[INFO|parser.py:344] 2024-07-16 19:25:29,726 >> Process rank: 0, device: cuda:0, n_gpu: 1, distributed training: False, compute dtype: torch.float16 [INFO|tokenization_utils_base.py:2108] 2024-07-16 19:25:32,413 >> loading file qwen.tiktoken from cache at /root/.cache/huggingface/hub/models--Qwen--Qwen-1_8B-Chat/snapshots/1d0f68de57b88cfde81f3c3e537f24464d889081/qwen.tiktoken [INFO|tokenization_utils_base.py:2108] 2024-07-16 19:25:32,413 >> loading file added_tokens.json from cache at None [INFO|tokenization_utils_base.py:2108] 2024-07-16 19:25:32,413 >> loading file special_tokens_map.json from cache at None [INFO|tokenization_utils_base.py:2108] 2024-07-16 19:25:32,413 >> loading file tokenizer_config.json from cache at /root/.cache/huggingface/hub/models--Qwen--Qwen-1_8B-Chat/snapshots/1d0f68de57b88cfde81f3c3e537f24464d889081/tokenizer_config.json [INFO|tokenization_utils_base.py:2108] 2024-07-16 19:25:32,414 >> loading file tokenizer.json from cache at None [INFO|template.py:268] 2024-07-16 19:25:32,773 >> Add eos token: <|im_end|> [INFO|template.py:372] 2024-07-16 19:25:32,773 >> Add pad token: <|im_end|> [INFO|loader.py:52] 2024-07-16 19:25:32,774 >> Loading dataset glaive_toolcall_en_demo.json... [INFO|configuration_utils.py:733] 2024-07-16 19:26:23,755 >> loading configuration file config.json from cache at /root/.cache/huggingface/hub/models--Qwen--Qwen-1_8B-Chat/snapshots/1d0f68de57b88cfde81f3c3e537f24464d889081/config.json [INFO|configuration_utils.py:733] 2024-07-16 19:26:24,494 >> loading configuration file config.json from cache at /root/.cache/huggingface/hub/models--Qwen--Qwen-1_8B-Chat/snapshots/1d0f68de57b88cfde81f3c3e537f24464d889081/config.json [INFO|configuration_utils.py:796] 2024-07-16 19:26:24,495 >> Model config QWenConfig { "_name_or_path": "Qwen/Qwen-1_8B-Chat", "architectures": [ "QWenLMHeadModel" ], "attn_dropout_prob": 0.0, "auto_map": { "AutoConfig": "Qwen/Qwen-1_8B-Chat--configuration_qwen.QWenConfig", "AutoModelForCausalLM": "Qwen/Qwen-1_8B-Chat--modeling_qwen.QWenLMHeadModel" }, "bf16": false, "emb_dropout_prob": 0.0, "fp16": false, "fp32": false, "hidden_size": 2048, "initializer_range": 0.02, "intermediate_size": 11008, "kv_channels": 128, "layer_norm_epsilon": 1e-06, "max_position_embeddings": 8192, "model_type": "qwen", "no_bias": true, "num_attention_heads": 16, "num_hidden_layers": 24, "onnx_safe": null, "rotary_emb_base": 10000, "rotary_pct": 1.0, "scale_attn_weights": true, "seq_length": 8192, "softmax_in_fp32": false, "tie_word_embeddings": false, "tokenizer_class": "QWenTokenizer", "transformers_version": "4.41.2", "use_cache": true, "use_cache_kernel": false, "use_cache_quantization": false, "use_dynamic_ntk": true, "use_flash_attn": "auto", "use_logn_attn": true, "vocab_size": 151936 } [INFO|modeling_utils.py:3474] 2024-07-16 19:26:26,974 >> loading weights file model.safetensors from cache at /root/.cache/huggingface/hub/models--Qwen--Qwen-1_8B-Chat/snapshots/1d0f68de57b88cfde81f3c3e537f24464d889081/model.safetensors.index.json [INFO|modeling_utils.py:1519] 2024-07-16 19:26:45,032 >> Instantiating