Upload folder using huggingface_hub

gpt2/gpt2-medium/dskd_v2_eta/forward_kl-bf16__teacher_qwen1.5__kd^rate0.5__kd^temp2.0__epoch20__bsz8x1x1x1_8__lr0.0005__proj^lr0.001/args.json

@@ -0,0 +1 @@
+{"model_path": "./model_hub/gpt2/gpt2-medium", "ckpt_name": null, "model_type": "gpt2", "teacher_model_type": "qwen", "n_gpu": 1, "n_nodes": 1, "teacher_model_path": "VoCuc/Qwen1.5_1.8B_SFT_Dolly", "teacher_model_fp16": true, "model_parallel": false, "model_parallel_size": null, "no_value": false, "dropout_path_rate": null, "fp32": false, "model_dtype": "bf16", "task": "dskd_v2_eta", "do_train": true, "do_valid": true, "do_eval": false, "base_path": ".", "load": null, "save_dir": "./outputs/gpt2/gpt2-medium/dskd_v2_eta/forward_kl-bf16__teacher_qwen1.5__kd^rate0.5__kd^temp2.0__epoch20__bsz8x1x1x1_8__lr0.0005__proj^lr0.001", "log_interval": 50, "save_interval": 1, "eval_interval": 1, "local_rank": 0, "save_additional_suffix": "", "save_rollout": false, "eb_sample_times": 3, "keep_best_n_checkpoints": 5, "criterion": "dual_space_kd_v2_with_eta", "eval_tqdm": false, "report_logits": false, "only_save_projector": false, "debug": false, "on_policy": false, "on_policy_after_n_epochs": 3, "stu_gen_ratio": 0.5, "data_dir": "./data/dolly/", "processed_data_dir": null, "force_process": false, "force_process_demo": false, "data_process_workers": -1, "train_num": -1, "train_ratio": 1, "dev_num": 1000, "dev_ratio": 1, "gen_num": -1, "data_names": null, "prompt_type": null, "num_workers": 0, "max_prompt_length": 256, "min_prompt_length": 128, "json_data": false, "bin_data": false, "txt_data": false, "prompt_data_dir": null, "pretrain_data_dir": null, "eval_ppl": false, "eval_rw": false, "eval_gen": true, "only_prompt": false, "batch_size": 8, "eval_batch_size": 32, "clip_grad": 1.0, "total_iters": null, "train_iters_per_epoch": -1, "max_length": 512, "seed": 10, "seed_order": 42, "seed_data": 42, "seed_ppo": 42, "seed_lm": 7, "num_epochs": 20, "training_epochs": 10000, "gradient_accumulation_steps": 1, "gradient_checkpointing": false, "attn_dtype": null, "lr": 0.0005, "lr_min": 1e-07, "weight_decay": 0.01, "loss_scale": 65536, "kd_rate": 0.5, "kd_temperature": 2.0, "kd_objective": "forward_kl", "teacher_temperature": 1.0, "label_smoothing": 0.0, "adaptive_kl_alpha": 0.5, "skew_lambda": 0.1, "warmup_iters": 0, "lr_decay_iters": null, "lr_decay_style": "cosine", "scheduler_name": "constant_trm", "top_k": 0, "top_p": 1.0, "do_sample": true, "no_repeat_ngram_size": 6, "repetition_penalty": null, "num_beams": 1, "temperature": 1.0, "eval_gen_repeat_times": 3, "peft": null, "peft_lora_r": 16, "peft_lora_alpha": 64, "peft_lora_dropout": 0.1, "peft_name": null, "peft_path": null, "teacher_peft_name": null, "teacher_peft_path": null, "deepspeed": true, "deepspeed_config": "./configs/deepspeed/ds_config_bf16.json", "deepscale": false, "deepscale_config": null, "projector_lr": 0.001, "vocab_alignment_path": null, "teacher_to_student_token_mapping": null, "teacher_to_student_id_mapping": null, "student_to_teacher_token_mapping": null, "student_to_teacher_id_mapping": null, "init_t2s_projector": true, "init_s2t_projector": true, "topk_vocab": -1, "only_stu_kd": false, "only_tea_kd": false, "t2s_agreement": 1.0, "student_layer_mapping": [-1], "teacher_layer_mapping": [-1], "split_layer_mapping": [0, 0, 0, 0], "w_span_loss": 1.0, "MTA_mode": false, "entropy_weight": false, "wo_span_weight": false, "rank": 0, "world_size": 1}

gpt2/gpt2-medium/dskd_v2_eta/forward_kl-bf16__teacher_qwen1.5__kd^rate0.5__kd^temp2.0__epoch20__bsz8x1x1x1_8__lr0.0005__proj^lr0.001/epoch1_step1430_loss3.5965_rougel21.3929/config.json

@@ -0,0 +1,44 @@
+{
+  "activation_function": "gelu_new",
+  "add_cross_attention": false,
+  "architectures": [
+    "GPT2LMHeadModel"
+  ],
+  "attn_pdrop": 0.1,
+  "bos_token_id": 50256,
+  "dtype": "bfloat16",
+  "embd_pdrop": 0.1,
+  "eos_token_id": 50256,
+  "initializer_range": 0.02,
+  "is_model_parallel": false,
+  "layer_norm_epsilon": 1e-05,
+  "model_type": "gpt2",
+  "n_ctx": 1024,
+  "n_embd": 1024,
+  "n_head": 16,
+  "n_inner": null,
+  "n_layer": 24,
+  "n_positions": 1024,
+  "n_special": 0,
+  "pad_token_id": null,
+  "predict_special_tokens": true,
+  "reorder_and_upcast_attn": false,
+  "resid_pdrop": 0.1,
+  "scale_attn_by_inverse_layer_idx": false,
+  "scale_attn_weights": true,
+  "summary_activation": null,
+  "summary_first_dropout": 0.1,
+  "summary_proj_to_labels": true,
+  "summary_type": "cls_index",
+  "summary_use_proj": true,
+  "task_specific_params": {
+    "text-generation": {
+      "do_sample": true,
+      "max_length": 50
+    }
+  },
+  "tie_word_embeddings": true,
+  "transformers_version": "5.9.0",
+  "use_cache": true,
+  "vocab_size": 50257
+}

gpt2/gpt2-medium/dskd_v2_eta/forward_kl-bf16__teacher_qwen1.5__kd^rate0.5__kd^temp2.0__epoch20__bsz8x1x1x1_8__lr0.0005__proj^lr0.001/epoch1_step1430_loss3.5965_rougel21.3929/generation_config.json

