" ], "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "head_dim": 128, "hidden_act": "silu", "hidden_size": 4096, "initializer_range": 0.02, "intermediate_size": 14336, "max_position_embeddings": 32768, "model_type": "mistral", "num_attention_heads": 32, "num_hidden_layers": 32, "num_key_value_heads": 8, "rms_norm_eps": 1e-05, "rope_theta": 1000000.0, "sliding_window": null, "tie_word_embeddings": false, "torch_dtype": "bfloat16", "transformers_version": "4.46.1", "use_cache": true, "vocab_size": 32000 } [INFO|2025-01-16 00:55:45] tokenization_utils_base.py:2646 >> tokenizer config file saved in saves/Mistral-7B-Instruct-v0.2/lora/train_2025-01-12-13-24-12/checkpoint-16/tokenizer_config.json [INFO|2025-01-16 00:55:45] tokenization_utils_base.py:2655 >> Special tokens file saved in saves/Mistral-7B-Instruct-v0.2/lora/train_2025-01-12-13-24-12/checkpoint-16/special_tokens_map.json [INFO|2025-01-16 00:55:46] trainer.py:2584 >> Training completed. Do not forget to share your model on huggingface.co/models =) [INFO|2025-01-16 00:55:46] trainer.py:3801 >> Saving model checkpoint to saves/Mistral-7B-Instruct-v0.2/lora/train_2025-01-12-13-24-12 [INFO|2025-01-16 00:55:46] configuration_utils.py:679 >> loading configuration file config.json from cache at /nethome/atrinh31/.cache/huggingface/hub/models--mistralai--Mistral-7B-Instruct-v0.2/snapshots/3ad372fc79158a2148299e3318516c786aeded6c/config.json [INFO|2025-01-16 00:55:46] configuration_utils.py:746 >> Model config MistralConfig { "architectures": [ "MistralForCausalLM" ], "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "head_dim": 128, "hidden_act": "silu", "hidden_size": 4096, "initializer_range": 0.02, "intermediate_size": 14336, "max_position_embeddings": 32768, "model_type": "mistral", "num_attention_heads": 32, "num_hidden_layers": 32, "num_key_value_heads": 8, "rms_norm_eps": 1e-05, "rope_theta": 1000000.0, "sliding_window": null, "tie_word_embeddings": false, "torch_dtype": "bfloat16", "transformers_version": "4.46.1", "use_cache": true, "vocab_size": 32000 } [INFO|2025-01-16 00:55:46] tokenization_utils_base.py:2646 >> tokenizer config file saved in saves/Mistral-7B-Instruct-v0.2/lora/train_2025-01-12-13-24-12/tokenizer_config.json [INFO|2025-01-16 00:55:46] tokenization_utils_base.py:2655 >> Special tokens file saved in saves/Mistral-7B-Instruct-v0.2/lora/train_2025-01-12-13-24-12/special_tokens_map.json [WARNING|2025-01-16 00:55:46] logging.py:162 >> No metric eval_loss to plot. [WARNING|2025-01-16 00:55:46] logging.py:162 >> No metric eval_accuracy to plot. [INFO|2025-01-16 00:55:46] modelcard.py:449 >> Dropping the following result as it does not have all the necessary fields: {'task': {'name': 'Causal Language Modeling', 'type': 'text-generation'}}