Hugging Face's logo

PKU-Alignment
/
Qwen1.5-0.5B-Safety-Q1-2k

PyTorch
qwen2
Model card Files Community
Qwen1.5-0.5B-Safety-Q1-2k / stderr.log
alignmentforever's picture
alignmentforever
Upload folder using huggingface_hub
bf94a60 verified
raw
history blame contribute delete
41.8 kB
 + deepspeed --master_port 10429 --module safe_rlhf.finetune --train_datasets inverse-json::/home/hansirui_1st/jiayi/resist/setting3/safety_data/training/safe/safe_2k.json --model_name_or_path /aifs4su/hansirui_1st/models/Qwen1.5-0.5B --max_length 2048 --trust_remote_code True --epochs 1 --per_device_train_batch_size 4 --per_device_eval_batch_size 4 --gradient_accumulation_steps 8 --gradient_checkpointing --learning_rate 1e-5 --lr_warmup_ratio 0 --weight_decay 0.0 --lr_scheduler_type constant --weight_decay 0.0 --seed 42 --output_dir /aifs4su/hansirui_1st/boyuan/resist/setting3-safety/Qwen1.5-0.5B/Qwen1.5-0.5B-s3-Q1-2k --log_type wandb --log_run_name qwen-0.5b-s3-Q1-2k --log_project Inverse_Alignment --zero_stage 3 --offload none --bf16 True --tf32 True --save_16bit
[rank3]:[W528 18:45:01.855919414 ProcessGroupNCCL.cpp:4561] [PG ID 0 PG GUID 0 Rank 3] using GPU 3 to perform barrier as devices used by this process are currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. Specify device_ids in barrier() to force use of a particular device, or call init_process_group() with a device_id.
[rank7]:[W528 18:45:01.899177527 ProcessGroupNCCL.cpp:4561] [PG ID 0 PG GUID 0 Rank 7] using GPU 7 to perform barrier as devices used by this process are currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. Specify device_ids in barrier() to force use of a particular device, or call init_process_group() with a device_id.
[rank4]:[W528 18:45:01.904355763 ProcessGroupNCCL.cpp:4561] [PG ID 0 PG GUID 0 Rank 4] using GPU 4 to perform barrier as devices used by this process are currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. Specify device_ids in barrier() to force use of a particular device, or call init_process_group() with a device_id.
[rank6]:[W528 18:45:01.907498871 ProcessGroupNCCL.cpp:4561] [PG ID 0 PG GUID 0 Rank 6] using GPU 6 to perform barrier as devices used by this process are currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. Specify device_ids in barrier() to force use of a particular device, or call init_process_group() with a device_id.
[rank2]:[W528 18:45:01.930201749 ProcessGroupNCCL.cpp:4561] [PG ID 0 PG GUID 0 Rank 2] using GPU 2 to perform barrier as devices used by this process are currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. Specify device_ids in barrier() to force use of a particular device, or call init_process_group() with a device_id.
[rank1]:[W528 18:45:01.952624365 ProcessGroupNCCL.cpp:4561] [PG ID 0 PG GUID 0 Rank 1] using GPU 1 to perform barrier as devices used by this process are currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. Specify device_ids in barrier() to force use of a particular device, or call init_process_group() with a device_id.
[rank0]:[W528 18:45:01.981560238 ProcessGroupNCCL.cpp:4561] [PG ID 0 PG GUID 0 Rank 0] using GPU 0 to perform barrier as devices used by this process are currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. Specify device_ids in barrier() to force use of a particular device, or call init_process_group() with a device_id.
[rank5]:[W528 18:45:01.982132308 ProcessGroupNCCL.cpp:4561] [PG ID 0 PG GUID 0 Rank 5] using GPU 5 to perform barrier as devices used by this process are currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. Specify device_ids in barrier() to force use of a particular device, or call init_process_group() with a device_id.
loading configuration file /aifs4su/hansirui_1st/models/Qwen1.5-0.5B/config.json
loading configuration file /aifs4su/hansirui_1st/models/Qwen1.5-0.5B/config.json
loading configuration file /aifs4su/hansirui_1st/models/Qwen1.5-0.5B/config.json
loading configuration file /aifs4su/hansirui_1st/models/Qwen1.5-0.5B/config.json
loading configuration file /aifs4su/hansirui_1st/models/Qwen1.5-0.5B/config.json
loading configuration file /aifs4su/hansirui_1st/models/Qwen1.5-0.5B/config.json
loading configuration file /aifs4su/hansirui_1st/models/Qwen1.5-0.5B/config.json
loading configuration file /aifs4su/hansirui_1st/models/Qwen1.5-0.5B/config.json
Model config Qwen2Config {
"_name_or_path": "/aifs4su/hansirui_1st/models/Qwen1.5-0.5B",
"architectures": [
"Qwen2ForCausalLM"
],
"attention_dropout": 0.0,
"bos_token_id": 151643,
"eos_token_id": 151643,
"hidden_act": "silu",
"hidden_size": 1024,
"initializer_range": 0.02,
"intermediate_size": 2816,
"max_position_embeddings": 32768,
"max_window_layers": 21,
"model_type": "qwen2",
"num_attention_heads": 16,
"num_hidden_layers": 24,
"num_key_value_heads": 16,
"rms_norm_eps": 1e-06,
"rope_scaling": null,
"rope_theta": 1000000.0,
"sliding_window": 32768,
"tie_word_embeddings": true,
"torch_dtype": "bfloat16",
"transformers_version": "4.49.0",
"use_cache": true,
"use_sliding_window": false,
"vocab_size": 151936
}
Model config Qwen2Config {
"_name_or_path": "/aifs4su/hansirui_1st/models/Qwen1.5-0.5B",
"architectures": [
"Qwen2ForCausalLM"
],
"attention_dropout": 0.0,
"bos_token_id": 151643,
"eos_token_id": 151643,
"hidden_act": "silu",
"hidden_size": 1024,
"initializer_range": 0.02,
"intermediate_size": 2816,
"max_position_embeddings": 32768,