@@ -0,0 +1,6 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 50256,
+  "eos_token_id": 50256,
+  "transformers_version": "5.9.0"
+}

gpt2/gpt2-medium/dskd_v2_eta/forward_kl-bf16__teacher_qwen1.5__kd^rate0.5__kd^temp2.0__epoch20__bsz8x1x1x1_8__lr0.0005__proj^lr0.001/epoch1_step1430_loss3.5965_rougel21.3929/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:53ea4fe0a8ea9be971dd064f3bc4b2a6fc7af4c0eb9abf3856746357fef6850b
+size 709676664

gpt2/gpt2-medium/dskd_v2_eta/forward_kl-bf16__teacher_qwen1.5__kd^rate0.5__kd^temp2.0__epoch20__bsz8x1x1x1_8__lr0.0005__proj^lr0.001/epoch1_step1430_loss3.5965_rougel21.3929/tokenizer_config.json

@@ -0,0 +1,13 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": "<|endoftext|>",
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "is_local": true,
+  "local_files_only": false,
+  "model_max_length": 1024,
+  "pad_token": "<|endoftext|>",
+  "tokenizer_class": "GPT2Tokenizer",
+  "unk_token": "<|endoftext|>"
+}

gpt2/gpt2-medium/dskd_v2_eta/forward_kl-bf16__teacher_qwen1.5__kd^rate0.5__kd^temp2.0__epoch20__bsz8x1x1x1_8__lr0.0005__proj^lr0.001/epoch2_step2860_loss3.9196_rougel22.8349/config.json

@@ -0,0 +1,44 @@
+{
+  "activation_function": "gelu_new",
+  "add_cross_attention": false,
+  "architectures": [
+    "GPT2LMHeadModel"
+  ],
+  "attn_pdrop": 0.1,
+  "bos_token_id": 50256,
+  "dtype": "bfloat16",
+  "embd_pdrop": 0.1,
+  "eos_token_id": 50256,
+  "initializer_range": 0.02,
+  "is_model_parallel": false,
+  "layer_norm_epsilon": 1e-05,
+  "model_type": "gpt2",
+  "n_ctx": 1024,
+  "n_embd": 1024,
+  "n_head": 16,
+  "n_inner": null,
+  "n_layer": 24,
+  "n_positions": 1024,
+  "n_special": 0,
+  "pad_token_id": null,
+  "predict_special_tokens": true,
+  "reorder_and_upcast_attn": false,
+  "resid_pdrop": 0.1,
+  "scale_attn_by_inverse_layer_idx": false,
+  "scale_attn_weights": true,
+  "summary_activation": null,
+  "summary_first_dropout": 0.1,
+  "summary_proj_to_labels": true,
+  "summary_type": "cls_index",
+  "summary_use_proj": true,
+  "task_specific_params": {
+    "text-generation": {
+      "do_sample": true,
+      "max_length": 50
+    }
+  },
+  "tie_word_embeddings": true,
+  "transformers_version": "5.9.0",
+  "use_cache": true,
+  "vocab_size": 50257
+}

gpt2/gpt2-medium/dskd_v2_eta/forward_kl-bf16__teacher_qwen1.5__kd^rate0.5__kd^temp2.0__epoch20__bsz8x1x1x1_8__lr0.0005__proj^lr0.001/epoch2_step2860_loss3.9196_rougel22.8349/generation_config.json

@@ -0,0 +1,6 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 50256,
+  "eos_token_id": 50256,
+  "transformers_version": "5.9.0"
+}

gpt2/gpt2-medium/dskd_v2_eta/forward_kl-bf16__teacher_qwen1.5__kd^rate0.5__kd^temp2.0__epoch20__bsz8x1x1x1_8__lr0.0005__proj^lr0.001/epoch2_step2860_loss3.9196_rougel22.8349/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a8d414e1f1c576184fcc01e7569e42ed4b85588c7366ff28eb4c5e62192b97c7
+size 709676664

gpt2/gpt2-medium/dskd_v2_eta/forward_kl-bf16__teacher_qwen1.5__kd^rate0.5__kd^temp2.0__epoch20__bsz8x1x1x1_8__lr0.0005__proj^lr0.001/epoch2_step2860_loss3.9196_rougel22.8349/tokenizer_config.json

@@ -0,0 +1,13 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": "<|endoftext|>",
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "is_local": true,
+  "local_files_only": false,
+  "model_max_length": 1024,
+  "pad_token": "<|endoftext|>",
+  "tokenizer_class": "GPT2Tokenizer",
+  "unk_token": "<|endoftext|>"
+}

gpt2/gpt2-medium/dskd_v2_eta/forward_kl-bf16__teacher_qwen1.5__kd^rate0.5__kd^temp2.0__epoch20__bsz8x1x1x1_8__lr0.0005__proj^lr0.001/epoch3_step4290_loss4.6196_rougel23.2497/config.json