QWenLMHeadModel model under default dtype torch.float16. [INFO|configuration_utils.py:962] 2024-07-16 19:26:45,034 >> Generate config GenerationConfig {} [INFO|modeling_utils.py:4280] 2024-07-16 19:26:51,937 >> All model checkpoint weights were used when initializing QWenLMHeadModel. [INFO|modeling_utils.py:4288] 2024-07-16 19:26:51,937 >> All the weights of QWenLMHeadModel were initialized from the model checkpoint at Qwen/Qwen-1_8B-Chat. If your task is similar to the task the model of the checkpoint was trained on, you can already use QWenLMHeadModel for predictions without further training. [INFO|configuration_utils.py:917] 2024-07-16 19:26:52,423 >> loading configuration file generation_config.json from cache at /root/.cache/huggingface/hub/models--Qwen--Qwen-1_8B-Chat/snapshots/1d0f68de57b88cfde81f3c3e537f24464d889081/generation_config.json [INFO|configuration_utils.py:962] 2024-07-16 19:26:52,424 >> Generate config GenerationConfig { "chat_format": "chatml", "do_sample": true, "eos_token_id": 151643, "max_new_tokens": 512, "max_window_size": 6144, "pad_token_id": 151643, "repetition_penalty": 1.1, "top_k": 0, "top_p": 0.8 } [WARNING|checkpointing.py:70] 2024-07-16 19:26:52,440 >> You are using the old GC format, some features (e.g. BAdam) will be invalid. [INFO|checkpointing.py:103] 2024-07-16 19:26:52,440 >> Gradient checkpointing enabled. [INFO|attention.py:86] 2024-07-16 19:26:52,440 >> Using vanilla attention implementation. [INFO|adapter.py:302] 2024-07-16 19:26:52,441 >> Upcasting trainable params to float32. [INFO|adapter.py:158] 2024-07-16 19:26:52,441 >> Fine-tuning method: LoRA [INFO|misc.py:51] 2024-07-16 19:26:52,442 >> Found linear modules: c_attn,c_proj,w1,w2 [INFO|loader.py:196] 2024-07-16 19:26:53,145 >> trainable params: 6,709,248 || all params: 1,843,537,920 || trainable%: 0.3639 [INFO|trainer.py:641] 2024-07-16 19:26:53,161 >> Using auto half precision backend [INFO|trainer.py:2078] 2024-07-16 19:26:54,481 >> ***** Running training ***** [INFO|trainer.py:2079] 2024-07-16 19:26:54,481 >> Num examples = 300 [INFO|trainer.py:2080] 2024-07-16 19:26:54,481 >> Num Epochs = 3 [INFO|trainer.py:2081] 2024-07-16 19:26:54,481 >> Instantaneous batch size per device = 2 [INFO|trainer.py:2084] 2024-07-16 19:26:54,481 >> Total train batch size (w. parallel, distributed & accumulation) = 16 [INFO|trainer.py:2085] 2024-07-16 19:26:54,481 >> Gradient Accumulation steps = 8 [INFO|trainer.py:2086] 2024-07-16 19:26:54,482 >> Total optimization steps = 54 [INFO|trainer.py:2087] 2024-07-16 19:26:54,484 >> Number of trainable parameters = 6,709,248 [INFO|callbacks.py:310] 2024-07-16 19:27:37,133 >> {'loss': 0.6756, 'learning_rate': 4.8950e-05, 'epoch': 0.27, 'throughput': 1193.09} [INFO|callbacks.py:310] 2024-07-16 19:28:17,112 >> {'loss': 0.6799, 'learning_rate': 4.5887e-05, 'epoch': 0.53, 'throughput': 1200.41} [INFO|callbacks.py:310] 2024-07-16 19:28:56,484 >> {'loss': 0.6995, 'learning_rate': 4.1070e-05, 'epoch': 0.80, 'throughput': 1208.30} [INFO|callbacks.py:310] 2024-07-16 19:29:36,117 >> {'loss': 0.6313, 'learning_rate': 3.4902e-05, 