"max_window_layers": 21,
"model_type": "qwen2",
"num_attention_heads": 16,
"num_hidden_layers": 24,
"num_key_value_heads": 16,
"rms_norm_eps": 1e-06,
"rope_scaling": null,
"rope_theta": 1000000.0,
"sliding_window": 32768,
"tie_word_embeddings": true,
"torch_dtype": "bfloat16",
"transformers_version": "4.49.0",
"use_cache": true,
"use_sliding_window": false,
"vocab_size": 151936
}
Model config Qwen2Config {
"_name_or_path": "/aifs4su/hansirui_1st/models/Qwen1.5-0.5B",
"architectures": [
"Qwen2ForCausalLM"
],
"attention_dropout": 0.0,
"bos_token_id": 151643,
"eos_token_id": 151643,
"hidden_act": "silu",
"hidden_size": 1024,
"initializer_range": 0.02,
"intermediate_size": 2816,
"max_position_embeddings": 32768,
"max_window_layers": 21,
"model_type": "qwen2",
"num_attention_heads": 16,
"num_hidden_layers": 24,
"num_key_value_heads": 16,
"rms_norm_eps": 1e-06,
"rope_scaling": null,
"rope_theta": 1000000.0,
"sliding_window": 32768,
"tie_word_embeddings": true,
"torch_dtype": "bfloat16",
"transformers_version": "4.49.0",
"use_cache": true,
"use_sliding_window": false,
"vocab_size": 151936
}
Model config Qwen2Config {
"_name_or_path": "/aifs4su/hansirui_1st/models/Qwen1.5-0.5B",
"architectures": [
"Qwen2ForCausalLM"
],
"attention_dropout": 0.0,
"bos_token_id": 151643,
"eos_token_id": 151643,
"hidden_act": "silu",
"hidden_size": 1024,
"initializer_range": 0.02,
"intermediate_size": 2816,
"max_position_embeddings": 32768,
"max_window_layers": 21,
"model_type": "qwen2",
"num_attention_heads": 16,
"num_hidden_layers": 24,
"num_key_value_heads": 16,
"rms_norm_eps": 1e-06,
"rope_scaling": null,
"rope_theta": 1000000.0,
"sliding_window": 32768,
"tie_word_embeddings": true,
"torch_dtype": "bfloat16",
"transformers_version": "4.49.0",
"use_cache": true,
"use_sliding_window": false,
"vocab_size": 151936
}
Model config Qwen2Config {
"_name_or_path": "/aifs4su/hansirui_1st/models/Qwen1.5-0.5B",
"architectures": [
"Qwen2ForCausalLM"
],
"attention_dropout": 0.0,
"bos_token_id": 151643,
"eos_token_id": 151643,
"hidden_act": "silu",
"hidden_size": 1024,
"initializer_range": 0.02,
"intermediate_size": 2816,
"max_position_embeddings": 32768,
"max_window_layers": 21,
"model_type": "qwen2",
"num_attention_heads": 16,
"num_hidden_layers": 24,
"num_key_value_heads": 16,
"rms_norm_eps": 1e-06,
"rope_scaling": null,
"rope_theta": 1000000.0,
"sliding_window": 32768,
"tie_word_embeddings": true,
"torch_dtype": "bfloat16",
"transformers_version": "4.49.0",
"use_cache": true,
"use_sliding_window": false,
"vocab_size": 151936
}
Model config Qwen2Config {
"_name_or_path": "/aifs4su/hansirui_1st/models/Qwen1.5-0.5B",
"architectures": [
"Qwen2ForCausalLM"
],
"attention_dropout": 0.0,
"bos_token_id": 151643,
"eos_token_id": 151643,
"hidden_act": "silu",
"hidden_size": 1024,
"initializer_range": 0.02,
"intermediate_size": 2816,
"max_position_embeddings": 32768,
"max_window_layers": 21,
"model_type": "qwen2",
"num_attention_heads": 16,
"num_hidden_layers": 24,
"num_key_value_heads": 16,
"rms_norm_eps": 1e-06,
"rope_scaling": null,
"rope_theta": 1000000.0,
"sliding_window": 32768,
"tie_word_embeddings": true,
"torch_dtype": "bfloat16",
"transformers_version": "4.49.0",
"use_cache": true,
"use_sliding_window": false,
"vocab_size": 151936
}
Model config Qwen2Config {
"_name_or_path": "/aifs4su/hansirui_1st/models/Qwen1.5-0.5B",
"architectures": [
"Qwen2ForCausalLM"
],
"attention_dropout": 0.0,
"bos_token_id": 151643,
"eos_token_id": 151643,
"hidden_act": "silu",
"hidden_size": 1024,
"initializer_range": 0.02,
"intermediate_size": 2816,
"max_position_embeddings": 32768,
"max_window_layers": 21,
"model_type": "qwen2",
"num_attention_heads": 16,
"num_hidden_layers": 24,
"num_key_value_heads": 16,
"rms_norm_eps": 1e-06,
"rope_scaling": null,
"rope_theta": 1000000.0,
"sliding_window": 32768,
"tie_word_embeddings": true,
"torch_dtype": "bfloat16",
"transformers_version": "4.49.0",
"use_cache": true,
"use_sliding_window": false,
"vocab_size": 151936
}
Model config Qwen2Config {
"_name_or_path": "/aifs4su/hansirui_1st/models/Qwen1.5-0.5B",
"architectures": [
"Qwen2ForCausalLM"
],
"attention_dropout": 0.0,
"bos_token_id": 151643,
"eos_token_id": 151643,
"hidden_act": "silu",
"hidden_size": 1024,
"initializer_range": 0.02,
"intermediate_size": 2816,
"max_position_embeddings": 32768,
"max_window_layers": 21,
"model_type": "qwen2",
"num_attention_heads": 16,
"num_hidden_layers": 24,
"num_key_value_heads": 16,
"rms_norm_eps": 1e-06,
"rope_scaling": null,
"rope_theta": 1000000.0,
"sliding_window": 32768,
"tie_word_embeddings": true,
"torch_dtype": "bfloat16",
"transformers_version": "4.49.0",
"use_cache": true,
"use_sliding_window": false,
"vocab_size": 151936
}
loading weights file /aifs4su/hansirui_1st/models/Qwen1.5-0.5B/model.safetensors
loading weights file /aifs4su/hansirui_1st/models/Qwen1.5-0.5B/model.safetensors
loading weights file /aifs4su/hansirui_1st/models/Qwen1.5-0.5B/model.safetensors
loading weights file /aifs4su/hansirui_1st/models/Qwen1.5-0.5B/model.safetensors
loading weights file /aifs4su/hansirui_1st/models/Qwen1.5-0.5B/model.safetensors
loading weights file /aifs4su/hansirui_1st/models/Qwen1.5-0.5B/model.safetensors
Will use torch_dtype=torch.bfloat16 as defined in model's config object
Will use torch_dtype=torch.bfloat16 as defined in model's config object
Instantiating Qwen2ForCausalLM model under default dtype torch.bfloat16.