@@ -0,0 +1,44 @@
+{
+  "activation_function": "gelu_new",
+  "add_cross_attention": false,
+  "architectures": [
+    "GPT2LMHeadModel"
+  ],
+  "attn_pdrop": 0.1,
+  "bos_token_id": 50256,
+  "dtype": "bfloat16",
+  "embd_pdrop": 0.1,
+  "eos_token_id": 50256,
+  "initializer_range": 0.02,
+  "is_model_parallel": false,
+  "layer_norm_epsilon": 1e-05,
+  "model_type": "gpt2",
+  "n_ctx": 1024,
+  "n_embd": 1024,
+  "n_head": 16,
+  "n_inner": null,
+  "n_layer": 24,
+  "n_positions": 1024,
+  "n_special": 0,
+  "pad_token_id": null,
+  "predict_special_tokens": true,
+  "reorder_and_upcast_attn": false,
+  "resid_pdrop": 0.1,
+  "scale_attn_by_inverse_layer_idx": false,
+  "scale_attn_weights": true,
+  "summary_activation": null,
+  "summary_first_dropout": 0.1,
+  "summary_proj_to_labels": true,
+  "summary_type": "cls_index",
+  "summary_use_proj": true,
+  "task_specific_params": {
+    "text-generation": {
+      "do_sample": true,
+      "max_length": 50
+    }
+  },
+  "tie_word_embeddings": true,
+  "transformers_version": "5.9.0",
+  "use_cache": true,
+  "vocab_size": 50257
+}

gpt2/gpt2-medium/dskd_v2_eta/forward_kl-bf16__teacher_qwen1.5__kd^rate0.5__kd^temp2.0__epoch20__bsz8x1x1x1_8__lr0.0005__proj^lr0.001/epoch3_step4290_loss4.6196_rougel23.2497/generation_config.json

@@ -0,0 +1,6 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 50256,
+  "eos_token_id": 50256,
+  "transformers_version": "5.9.0"
+}

gpt2/gpt2-medium/dskd_v2_eta/forward_kl-bf16__teacher_qwen1.5__kd^rate0.5__kd^temp2.0__epoch20__bsz8x1x1x1_8__lr0.0005__proj^lr0.001/epoch3_step4290_loss4.6196_rougel23.2497/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1c30b6c4e81b0a6b91f20ce98d295e42e0d23df43e85c11384f942732824e857
+size 709676664

gpt2/gpt2-medium/dskd_v2_eta/forward_kl-bf16__teacher_qwen1.5__kd^rate0.5__kd^temp2.0__epoch20__bsz8x1x1x1_8__lr0.0005__proj^lr0.001/epoch3_step4290_loss4.6196_rougel23.2497/tokenizer_config.json

@@ -0,0 +1,13 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": "<|endoftext|>",
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "is_local": true,
+  "local_files_only": false,
+  "model_max_length": 1024,
+  "pad_token": "<|endoftext|>",
+  "tokenizer_class": "GPT2Tokenizer",
+  "unk_token": "<|endoftext|>"
+}

gpt2/gpt2-medium/dskd_v2_eta/forward_kl-bf16__teacher_qwen1.5__kd^rate0.5__kd^temp2.0__epoch20__bsz8x1x1x1_8__lr0.0005__proj^lr0.001/train.log