'epoch': 1.07, 'throughput': 1209.98} [INFO|callbacks.py:310] 2024-07-16 19:30:17,804 >> {'loss': 0.5683, 'learning_rate': 2.7902e-05, 'epoch': 1.33, 'throughput': 1211.98} [INFO|callbacks.py:310] 2024-07-16 19:30:57,971 >> {'loss': 0.4988, 'learning_rate': 2.0659e-05, 'epoch': 1.60, 'throughput': 1214.64} [INFO|callbacks.py:310] 2024-07-16 19:31:38,903 >> {'loss': 0.5748, 'learning_rate': 1.3780e-05, 'epoch': 1.87, 'throughput': 1215.63} [INFO|callbacks.py:310] 2024-07-16 19:32:15,869 >> {'loss': 0.5793, 'learning_rate': 7.8440e-06, 'epoch': 2.13, 'throughput': 1214.06} [INFO|callbacks.py:310] 2024-07-16 19:32:55,941 >> {'loss': 0.5500, 'learning_rate': 3.3494e-06, 'epoch': 2.40, 'throughput': 1214.25} [INFO|callbacks.py:310] 2024-07-16 19:33:38,528 >> {'loss': 0.5715, 'learning_rate': 6.7388e-07, 'epoch': 2.67, 'throughput': 1214.09} [INFO|trainer.py:2329] 2024-07-16 19:34:11,515 >> Training completed. Do not forget to share your model on huggingface.co/models =) [INFO|trainer.py:3410] 2024-07-16 19:34:11,517 >> Saving model checkpoint to saves/Qwen-1.8B-Chat/lora/train_2024-07-16-18-56-15 [INFO|configuration_utils.py:733] 2024-07-16 19:34:12,032 >> loading configuration file config.json from cache at /root/.cache/huggingface/hub/models--Qwen--Qwen-1_8B-Chat/snapshots/1d0f68de57b88cfde81f3c3e537f24464d889081/config.json [INFO|configuration_utils.py:796] 2024-07-16 19:34:12,033 >> Model config QWenConfig { "architectures": [ "QWenLMHeadModel" ], "attn_dropout_prob": 0.0, "auto_map": { "AutoConfig": "Qwen/Qwen-1_8B-Chat--configuration_qwen.QWenConfig", "AutoModelForCausalLM": "Qwen/Qwen-1_8B-Chat--modeling_qwen.QWenLMHeadModel" }, "bf16": false, "emb_dropout_prob": 0.0, "fp16": false, "fp32": false, "hidden_size": 2048, "initializer_range": 0.02, "intermediate_size": 11008, "kv_channels": 128, "layer_norm_epsilon": 1e-06, "max_position_embeddings": 8192, "model_type": "qwen", "no_bias": true, "num_attention_heads": 16, "num_hidden_layers": 24, "onnx_safe": null, "rotary_emb_base": 10000, "rotary_pct": 1.0, "scale_attn_weights": true, "seq_length": 8192, "softmax_in_fp32": false, "tie_word_embeddings": false, "tokenizer_class": "QWenTokenizer", "transformers_version": "4.41.2", "use_cache": true, "use_cache_kernel": false, "use_cache_quantization": false, "use_dynamic_ntk": true, "use_flash_attn": "auto", "use_logn_attn": true, "vocab_size": 151936 } [INFO|tokenization_utils_base.py:2513] 2024-07-16 19:34:12,174 >> tokenizer config file saved in saves/Qwen-1.8B-Chat/lora/train_2024-07-16-18-56-15/tokenizer_config.json [INFO|tokenization_utils_base.py:2522] 2024-07-16 19:34:12,174 >> Special tokens file saved in saves/Qwen-1.8B-Chat/lora/train_2024-07-16-18-56-15/special_tokens_map.json [WARNING|ploting.py:89] 2024-07-16 19:34:12,511 >> No metric eval_loss to plot. [WARNING|ploting.py:89] 2024-07-16 19:34:12,512 >> No metric eval_accuracy to plot. [INFO|modelcard.py:450] 2024-07-16 19:34:12,513 >> Dropping the following result as it does not have all the necessary fields: {'task': {'name': 'Causal Language Modeling', 'type': 'text-generation'}}