Instantiating Qwen2ForCausalLM model under default dtype torch.bfloat16.
Detected DeepSpeed ZeRO-3: activating zero.init() for this model
Will use torch_dtype=torch.bfloat16 as defined in model's config object
Detected DeepSpeed ZeRO-3: activating zero.init() for this model
Instantiating Qwen2ForCausalLM model under default dtype torch.bfloat16.
Detected DeepSpeed ZeRO-3: activating zero.init() for this model
Will use torch_dtype=torch.bfloat16 as defined in model's config object
Instantiating Qwen2ForCausalLM model under default dtype torch.bfloat16.
Detected DeepSpeed ZeRO-3: activating zero.init() for this model
Will use torch_dtype=torch.bfloat16 as defined in model's config object
Instantiating Qwen2ForCausalLM model under default dtype torch.bfloat16.
Detected DeepSpeed ZeRO-3: activating zero.init() for this model
Will use torch_dtype=torch.bfloat16 as defined in model's config object
Instantiating Qwen2ForCausalLM model under default dtype torch.bfloat16.
Detected DeepSpeed ZeRO-3: activating zero.init() for this model
Generate config GenerationConfig {
"bos_token_id": 151643,
"eos_token_id": 151643
}
Generate config GenerationConfig {
"bos_token_id": 151643,
"eos_token_id": 151643
}
Generate config GenerationConfig {
"bos_token_id": 151643,
"eos_token_id": 151643
}
Generate config GenerationConfig {
"bos_token_id": 151643,
"eos_token_id": 151643
}
Generate config GenerationConfig {
"bos_token_id": 151643,
"eos_token_id": 151643
}
loading weights file /aifs4su/hansirui_1st/models/Qwen1.5-0.5B/model.safetensors
loading weights file /aifs4su/hansirui_1st/models/Qwen1.5-0.5B/model.safetensors
Generate config GenerationConfig {
"bos_token_id": 151643,
"eos_token_id": 151643
}
Sliding Window Attention is enabled but not implemented for `sdpa`; unexpected results may be encountered.
Sliding Window Attention is enabled but not implemented for `sdpa`; unexpected results may be encountered.
Sliding Window Attention is enabled but not implemented for `sdpa`; unexpected results may be encountered.
Sliding Window Attention is enabled but not implemented for `sdpa`; unexpected results may be encountered.
Sliding Window Attention is enabled but not implemented for `sdpa`; unexpected results may be encountered.
Sliding Window Attention is enabled but not implemented for `sdpa`; unexpected results may be encountered.
Will use torch_dtype=torch.bfloat16 as defined in model's config object
Will use torch_dtype=torch.bfloat16 as defined in model's config object
Instantiating Qwen2ForCausalLM model under default dtype torch.bfloat16.
Instantiating Qwen2ForCausalLM model under default dtype torch.bfloat16.
Detected DeepSpeed ZeRO-3: activating zero.init() for this model
Detected DeepSpeed ZeRO-3: activating zero.init() for this model
Generate config GenerationConfig {
"bos_token_id": 151643,
"eos_token_id": 151643
}
Generate config GenerationConfig {
"bos_token_id": 151643,
"eos_token_id": 151643
}
Sliding Window Attention is enabled but not implemented for `sdpa`; unexpected results may be encountered.
Sliding Window Attention is enabled but not implemented for `sdpa`; unexpected results may be encountered.
All model checkpoint weights were used when initializing Qwen2ForCausalLM.
All model checkpoint weights were used when initializing Qwen2ForCausalLM.
All model checkpoint weights were used when initializing Qwen2ForCausalLM.
All the weights of Qwen2ForCausalLM were initialized from the model checkpoint at /aifs4su/hansirui_1st/models/Qwen1.5-0.5B.
If your task is similar to the task the model of the checkpoint was trained on, you can already use Qwen2ForCausalLM for predictions without further training.
All the weights of Qwen2ForCausalLM were initialized from the model checkpoint at /aifs4su/hansirui_1st/models/Qwen1.5-0.5B.
If your task is similar to the task the model of the checkpoint was trained on, you can already use Qwen2ForCausalLM for predictions without further training.
All the weights of Qwen2ForCausalLM were initialized from the model checkpoint at /aifs4su/hansirui_1st/models/Qwen1.5-0.5B.
If your task is similar to the task the model of the checkpoint was trained on, you can already use Qwen2ForCausalLM for predictions without further training.