@@ -0,0 +1,227 @@
+[2026-05-28 10:33:17] [INFO]  Using default tokenizer.
+[2026-05-28 10:33:17] [INFO]  Using world size: 1
+
+
+============================== EXP at 2026-05-28 10:33:17 ==============================
+[2026-05-28 10:33:17] [INFO]  Namespace(model_path='./model_hub/gpt2/gpt2-medium', ckpt_name=None, model_type='gpt2', teacher_model_type='qwen', n_gpu=1, n_nodes=1, teacher_model_path='VoCuc/Qwen1.5_1.8B_SFT_Dolly', teacher_model_fp16=True, model_parallel=False, model_parallel_size=None, no_value=False, dropout_path_rate=None, fp32=False, model_dtype='bf16', task='dskd_v2_eta', do_train=True, do_valid=True, do_eval=False, base_path='.', load=None, save_dir='./outputs/gpt2/gpt2-medium/dskd_v2_eta/forward_kl-bf16__teacher_qwen1.5__kd^rate0.5__kd^temp2.0__epoch20__bsz8x1x1x1_8__lr0.0005__proj^lr0.001', log_interval=50, save_interval=1, eval_interval=1, local_rank=0, save_additional_suffix='', save_rollout=False, eb_sample_times=3, keep_best_n_checkpoints=5, criterion='dual_space_kd_v2_with_eta', eval_tqdm=False, report_logits=False, only_save_projector=False, debug=False, on_policy=False, on_policy_after_n_epochs=3, stu_gen_ratio=0.5, data_dir='./data/dolly/', processed_data_dir=None, force_process=False, force_process_demo=False, data_process_workers=-1, train_num=-1, train_ratio=1, dev_num=1000, dev_ratio=1, gen_num=-1, data_names=None, prompt_type=None, num_workers=0, max_prompt_length=256, min_prompt_length=128, json_data=False, bin_data=False, txt_data=False, prompt_data_dir=None, pretrain_data_dir=None, eval_ppl=False, eval_rw=False, eval_gen=True, only_prompt=False, batch_size=8, eval_batch_size=32, clip_grad=1.0, total_iters=None, train_iters_per_epoch=-1, max_length=512, seed=10, seed_order=42, seed_data=42, seed_ppo=42, seed_lm=7, num_epochs=20, training_epochs=10000, gradient_accumulation_steps=1, gradient_checkpointing=False, attn_dtype=None, lr=0.0005, lr_min=1e-07, weight_decay=0.01, loss_scale=65536, kd_rate=0.5, kd_temperature=2.0, kd_objective='forward_kl', teacher_temperature=1.0, label_smoothing=0.0, adaptive_kl_alpha=0.5, skew_lambda=0.1, warmup_iters=0, lr_decay_iters=None, lr_decay_style='cosine', scheduler_name='constant_trm', top_k=0, top_p=1.0, do_sample=True, no_repeat_ngram_size=6, repetition_penalty=None, num_beams=1, temperature=1.0, eval_gen_repeat_times=3, peft=None, peft_lora_r=16, peft_lora_alpha=64, peft_lora_dropout=0.1, peft_name=None, peft_path=None, teacher_peft_name=None, teacher_peft_path=None, deepspeed=True, deepspeed_config='./configs/deepspeed/ds_config_bf16.json', deepscale=False, deepscale_config=None, projector_lr=0.001, vocab_alignment_path=None, teacher_to_student_token_mapping=None, teacher_to_student_id_mapping=None, student_to_teacher_token_mapping=None, student_to_teacher_id_mapping=None, init_t2s_projector=True, init_s2t_projector=True, topk_vocab=-1, only_stu_kd=False, only_tea_kd=False, t2s_agreement=1.0, student_layer_mapping=[-1], teacher_layer_mapping=[-1], split_layer_mapping=[0, 0, 0, 0], w_span_loss=1.0, MTA_mode=False, entropy_weight=False, wo_span_weight=False, rank=0, world_size=1)
+[2026-05-28 10:33:17] [INFO]  Initializing a distiller for knowledge distillation...
+[2026-05-28 10:33:18] [INFO]  Loading student model with bf16...
+
+[2026-05-28 10:33:24] [INFO]  trainable params: 354,823,168 || all params: 354,823,168 || trainable%: 100.0000
+[2026-05-28 10:33:24] [INFO]  HTTP Request: HEAD https://huggingface.co/VoCuc/Qwen1.5_1.8B_SFT_Dolly/resolve/main/config.json "HTTP/1.1 307 Temporary Redirect"
+[2026-05-28 10:33:24] [INFO]  HTTP Request: HEAD https://huggingface.co/api/resolve-cache/models/VoCuc/Qwen1.5_1.8B_SFT_Dolly/d87ed7636245a30e577150f1c841051fde53ed63/config.json "HTTP/1.1 200 OK"
+[2026-05-28 10:33:24] [INFO]  HTTP Request: HEAD https://huggingface.co/VoCuc/Qwen1.5_1.8B_SFT_Dolly/resolve/main/config.json "HTTP/1.1 307 Temporary Redirect"
+[2026-05-28 10:33:24] [INFO]  HTTP Request: HEAD https://huggingface.co/api/resolve-cache/models/VoCuc/Qwen1.5_1.8B_SFT_Dolly/d87ed7636245a30e577150f1c841051fde53ed63/config.json "HTTP/1.1 200 OK"
+[2026-05-28 10:33:24] [INFO]  HTTP Request: HEAD https://huggingface.co/VoCuc/Qwen1.5_1.8B_SFT_Dolly/resolve/main/tokenizer_config.json "HTTP/1.1 307 Temporary Redirect"
+[2026-05-28 10:33:24] [INFO]  HTTP Request: HEAD https://huggingface.co/api/resolve-cache/models/VoCuc/Qwen1.5_1.8B_SFT_Dolly/d87ed7636245a30e577150f1c841051fde53ed63/tokenizer_config.json "HTTP/1.1 200 OK"
+[2026-05-28 10:33:25] [INFO]  HTTP Request: GET https://huggingface.co/api/models/VoCuc/Qwen1.5_1.8B_SFT_Dolly/tree/main/additional_chat_templates?recursive=false&expand=false "HTTP/1.1 404 Not Found"
+[2026-05-28 10:33:25] [INFO]  HTTP Request: GET https://huggingface.co/api/models/VoCuc/Qwen1.5_1.8B_SFT_Dolly/tree/main?recursive=true&expand=false "HTTP/1.1 200 OK"
+[2026-05-28 10:33:25] [INFO]  HTTP Request: GET https://huggingface.co/api/models/VoCuc/Qwen1.5_1.8B_SFT_Dolly "HTTP/1.1 200 OK"
+[2026-05-28 10:33:25] [INFO]  Loading teacher model with bf16...
+[2026-05-28 10:33:26] [INFO]  HTTP Request: HEAD https://huggingface.co/VoCuc/Qwen1.5_1.8B_SFT_Dolly/resolve/main/config.json "HTTP/1.1 307 Temporary Redirect"
+[2026-05-28 10:33:26] [INFO]  HTTP Request: HEAD https://huggingface.co/api/resolve-cache/models/VoCuc/Qwen1.5_1.8B_SFT_Dolly/d87ed7636245a30e577150f1c841051fde53ed63/config.json "HTTP/1.1 200 OK"
+[2026-05-28 10:33:26] [INFO]  HTTP Request: HEAD https://huggingface.co/VoCuc/Qwen1.5_1.8B_SFT_Dolly/resolve/main/model.safetensors "HTTP/1.1 404 Not Found"
+
+[2026-05-28 10:33:26] [INFO]  HTTP Request: GET https://huggingface.co/api/models/VoCuc/Qwen1.5_1.8B_SFT_Dolly/commits/main "HTTP/1.1 200 OK"
+
+[2026-05-28 10:33:26] [INFO]  HTTP Request: GET https://huggingface.co/api/models/VoCuc/Qwen1.5_1.8B_SFT_Dolly/discussions?p=0 "HTTP/1.1 200 OK"
+[2026-05-28 10:33:26] [INFO]  HTTP Request: HEAD https://huggingface.co/VoCuc/Qwen1.5_1.8B_SFT_Dolly/resolve/main/generation_config.json "HTTP/1.1 307 Temporary Redirect"
+[2026-05-28 10:33:26] [INFO]  HTTP Request: HEAD https://huggingface.co/api/resolve-cache/models/VoCuc/Qwen1.5_1.8B_SFT_Dolly/d87ed7636245a30e577150f1c841051fde53ed63/generation_config.json "HTTP/1.1 200 OK"
+[2026-05-28 10:33:26] [INFO]   > number of parameters of the teacher model: 1,836,828,672
+[2026-05-28 10:33:26] [INFO]  student tokenizer: GPT2Tokenizer(name_or_path='./model_hub/gpt2/gpt2-medium', vocab_size=50257, model_max_length=1024, padding_side='right', truncation_side='right', special_tokens={'bos_token': '<|endoftext|>', 'eos_token': '<|endoftext|>', 'unk_token': '<|endoftext|>', 'pad_token': '<|endoftext|>'}, added_tokens_decoder={
+	50256: AddedToken("<|endoftext|>", rstrip=False, lstrip=False, single_word=False, normalized=True, special=True),
+})
+[2026-05-28 10:33:26] [INFO]  teacher tokenizer: Qwen2Tokenizer(name_or_path='VoCuc/Qwen1.5_1.8B_SFT_Dolly', vocab_size=151643, model_max_length=32768, padding_side='right', truncation_side='right', special_tokens={'eos_token': '<|endoftext|>', 'pad_token': '<|endoftext|>'}, added_tokens_decoder={
+	151643: AddedToken("<|endoftext|>", rstrip=False, lstrip=False, single_word=False, normalized=False, special=True),
+	151644: AddedToken("<|im_start|>", rstrip=False, lstrip=False, single_word=False, normalized=False, special=True),
+	151645: AddedToken("<|im_end|>", rstrip=False, lstrip=False, single_word=False, normalized=False, special=True),
+})
+[2026-05-28 10:33:27] [INFO]  HTTP Request: GET https://huggingface.co/api/models/VoCuc/Qwen1.5_1.8B_SFT_Dolly/commits/refs%2Fpr%2F2 "HTTP/1.1 200 OK"
+[2026-05-28 10:33:27] [INFO]  Found overlap tokens of two tokenizers: 42257
+[2026-05-28 10:33:27] [INFO]  Init t2s projector through pseudo inverse
+[2026-05-28 10:33:27] [INFO]  HTTP Request: HEAD https://huggingface.co/VoCuc/Qwen1.5_1.8B_SFT_Dolly/resolve/refs%2Fpr%2F2/model.safetensors.index.json "HTTP/1.1 404 Not Found"
+[2026-05-28 10:33:27] [INFO]  Projector Initialization Finished
+[2026-05-28 10:33:27] [INFO]  HTTP Request: HEAD https://huggingface.co/VoCuc/Qwen1.5_1.8B_SFT_Dolly/resolve/refs%2Fpr%2F2/model.safetensors "HTTP/1.1 302 Found"
+[2026-05-28 10:33:27] [INFO]  t2s projector structure: Linear(in_features=2048, out_features=1024, bias=False)
+[2026-05-28 10:33:27] [INFO]  s2t projector structure: Linear(in_features=1024, out_features=2048, bias=False)
+[2026-05-28 10:33:27] [INFO]  Processing dataset for student model (and teacher model)...
+
  0%|          | 0/11435 [00:00<?, ?it/s]
  1%|▏         | 144/11435 [00:00<00:07, 1434.93it/s][transformers] Token indices sequence length is longer than the specified maximum sequence length for this model (1092 > 1024). Running this sequence through the model will result in indexing errors
+
  3%|▎         | 300/11435 [00:00<00:07, 1507.21it/s]
  4%|▍         | 458/11435 [00:00<00:07, 1538.78it/s]
  5%|▌         | 624/11435 [00:00<00:06, 1584.65it/s]
  7%|▋         | 800/11435 [00:00<00:06, 1642.38it/s]
  8%|▊         | 965/11435 [00:00<00:06, 1643.53it/s]
 10%|▉         | 1130/11435 [00:00<00:06, 1622.84it/s]
 11%|█▏        | 1299/11435 [00:00<00:06, 1642.57it/s]
 13%|█▎        | 1464/11435 [00:00<00:06, 1637.34it/s]