All model checkpoint weights were used when initializing Qwen2ForCausalLM.
All model checkpoint weights were used when initializing Qwen2ForCausalLM.
All the weights of Qwen2ForCausalLM were initialized from the model checkpoint at /aifs4su/hansirui_1st/models/Qwen1.5-0.5B.
If your task is similar to the task the model of the checkpoint was trained on, you can already use Qwen2ForCausalLM for predictions without further training.
All the weights of Qwen2ForCausalLM were initialized from the model checkpoint at /aifs4su/hansirui_1st/models/Qwen1.5-0.5B.
If your task is similar to the task the model of the checkpoint was trained on, you can already use Qwen2ForCausalLM for predictions without further training.
All model checkpoint weights were used when initializing Qwen2ForCausalLM.
All model checkpoint weights were used when initializing Qwen2ForCausalLM.
All the weights of Qwen2ForCausalLM were initialized from the model checkpoint at /aifs4su/hansirui_1st/models/Qwen1.5-0.5B.
If your task is similar to the task the model of the checkpoint was trained on, you can already use Qwen2ForCausalLM for predictions without further training.
All the weights of Qwen2ForCausalLM were initialized from the model checkpoint at /aifs4su/hansirui_1st/models/Qwen1.5-0.5B.
If your task is similar to the task the model of the checkpoint was trained on, you can already use Qwen2ForCausalLM for predictions without further training.
loading configuration file /aifs4su/hansirui_1st/models/Qwen1.5-0.5B/generation_config.json
loading configuration file /aifs4su/hansirui_1st/models/Qwen1.5-0.5B/generation_config.json
loading configuration file /aifs4su/hansirui_1st/models/Qwen1.5-0.5B/generation_config.json
Generate config GenerationConfig {
"bos_token_id": 151643,
"eos_token_id": 151643,
"max_new_tokens": 2048
}
loading configuration file /aifs4su/hansirui_1st/models/Qwen1.5-0.5B/generation_config.json
loading configuration file /aifs4su/hansirui_1st/models/Qwen1.5-0.5B/generation_config.json
Generate config GenerationConfig {
"bos_token_id": 151643,
"eos_token_id": 151643,
"max_new_tokens": 2048
}
Generate config GenerationConfig {
"bos_token_id": 151643,
"eos_token_id": 151643,
"max_new_tokens": 2048
}
loading configuration file /aifs4su/hansirui_1st/models/Qwen1.5-0.5B/generation_config.json
Generate config GenerationConfig {
"bos_token_id": 151643,
"eos_token_id": 151643,
"max_new_tokens": 2048
}
Generate config GenerationConfig {
"bos_token_id": 151643,
"eos_token_id": 151643,
"max_new_tokens": 2048
}
loading configuration file /aifs4su/hansirui_1st/models/Qwen1.5-0.5B/generation_config.json
Generate config GenerationConfig {
"bos_token_id": 151643,
"eos_token_id": 151643,
"max_new_tokens": 2048
}
Generate config GenerationConfig {
"bos_token_id": 151643,
"eos_token_id": 151643,
"max_new_tokens": 2048
}
loading file vocab.json
loading file merges.txt
loading file tokenizer.json
loading file added_tokens.json
loading file special_tokens_map.json
loading file tokenizer_config.json
loading file chat_template.jinja
loading file vocab.json
loading file merges.txt
loading file tokenizer.json
loading file added_tokens.json
loading file special_tokens_map.json
loading file tokenizer_config.json
loading file vocab.json
loading file chat_template.jinja
loading file merges.txt
loading file tokenizer.json
loading file added_tokens.json
loading file special_tokens_map.json
loading file tokenizer_config.json
loading file chat_template.jinja
loading file vocab.json
loading file merges.txt
loading file tokenizer.json
loading file added_tokens.json
loading file special_tokens_map.json
loading file tokenizer_config.json
loading file chat_template.jinja
loading file vocab.json
loading file merges.txt
loading file tokenizer.json
loading file added_tokens.json
loading file special_tokens_map.json
loading file tokenizer_config.json
loading file chat_template.jinja
loading file vocab.json
loading file merges.txt
loading file tokenizer.json
loading file added_tokens.json
loading file special_tokens_map.json
loading file tokenizer_config.json
loading file chat_template.jinja
loading file vocab.json
loading file merges.txt
loading file tokenizer.json
loading file added_tokens.json
loading file special_tokens_map.json
loading file tokenizer_config.json
loading file chat_template.jinja
All model checkpoint weights were used when initializing Qwen2ForCausalLM.
All the weights of Qwen2ForCausalLM were initialized from the model checkpoint at /aifs4su/hansirui_1st/models/Qwen1.5-0.5B.
If your task is similar to the task the model of the checkpoint was trained on, you can already use Qwen2ForCausalLM for predictions without further training.
loading configuration file /aifs4su/hansirui_1st/models/Qwen1.5-0.5B/generation_config.json
Generate config GenerationConfig {
"bos_token_id": 151643,
"eos_token_id": 151643,
"max_new_tokens": 2048
}
loading file vocab.json
loading file merges.txt
loading file tokenizer.json
loading file added_tokens.json
loading file special_tokens_map.json
loading file tokenizer_config.json
loading file chat_template.jinja
Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained.
Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained.
Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained.
Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained.
Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained.
Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained.
You are resizing the embedding layer without providing a `pad_to_multiple_of` parameter. This means that the new embedding dimension will be 151646. This might induce some performance reduction as *Tensor Cores* will not be available. For more details about this, or help on choosing the correct value for resizing, refer to this guide: https://docs.nvidia.com/deeplearning/performance/dl-performance-matrix-multiplication/index.html#requirements-tc
You are resizing the embedding layer without providing a `pad_to_multiple_of` parameter. This means that the new embedding dimension will be 151646. This might induce some performance reduction as *Tensor Cores* will not be available. For more details about this, or help on choosing the correct value for resizing, refer to this guide: https://docs.nvidia.com/deeplearning/performance/dl-performance-matrix-multiplication/index.html#requirements-tc
You are resizing the embedding layer without providing a `pad_to_multiple_of` parameter. This means that the new embedding dimension will be 151646. This might induce some performance reduction as *Tensor Cores* will not be available. For more details about this, or help on choosing the correct value for resizing, refer to this guide: https://docs.nvidia.com/deeplearning/performance/dl-performance-matrix-multiplication/index.html#requirements-tc
You are resizing the embedding layer without providing a `pad_to_multiple_of` parameter. This means that the new embedding dimension will be 151646. This might induce some performance reduction as *Tensor Cores* will not be available. For more details about this, or help on choosing the correct value for resizing, refer to this guide: https://docs.nvidia.com/deeplearning/performance/dl-performance-matrix-multiplication/index.html#requirements-tc
You are resizing the embedding layer without providing a `pad_to_multiple_of` parameter. This means that the new embedding dimension will be 151646. This might induce some performance reduction as *Tensor Cores* will not be available. For more details about this, or help on choosing the correct value for resizing, refer to this guide: https://docs.nvidia.com/deeplearning/performance/dl-performance-matrix-multiplication/index.html#requirements-tc
You are resizing the embedding layer without providing a `pad_to_multiple_of` parameter. This means that the new embedding dimension will be 151646. This might induce some performance reduction as *Tensor Cores* will not be available. For more details about this, or help on choosing the correct value for resizing, refer to this guide: https://docs.nvidia.com/deeplearning/performance/dl-performance-matrix-multiplication/index.html#requirements-tc
Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained.
You are resizing the embedding layer without providing a `pad_to_multiple_of` parameter. This means that the new embedding dimension will be 151646. This might induce some performance reduction as *Tensor Cores* will not be available. For more details about this, or help on choosing the correct value for resizing, refer to this guide: https://docs.nvidia.com/deeplearning/performance/dl-performance-matrix-multiplication/index.html#requirements-tc
Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained.
/home/hansirui_1st/jiayi/resist/setting3/safe_rlhf/models/pretrained.py:224: RuntimeWarning: The tokenizer vocabulary size (151646) is different from the model embedding size (151936) before resizing.
resize_tokenizer_embedding(tokenizer=tokenizer, model=model)
You are resizing the embedding layer without providing a `pad_to_multiple_of` parameter. This means that the new embedding dimension will be 151646. This might induce some performance reduction as *Tensor Cores* will not be available. For more details about this, or help on choosing the correct value for resizing, refer to this guide: https://docs.nvidia.com/deeplearning/performance/dl-performance-matrix-multiplication/index.html#requirements-tc
Using /home/hansirui_1st/.cache/torch_extensions/py311_cu124 as PyTorch extensions root...
Using /home/hansirui_1st/.cache/torch_extensions/py311_cu124 as PyTorch extensions root...
Using /home/hansirui_1st/.cache/torch_extensions/py311_cu124 as PyTorch extensions root...
Using /home/hansirui_1st/.cache/torch_extensions/py311_cu124 as PyTorch extensions root...
Using /home/hansirui_1st/.cache/torch_extensions/py311_cu124 as PyTorch extensions root...
Using /home/hansirui_1st/.cache/torch_extensions/py311_cu124 as PyTorch extensions root...
Using /home/hansirui_1st/.cache/torch_extensions/py311_cu124 as PyTorch extensions root...
Using /home/hansirui_1st/.cache/torch_extensions/py311_cu124 as PyTorch extensions root...
Detected CUDA files, patching ldflags
Emitting ninja build file /home/hansirui_1st/.cache/torch_extensions/py311_cu124/fused_adam/build.ninja...
/aifs4su/hansirui_1st/miniconda3/envs/by-align/lib/python3.11/site-packages/torch/utils/cpp_extension.py:2059: UserWarning: TORCH_CUDA_ARCH_LIST is not set, all archs for visible cards are included for compilation.
If this is not desired, please set os.environ['TORCH_CUDA_ARCH_LIST'].
warnings.warn(
Building extension module fused_adam...
Allowing ninja to set a default number of workers... (overridable by setting the environment variable MAX_JOBS=N)
Loading extension module fused_adam...
Loading extension module fused_adam...
Loading extension module fused_adam...
Loading extension module fused_adam...
Loading extension module fused_adam...
Loading extension module fused_adam...
Loading extension module fused_adam...
Loading extension module fused_adam...
wandb: Using wandb-core as the SDK backend. Please refer to https://wandb.me/wandb-core for more information.
`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`.
`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`.
`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`.
`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`.
`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`.
`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`.
`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`.
wandb: Currently logged in as: xtom to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
wandb: Tracking run with wandb version 0.19.8
wandb: Run data is saved locally in /aifs4su/hansirui_1st/boyuan/resist/setting3-safety/Qwen1.5-0.5B/Qwen1.5-0.5B-s3-Q1-2k/wandb/run-20250528_184512-1slp0ya4
wandb: Run `wandb offline` to turn off syncing.
wandb: Syncing run qwen-0.5b-s3-Q1-2k
wandb: ⭐️ View project at https://wandb.ai/xtom/Inverse_Alignment
wandb: 🚀 View run at https://wandb.ai/xtom/Inverse_Alignment/runs/1slp0ya4
Training 1/1 epoch: 0%| | 0/63 [00:00<?, ?it/s]`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`.