 14%|█▍        | 1628/11435 [00:01<00:06, 1561.12it/s]
 16%|█▌        | 1786/11435 [00:01<00:06, 1566.21it/s]
 17%|█▋        | 1947/11435 [00:01<00:06, 1575.09it/s]
 18%|█▊        | 2110/11435 [00:01<00:05, 1589.74it/s]
 20%|█▉        | 2270/11435 [00:01<00:05, 1592.68it/s]
 21%|██▏       | 2434/11435 [00:01<00:05, 1603.87it/s]
 23%|██▎       | 2596/11435 [00:01<00:05, 1608.60it/s]
 24%|██▍       | 2757/11435 [00:01<00:05, 1544.75it/s]
 26%|██▌       | 2921/11435 [00:01<00:05, 1567.13it/s]
 27%|██▋       | 3079/11435 [00:01<00:05, 1566.32it/s]
 28%|██▊       | 3236/11435 [00:02<00:05, 1555.48it/s]
 30%|██▉       | 3396/11435 [00:02<00:05, 1568.19it/s]
 31%|███       | 3561/11435 [00:02<00:04, 1590.44it/s]
 33%|███▎      | 3721/11435 [00:02<00:04, 1558.61it/s]
 34%|███▍      | 3881/11435 [00:02<00:04, 1570.67it/s]
 35%|███▌      | 4039/11435 [00:02<00:05, 1478.42it/s]
 37%|███▋      | 4189/11435 [00:02<00:04, 1471.77it/s]
 38%|███▊      | 4347/11435 [00:02<00:04, 1497.25it/s]
 39%|███▉      | 4498/11435 [00:02<00:04, 1489.74it/s]
 41%|████      | 4648/11435 [00:02<00:04, 1485.93it/s]
 42%|████▏     | 4810/11435 [00:03<00:04, 1521.52it/s]
 43%|████▎     | 4963/11435 [00:03<00:04, 1421.63it/s]
 45%|████▍     | 5114/11435 [00:03<00:04, 1444.06it/s]
 46%|████▌     | 5265/11435 [00:03<00:04, 1462.57it/s]
 47%|████▋     | 5413/11435 [00:03<00:04, 1453.45it/s]
 49%|████▊     | 5569/11435 [00:03<00:03, 1483.87it/s]
 50%|█████     | 5718/11435 [00:03<00:04, 1391.27it/s]
 51%|█████▏    | 5875/11435 [00:03<00:03, 1438.49it/s]
 53%|█████▎    | 6025/11435 [00:03<00:03, 1454.60it/s]
 54%|█████▍    | 6179/11435 [00:04<00:03, 1470.49it/s]
 55%|█████▌    | 6327/11435 [00:04<00:03, 1470.74it/s]
 57%|█████▋    | 6480/11435 [00:04<00:03, 1486.65it/s]
 58%|█████▊    | 6633/11435 [00:04<00:03, 1498.80it/s]
 59%|█████▉    | 6784/11435 [00:04<00:03, 1477.05it/s]
 61%|██████    | 6937/11435 [00:04<00:03, 1490.79it/s]
 62%|██████▏   | 7095/11435 [00:04<00:02, 1511.82it/s]
 63%|██████▎   | 7247/11435 [00:04<00:02, 1502.35it/s]
 65%|██████▍   | 7398/11435 [00:04<00:02, 1483.97it/s]
 66%|██████▌   | 7551/11435 [00:04<00:02, 1496.81it/s]
 67%|██████▋   | 7711/11435 [00:05<00:02, 1523.32it/s]
 69%|██████▉   | 7864/11435 [00:05<00:02, 1497.13it/s]
 70%|███████   | 8014/11435 [00:05<00:02, 1471.17it/s]
 71%|███████▏  | 8162/11435 [00:05<00:02, 1462.79it/s]
 73%|███████▎  | 8309/11435 [00:05<00:02, 1442.11it/s]
 74%|███████▍  | 8462/11435 [00:05<00:02, 1466.90it/s]
 75%|███████▌  | 8611/11435 [00:05<00:01, 1471.66it/s]
 77%|███████▋  | 8775/11435 [00:05<00:01, 1518.64it/s]
 78%|███████▊  | 8927/11435 [00:05<00:01, 1469.50it/s]
 79%|███████▉  | 9078/11435 [00:05<00:01, 1478.81it/s]
 81%|████████  | 9229/11435 [00:06<00:01, 1487.65it/s]
 82%|████████▏ | 9378/11435 [00:06<00:01, 1463.35it/s]
 83%|████████▎ | 9525/11435 [00:06<00:01, 1455.23it/s]
 85%|████████▍ | 9677/11435 [00:06<00:01, 1473.14it/s]
 86%|████████▌ | 9825/11435 [00:06<00:01, 1464.93it/s]
 87%|████████▋ | 9993/11435 [00:06<00:00, 1527.67it/s]
 89%|████████▉ | 10159/11435 [00:06<00:00, 1566.19it/s]
 90%|█████████ | 10316/11435 [00:06<00:00, 1566.49it/s]
 92%|█████████▏| 10473/11435 [00:06<00:00, 1525.99it/s]
 93%|█████████▎| 10626/11435 [00:07<00:00, 1518.31it/s]
 94%|███��█████▍| 10779/11435 [00:07<00:00, 1449.46it/s]
 96%|█████████▌| 10937/11435 [00:07<00:00, 1484.37it/s]
 97%|█████████▋| 11096/11435 [00:07<00:00, 1514.07it/s]
 98%|█████████▊| 11253/11435 [00:07<00:00, 1529.54it/s]
+[2026-05-28 10:33:35] [INFO]  Num of train data: 11435
+[2026-05-28 10:33:35] [INFO]  Processing dataset for student model (and teacher model)...
+
  0%|          | 0/1000 [00:00<?, ?it/s]
 13%|█▎        | 127/1000 [00:00<00:00, 1269.95it/s]
 28%|██▊       | 277/1000 [00:00<00:00, 1400.10it/s]
 42%|████▏     | 420/1000 [00:00<00:00, 1413.16it/s]
 56%|█████▌    | 562/1000 [00:00<00:00, 1368.66it/s]
 71%|███████▏  | 714/1000 [00:00<00:00, 1421.83it/s]
 86%|████████▌ | 859/1000 [00:00<00:00, 1427.82it/s]
+[2026-05-28 10:33:36] [INFO]  Num of dev data: 1000
+[2026-05-28 10:33:36] [INFO]  Train iters per epoch = 1429
+[2026-05-28 10:33:36] [INFO]  Total_iters = 28580
+[2026-05-28 10:33:36] [INFO]  Optimizer = AdamW
+[2026-05-28 10:33:36] [INFO]  Start Fine-tuning
+[2026-05-28 10:33:36] [INFO]  Start iterations of epoch 1
+/home/ubuntu/mta_dskd_v2/env/lib/python3.13/site-packages/torch/autograd/graph.py:882: UserWarning: c10d::allreduce_: an autograd kernel was not registered to the Autograd key(s) but we are trying to backprop through it. This may lead to silently incorrect behavior. This behavior is deprecated and will be removed in a future version of PyTorch. If your operator is differentiable, please ensure you have registered an autograd kernel to the correct Autograd key (e.g. DispatchKey::Autograd, DispatchKey::CompositeImplicitAutograd). If your operator is not differentiable, or to squash this warning and use the previous behavior, please register torch::CppFunction::makeFallthrough() to DispatchKey::Autograd. (Triggered internally at /pytorch/torch/csrc/autograd/autograd_not_implemented_fallback.cpp:76.)
+  return Variable._execution_engine.run_backward(  # Calls into the C++ engine to run the backward pass
+[2026-05-28 10:34:00] [INFO]  train | epoch 001:      50 /  1429  global_step=50, loss=7.9767, nll_loss=4.0463, kd_loss=11.9071, accuracy=0.3663, micro_step_time=0.4679, step_time=0.4712, align_ratio=0.9013, t2s_agreement=0.5851, t2s_agreement_ratio=0.5321, t2s_acc=0.5322, t2s_acc_ratio=0.5322, t_preds_as_label_acc=0.9997, t2s_ce_loss=3.5865, t2s_kd_loss=2.4607, s2t_kd_loss=5.8599, lr=5.0000e-04, projector_lr=5.0000e-04, scale=1.0000
+[2026-05-28 10:34:20] [INFO]  train | epoch 001:     100 /  1429  global_step=100, loss=6.2158, nll_loss=3.5321, kd_loss=8.8994, accuracy=0.4337, micro_step_time=0.4107, step_time=0.4131, align_ratio=0.9007, t2s_agreement=0.6672, t2s_agreement_ratio=0.6093, t2s_acc=0.6093, t2s_acc_ratio=0.6093, t_preds_as_label_acc=0.9998, t2s_ce_loss=3.7271, t2s_kd_loss=2.2629, s2t_kd_loss=2.9095, lr=4.9998e-04, projector_lr=4.9998e-04, scale=1.0000
+[2026-05-28 10:34:41] [INFO]  train | epoch 001:     150 /  1429  global_step=150, loss=6.3890, nll_loss=3.6902, kd_loss=9.0878, accuracy=0.4183, micro_step_time=0.4193, step_time=0.4216, align_ratio=0.9053, t2s_agreement=0.6994, t2s_agreement_ratio=0.6379, t2s_acc=0.6380, t2s_acc_ratio=0.6380, t_preds_as_label_acc=0.9996, t2s_ce_loss=3.7770, t2s_kd_loss=2.4357, s2t_kd_loss=2.8751, lr=4.9997e-04, projector_lr=4.9997e-04, scale=1.0000
+[2026-05-28 10:35:03] [INFO]  train | epoch 001:     200 /  1429  global_step=200, loss=6.2654, nll_loss=3.5878, kd_loss=8.9429, accuracy=0.4303, micro_step_time=0.4358, step_time=0.4382, align_ratio=0.8957, t2s_agreement=0.7243, t2s_agreement_ratio=0.6541, t2s_acc=0.6544, t2s_acc_ratio=0.6544, t_preds_as_label_acc=0.9994, t2s_ce_loss=3.7484, t2s_kd_loss=2.4130, s2t_kd_loss=2.7815, lr=4.9994e-04, projector_lr=4.9994e-04, scale=1.0000
+[2026-05-28 10:35:25] [INFO]  train | epoch 001:     250 /  1429  global_step=250, loss=6.4831, nll_loss=3.6988, kd_loss=9.2675, accuracy=0.4177, micro_step_time=0.4389, step_time=0.4414, align_ratio=0.8962, t2s_agreement=0.7276, t2s_agreement_ratio=0.6558, t2s_acc=0.6557, t2s_acc_ratio=0.6557, t_preds_as_label_acc=0.9993, t2s_ce_loss=3.9858, t2s_kd_loss=2.4747, s2t_kd_loss=2.8070, lr=4.9991e-04, projector_lr=4.9991e-04, scale=1.0000
+[2026-05-28 10:35:48] [INFO]  train | epoch 001:     300 /  1429  global_step=300, loss=6.5716, nll_loss=3.7755, kd_loss=9.3677, accuracy=0.3995, micro_step_time=0.4544, step_time=0.4569, align_ratio=0.8976, t2s_agreement=0.7394, t2s_agreement_ratio=0.6686, t2s_acc=0.6687, t2s_acc_ratio=0.6687, t_preds_as_label_acc=0.9994, t2s_ce_loss=3.9656, t2s_kd_loss=2.5738, s2t_kd_loss=2.8284, lr=4.9986e-04, projector_lr=4.9986e-04, scale=1.0000
+[2026-05-28 10:36:10] [INFO]  train | epoch 001:     350 /  1429  global_step=350, loss=6.3975, nll_loss=3.7021, kd_loss=9.0929, accuracy=0.4149, micro_step_time=0.4334, step_time=0.4358, align_ratio=0.9060, t2s_agreement=0.7587, t2s_agreement_ratio=0.6927, t2s_acc=0.6925, t2s_acc_ratio=0.6925, t_preds_as_label_acc=0.9995, t2s_ce_loss=3.7648, t2s_kd_loss=2.5440, s2t_kd_loss=2.7841, lr=4.9982e-04, projector_lr=4.9982e-04, scale=1.0000
+[2026-05-28 10:36:31] [INFO]  train | epoch 001:     400 /  1429  global_step=400, loss=6.3284, nll_loss=3.6338, kd_loss=9.0229, accuracy=0.4276, micro_step_time=0.4110, step_time=0.4134, align_ratio=0.8974, t2s_agreement=0.7643, t2s_agreement_ratio=0.6925, t2s_acc=0.6923, t2s_acc_ratio=0.6923, t_preds_as_label_acc=0.9992, t2s_ce_loss=3.8045, t2s_kd_loss=2.4837, s2t_kd_loss=2.7347, lr=4.9976e-04, projector_lr=4.9976e-04, scale=1.0000
+[2026-05-28 10:36:52] [INFO]  train | epoch 001:     450 /  1429  global_step=450, loss=6.5351, nll_loss=3.7205, kd_loss=9.3496, accuracy=0.4094, micro_step_time=0.4332, step_time=0.4357, align_ratio=0.8952, t2s_agreement=0.7618, t2s_agreement_ratio=0.6918, t2s_acc=0.6916, t2s_acc_ratio=0.6916, t_preds_as_label_acc=0.9996, t2s_ce_loss=3.9805, t2s_kd_loss=2.5894, s2t_kd_loss=2.7797, lr=4.9969e-04, projector_lr=4.9969e-04, scale=1.0000