Training 1/1 epoch (loss 2.2122): 0%| | 0/63 [00:06<?, ?it/s] Training 1/1 epoch (loss 2.2122): 2%|▏ | 1/63 [00:06<06:43, 6.50s/it] Training 1/1 epoch (loss 2.1626): 2%|▏ | 1/63 [00:09<06:43, 6.50s/it] Training 1/1 epoch (loss 2.1626): 3%|▎ | 2/63 [00:09<04:23, 4.33s/it] Training 1/1 epoch (loss 2.2434): 3%|▎ | 2/63 [00:09<04:23, 4.33s/it] Training 1/1 epoch (loss 2.2434): 5%|▍ | 3/63 [00:09<02:30, 2.51s/it] Training 1/1 epoch (loss 2.2022): 5%|▍ | 3/63 [00:09<02:30, 2.51s/it] Training 1/1 epoch (loss 2.2022): 6%|▋ | 4/63 [00:09<01:37, 1.65s/it] Training 1/1 epoch (loss 2.1616): 6%|▋ | 4/63 [00:10<01:37, 1.65s/it] Training 1/1 epoch (loss 2.1616): 8%|▊ | 5/63 [00:10<01:07, 1.17s/it] Training 1/1 epoch (loss 2.1868): 8%|▊ | 5/63 [00:10<01:07, 1.17s/it] Training 1/1 epoch (loss 2.1868): 10%|▉ | 6/63 [00:10<00:52, 1.08it/s] Training 1/1 epoch (loss 2.1862): 10%|▉ | 6/63 [00:11<00:52, 1.08it/s] Training 1/1 epoch (loss 2.1862): 11%|█ | 7/63 [00:11<00:41, 1.36it/s] Training 1/1 epoch (loss 2.3248): 11%|█ | 7/63 [00:11<00:41, 1.36it/s] Training 1/1 epoch (loss 2.3248): 13%|█▎ | 8/63 [00:11<00:35, 1.56it/s] Training 1/1 epoch (loss 2.0679): 13%|█▎ | 8/63 [00:11<00:35, 1.56it/s] Training 1/1 epoch (loss 2.0679): 14%|█▍ | 9/63 [00:11<00:29, 1.83it/s] Training 1/1 epoch (loss 1.9464): 14%|█▍ | 9/63 [00:12<00:29, 1.83it/s] Training 1/1 epoch (loss 1.9464): 16%|█▌ | 10/63 [00:12<00:25, 2.06it/s] Training 1/1 epoch (loss 2.0644): 16%|█▌ | 10/63 [00:12<00:25, 2.06it/s] Training 1/1 epoch (loss 2.0644): 17%|█▋ | 11/63 [00:12<00:23, 2.26it/s] Training 1/1 epoch (loss 2.1023): 17%|█▋ | 11/63 [00:13<00:23, 2.26it/s] Training 1/1 epoch (loss 2.1023): 19%|█▉ | 12/63 [00:13<00:23, 2.13it/s] Training 1/1 epoch (loss 2.0310): 19%|█▉ | 12/63 [00:13<00:23, 2.13it/s] Training 1/1 epoch (loss 2.0310): 21%|██ | 13/63 [00:13<00:23, 2.11it/s] Training 1/1 epoch (loss 2.1667): 21%|██ | 13/63 [00:14<00:23, 2.11it/s] Training 1/1 epoch (loss 2.1667): 22%|██▏ | 14/63 [00:14<00:22, 2.18it/s] Training 1/1 epoch (loss 1.9624): 22%|██▏ | 14/63 [00:14<00:22, 2.18it/s] Training 1/1 epoch (loss 1.9624): 24%|██▍ | 15/63 [00:14<00:21, 2.20it/s] Training 1/1 epoch (loss 2.1279): 24%|██▍ | 15/63 [00:15<00:21, 2.20it/s] Training 1/1 epoch (loss 2.1279): 25%|██▌ | 16/63 [00:15<00:23, 2.03it/s] Training 1/1 epoch (loss 2.0294): 25%|██▌ | 16/63 [00:15<00:23, 2.03it/s] Training 1/1 epoch (loss 2.0294): 27%|██▋ | 17/63 [00:15<00:22, 2.05it/s] Training 1/1 epoch (loss 1.9367): 27%|██▋ | 17/63 [00:15<00:22, 2.05it/s] Training 1/1 epoch (loss 1.9367): 29%|██▊ | 18/63 [00:15<00:20, 2.16it/s] Training 1/1 epoch (loss 1.9497): 29%|██▊ | 18/63 [00:16<00:20, 2.16it/s] Training 1/1 epoch (loss 1.9497): 30%|███ | 19/63 [00:16<00:20, 2.15it/s] Training 1/1 epoch (loss 1.9627): 30%|███ | 19/63 [00:16<00:20, 2.15it/s] Training 1/1 epoch (loss 1.9627): 32%|███▏ | 20/63 [00:16<00:20, 2.11it/s] Training 1/1 epoch (loss 1.9898): 32%|███▏ | 20/63 [00:17<00:20, 2.11it/s] Training 1/1 epoch (loss 1.9898): 33%|███▎ | 21/63 [00:17<00:18, 2.21it/s] Training 1/1 epoch (loss 2.1427): 33%|███▎ | 21/63 [00:17<00:18, 2.21it/s] Training 1/1 epoch (loss 2.1427): 35%|███▍ | 22/63 [00:17<00:17, 2.38it/s] Training 1/1 epoch (loss 1.9799): 35%|███▍ | 22/63 [00:18<00:17, 2.38it/s] Training 1/1 epoch (loss 1.9799): 37%|███▋ | 23/63 [00:18<00:16, 2.37it/s] Training 1/1 epoch (loss 2.0389): 37%|███▋ | 23/63 [00:18<00:16, 2.37it/s] Training 1/1 epoch (loss 2.0389): 38%|███▊ | 24/63 [00:18<00:16, 2.41it/s] Training 1/1 epoch (loss 1.9805): 38%|███▊ | 24/63 [00:18<00:16, 2.41it/s] Training 1/1 epoch (loss 1.9805): 40%|███▉ | 25/63 [00:18<00:15, 2.47it/s] Training 1/1 epoch (loss 2.0551): 40%|███▉ | 25/63 [00:19<00:15, 2.47it/s] Training 1/1 epoch (loss 2.0551): 41%|████▏ | 26/63 [00:19<00:14, 2.52it/s] Training 1/1 epoch (loss 2.0303): 41%|████▏ | 26/63 [00:19<00:14, 2.52it/s] Training 1/1 epoch (loss 2.0303): 43%|████▎ | 27/63 [00:19<00:15, 2.38it/s] Training 1/1 epoch (loss 1.8327): 43%|████▎ | 27/63 [00:20<00:15, 2.38it/s] Training 1/1 epoch (loss 1.8327): 44%|████▍ | 28/63 [00:20<00:14, 2.40it/s] Training 1/1 epoch (loss 2.0123): 44%|████▍ | 28/63 [00:20<00:14, 2.40it/s] Training 1/1 epoch (loss 2.0123): 46%|████▌ | 29/63 [00:20<00:13, 2.50it/s] Training 1/1 epoch (loss 1.9087): 46%|████▌ | 29/63 [00:20<00:13, 2.50it/s] Training 1/1 epoch (loss 1.9087): 48%|████▊ | 30/63 [00:20<00:12, 2.57it/s] Training 1/1 epoch (loss 2.0823): 48%|████▊ | 30/63 [00:21<00:12, 2.57it/s] Training 1/1 epoch (loss 2.0823): 49%|████▉ | 31/63 [00:21<00:12, 2.53it/s] Training 1/1 epoch (loss 2.0103): 49%|████▉ | 31/63 [00:21<00:12, 2.53it/s] Training 1/1 epoch (loss 2.0103): 51%|█████ | 32/63 [00:21<00:12, 2.55it/s] Training 1/1 epoch (loss 2.0080): 51%|█████ | 32/63 [00:22<00:12, 2.55it/s] Training 1/1 epoch (loss 2.0080): 52%|█████▏ | 33/63 [00:22<00:11, 2.53it/s] Training 1/1 epoch (loss 1.8523): 52%|█████▏ | 33/63 [00:22<00:11, 2.53it/s] Training 1/1 epoch (loss 1.8523): 54%|█████▍ | 34/63 [00:22<00:11, 2.59it/s] Training 1/1 epoch (loss 2.0902): 54%|█████▍ | 34/63 [00:22<00:11, 2.59it/s] Training 1/1 epoch (loss 2.0902): 56%|█████▌ | 35/63 [00:22<00:11, 2.42it/s] Training 1/1 epoch (loss 1.9712): 56%|█████▌ | 35/63 [00:23<00:11, 2.42it/s] Training 1/1 epoch (loss 1.9712): 57%|█████▋ | 36/63 [00:23<00:11, 2.38it/s] Training 1/1 epoch (loss 2.1606): 57%|█████▋ | 36/63 [00:23<00:11, 2.38it/s] Training 1/1 epoch (loss 2.1606): 59%|█████▊ | 37/63 [00:23<00:10, 2.39it/s] Training 1/1 epoch (loss 1.8062): 59%|█████▊ | 37/63 [00:24<00:10, 2.39it/s] Training 1/1 epoch (loss 1.8062): 60%|██████ | 38/63 [00:24<00:11, 2.26it/s] Training 1/1 epoch (loss 1.9797): 60%|██████ | 38/63 [00:24<00:11, 2.26it/s] Training 1/1 epoch (loss 1.9797): 62%|██████▏ | 39/63 [00:24<00:11, 2.15it/s] Training 1/1 epoch (loss 1.9174): 62%|██████▏ | 39/63 [00:25<00:11, 2.15it/s] Training 1/1 epoch (loss 1.9174): 63%|██████▎ | 40/63 [00:25<00:10, 2.12it/s] Training 1/1 epoch (loss 1.8677): 63%|██████▎ | 40/63 [00:25<00:10, 2.12it/s] Training 1/1 epoch (loss 1.8677): 65%|██████▌ | 41/63 [00:25<00:09, 2.21it/s] Training 1/1 epoch (loss 1.9840): 65%|██████▌ | 41/63 [00:25<00:09, 2.21it/s] Training 1/1 epoch (loss 1.9840): 67%|██████▋ | 42/63 [00:25<00:08, 2.37it/s] Training 1/1 epoch (loss 2.0601): 67%|██████▋ | 42/63 [00:26<00:08, 2.37it/s] Training 1/1 epoch (loss 2.0601): 68%|██████▊ | 43/63 [00:26<00:08, 2.45it/s] Training 1/1 epoch (loss 1.9732): 68%|██████▊ | 43/63 [00:26<00:08, 2.45it/s] Training 1/1 epoch (loss 1.9732): 70%|██████▉ | 44/63 [00:26<00:07, 2.38it/s] Training 1/1 epoch (loss 1.9440): 70%|██████▉ | 44/63 [00:27<00:07, 2.38it/s] Training 1/1 epoch (loss 1.9440): 71%|███████▏ | 45/63 [00:27<00:08, 2.25it/s] Training 1/1 epoch (loss 1.9001): 71%|███████▏ | 45/63 [00:27<00:08, 2.25it/s] Training 1/1 epoch (loss 1.9001): 73%|███████▎ | 46/63 [00:27<00:07, 2.41it/s] Training 1/1 epoch (loss 1.8930): 73%|███████▎ | 46/63 [00:27<00:07, 2.41it/s] Training 1/1 epoch (loss 1.8930): 75%|███████▍ | 47/63 [00:27<00:06, 2.51it/s] Training 1/1 epoch (loss 2.0864): 75%|███████▍ | 47/63 [00:28<00:06, 2.51it/s] Training 1/1 epoch (loss 2.0864): 76%|███████▌ | 48/63 [00:28<00:05, 2.55it/s] Training 1/1 epoch (loss 1.8801): 76%|███████▌ | 48/63 [00:28<00:05, 2.55it/s] Training 1/1 epoch (loss 1.8801): 78%|███████▊ | 49/63 [00:28<00:05, 2.55it/s] Training 1/1 epoch (loss 1.9363): 78%|███████▊ | 49/63 [00:29<00:05, 2.55it/s] Training 1/1 epoch (loss 1.9363): 79%|███████▉ | 50/63 [00:29<00:05, 2.37it/s] Training 1/1 epoch (loss 1.9833): 79%|███████▉ | 50/63 [00:29<00:05, 2.37it/s] Training 1/1 epoch (loss 1.9833): 81%|████████ | 51/63 [00:29<00:04, 2.49it/s] Training 1/1 epoch (loss 1.9821): 81%|████████ | 51/63 [00:29<00:04, 2.49it/s] Training 1/1 epoch (loss 1.9821): 83%|████████▎ | 52/63 [00:29<00:04, 2.59it/s] Training 1/1 epoch (loss 1.8717): 83%|████████▎ | 52/63 [00:30<00:04, 2.59it/s] Training 1/1 epoch (loss 1.8717): 84%|████████▍ | 53/63 [00:30<00:03, 2.68it/s] Training 1/1 epoch (loss 1.9399): 84%|████████▍ | 53/63 [00:30<00:03, 2.68it/s] Training 1/1 epoch (loss 1.9399): 86%|████████▌ | 54/63 [00:30<00:03, 2.60it/s] Training 1/1 epoch (loss 2.0007): 86%|████████▌ | 54/63 [00:31<00:03, 2.60it/s] Training 1/1 epoch (loss 2.0007): 87%|████████▋ | 55/63 [00:31<00:03, 2.60it/s] Training 1/1 epoch (loss 1.7834): 87%|████████▋ | 55/63 [00:31<00:03, 2.60it/s] Training 1/1 epoch (loss 1.7834): 89%|████████▉ | 56/63 [00:31<00:02, 2.48it/s] Training 1/1 epoch (loss 1.9033): 89%|████████▉ | 56/63 [00:31<00:02, 2.48it/s] Training 1/1 epoch (loss 1.9033): 90%|█████████ | 57/63 [00:31<00:02, 2.60it/s] Training 1/1 epoch (loss 1.9213): 90%|█████████ | 57/63 [00:32<00:02, 2.60it/s] Training 1/1 epoch (loss 1.9213): 92%|█████████▏| 58/63 [00:32<00:01, 2.66it/s] Training 1/1 epoch (loss 1.9180): 92%|█████████▏| 58/63 [00:32<00:01, 2.66it/s] Training 1/1 epoch (loss 1.9180): 94%|█████████▎| 59/63 [00:32<00:01, 2.66it/s] Training 1/1 epoch (loss 1.9246): 94%|█████████▎| 59/63 [00:33<00:01, 2.66it/s] Training 1/1 epoch (loss 1.9246): 95%|█████████▌| 60/63 [00:33<00:01, 2.53it/s] Training 1/1 epoch (loss 1.9254): 95%|█████████▌| 60/63 [00:33<00:01, 2.53it/s] Training 1/1 epoch (loss 1.9254): 97%|█████████▋| 61/63 [00:33<00:00, 2.53it/s] Training 1/1 epoch (loss 2.0030): 97%|█████████▋| 61/63 [00:33<00:00, 2.53it/s] Training 1/1 epoch (loss 2.0030): 98%|█████████▊| 62/63 [00:33<00:00, 2.53it/s] Training 1/1 epoch (loss 1.9616): 98%|█████████▊| 62/63 [00:34<00:00, 2.53it/s] Training 1/1 epoch (loss 1.9616): 100%|██████████| 63/63 [00:34<00:00, 2.62it/s] Training 1/1 epoch (loss 1.9616): 100%|██████████| 63/63 [00:34<00:00, 1.84it/s]
tokenizer config file saved in /aifs4su/hansirui_1st/boyuan/resist/setting3-safety/Qwen1.5-0.5B/Qwen1.5-0.5B-s3-Q1-2k/tokenizer_config.json
Special tokens file saved in /aifs4su/hansirui_1st/boyuan/resist/setting3-safety/Qwen1.5-0.5B/Qwen1.5-0.5B-s3-Q1-2k/special_tokens_map.json
wandb:
wandb:
wandb: Run history:
wandb: train/epoch ▁▁▁▁▂▂▂▂▂▃▃▃▃▃▃▄▄▄▄▄▄▅▅▅▅▆▆▆▆▆▆▇▇▇▇▇▇▇██
wandb: train/loss ▇█▇█▆▆▇▄▇▄▄▇▄▅▄▅▆▅▅▂▇▁▄▃▂▄▄▃▆▃▄▂▄▅▁▃▃▃▃▄
wandb: train/lr ▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁
wandb: train/step ▁▁▁▂▂▂▂▂▂▂▃▃▃▃▃▃▄▄▄▄▄▄▅▅▅▆▆▆▆▆▇▇▇▇▇▇████
wandb:
wandb: Run summary:
wandb: train/epoch 1
wandb: train/loss 1.96162
wandb: train/lr 1e-05
wandb: train/step 63
wandb:
wandb: 🚀 View run qwen-0.5b-s3-Q1-2k at: https://wandb.ai/xtom/Inverse_Alignment/runs/1slp0ya4
wandb: ⭐️ View project at: https://wandb.ai/xtom/Inverse_Alignment
wandb: Synced 5 W&B file(s), 0 media file(s), 0 artifact file(s) and 0 other file(s)
wandb: Find logs at: /aifs4su/hansirui_1st/boyuan/resist/setting3-safety/Qwen1.5-0.5B/Qwen1.5-0.5B-s3-Q1-2k/wandb/run-20250528_184512-1slp0ya4/logs