+[2026-05-28 10:37:14] [INFO]  train | epoch 001:     500 /  1429  global_step=500, loss=6.7350, nll_loss=3.7840, kd_loss=9.6861, accuracy=0.4156, micro_step_time=0.4273, step_time=0.4296, align_ratio=0.8914, t2s_agreement=0.7598, t2s_agreement_ratio=0.6823, t2s_acc=0.6823, t2s_acc_ratio=0.6823, t_preds_as_label_acc=0.9995, t2s_ce_loss=4.3116, t2s_kd_loss=2.5900, s2t_kd_loss=2.7845, lr=4.9962e-04, projector_lr=4.9962e-04, scale=1.0000
+[2026-05-28 10:37:36] [INFO]  train | epoch 001:     550 /  1429  global_step=550, loss=6.6522, nll_loss=3.7431, kd_loss=9.5613, accuracy=0.4107, micro_step_time=0.4448, step_time=0.4472, align_ratio=0.8981, t2s_agreement=0.7729, t2s_agreement_ratio=0.6983, t2s_acc=0.6984, t2s_acc_ratio=0.6984, t_preds_as_label_acc=0.9992, t2s_ce_loss=4.1336, t2s_kd_loss=2.6373, s2t_kd_loss=2.7904, lr=4.9954e-04, projector_lr=4.9954e-04, scale=1.0000
+[2026-05-28 10:37:58] [INFO]  train | epoch 001:     600 /  1429  global_step=600, loss=6.4851, nll_loss=3.6974, kd_loss=9.2728, accuracy=0.4140, micro_step_time=0.4329, step_time=0.4352, align_ratio=0.9044, t2s_agreement=0.7768, t2s_agreement_ratio=0.7084, t2s_acc=0.7084, t2s_acc_ratio=0.7084, t_preds_as_label_acc=0.9997, t2s_ce_loss=3.9834, t2s_kd_loss=2.5582, s2t_kd_loss=2.7312, lr=4.9946e-04, projector_lr=4.9946e-04, scale=1.0000
+[2026-05-28 10:38:20] [INFO]  train | epoch 001:     650 /  1429  global_step=650, loss=6.8022, nll_loss=3.8282, kd_loss=9.7763, accuracy=0.4024, micro_step_time=0.4317, step_time=0.4340, align_ratio=0.9012, t2s_agreement=0.7716, t2s_agreement_ratio=0.6985, t2s_acc=0.6985, t2s_acc_ratio=0.6985, t_preds_as_label_acc=0.9999, t2s_ce_loss=4.2997, t2s_kd_loss=2.6711, s2t_kd_loss=2.8055, lr=4.9936e-04, projector_lr=4.9936e-04, scale=1.0000
+[2026-05-28 10:38:41] [INFO]  train | epoch 001:     700 /  1429  global_step=700, loss=6.7588, nll_loss=3.7377, kd_loss=9.7799, accuracy=0.4096, micro_step_time=0.4318, step_time=0.4341, align_ratio=0.8975, t2s_agreement=0.7753, t2s_agreement_ratio=0.6978, t2s_acc=0.6977, t2s_acc_ratio=0.6977, t_preds_as_label_acc=0.9995, t2s_ce_loss=4.3832, t2s_kd_loss=2.6306, s2t_kd_loss=2.7661, lr=4.9926e-04, projector_lr=4.9926e-04, scale=1.0000
+[2026-05-28 10:39:02] [INFO]  train | epoch 001:     750 /  1429  global_step=750, loss=6.8013, nll_loss=3.7957, kd_loss=9.8069, accuracy=0.4105, micro_step_time=0.4051, step_time=0.4074, align_ratio=0.8905, t2s_agreement=0.7781, t2s_agreement_ratio=0.6970, t2s_acc=0.6969, t2s_acc_ratio=0.6969, t_preds_as_label_acc=0.9996, t2s_ce_loss=4.3819, t2s_kd_loss=2.6398, s2t_kd_loss=2.7852, lr=4.9915e-04, projector_lr=4.9915e-04, scale=1.0000
+[2026-05-28 10:39:22] [INFO]  train | epoch 001:     800 /  1429  global_step=800, loss=6.8086, nll_loss=3.7881, kd_loss=9.8291, accuracy=0.4126, micro_step_time=0.4073, step_time=0.4095, align_ratio=0.9002, t2s_agreement=0.7867, t2s_agreement_ratio=0.7137, t2s_acc=0.7136, t2s_acc_ratio=0.7136, t_preds_as_label_acc=0.9994, t2s_ce_loss=4.3678, t2s_kd_loss=2.6772, s2t_kd_loss=2.7841, lr=4.9903e-04, projector_lr=4.9903e-04, scale=1.0000
+[2026-05-28 10:39:44] [INFO]  train | epoch 001:     850 /  1429  global_step=850, loss=6.5781, nll_loss=3.7212, kd_loss=9.4351, accuracy=0.4236, micro_step_time=0.4257, step_time=0.4279, align_ratio=0.9069, t2s_agreement=0.7922, t2s_agreement_ratio=0.7231, t2s_acc=0.7230, t2s_acc_ratio=0.7230, t_preds_as_label_acc=0.9998, t2s_ce_loss=4.1387, t2s_kd_loss=2.5857, s2t_kd_loss=2.7107, lr=4.9891e-04, projector_lr=4.9891e-04, scale=1.0000
+[2026-05-28 10:40:05] [INFO]  train | epoch 001:     900 /  1429  global_step=900, loss=6.7066, nll_loss=3.7847, kd_loss=9.6285, accuracy=0.4131, micro_step_time=0.4214, step_time=0.4236, align_ratio=0.8935, t2s_agreement=0.7954, t2s_agreement_ratio=0.7211, t2s_acc=0.7211, t2s_acc_ratio=0.7211, t_preds_as_label_acc=0.9997, t2s_ce_loss=4.1757, t2s_kd_loss=2.6772, s2t_kd_loss=2.7757, lr=4.9878e-04, projector_lr=4.9878e-04, scale=1.0000
+[2026-05-28 10:40:27] [INFO]  train | epoch 001:     950 /  1429  global_step=950, loss=6.5141, nll_loss=3.6567, kd_loss=9.3714, accuracy=0.4285, micro_step_time=0.4382, step_time=0.4405, align_ratio=0.9028, t2s_agreement=0.8026, t2s_agreement_ratio=0.7212, t2s_acc=0.7211, t2s_acc_ratio=0.7211, t_preds_as_label_acc=0.9999, t2s_ce_loss=4.0678, t2s_kd_loss=2.6031, s2t_kd_loss=2.7005, lr=4.9864e-04, projector_lr=4.9864e-04, scale=1.0000
+[2026-05-28 10:40:48] [INFO]  train | epoch 001:    1000 /  1429  global_step=1000, loss=6.9789, nll_loss=3.8657, kd_loss=10.0921, accuracy=0.4006, micro_step_time=0.4117, step_time=0.4140, align_ratio=0.8947, t2s_agreement=0.7898, t2s_agreement_ratio=0.7127, t2s_acc=0.7126, t2s_acc_ratio=0.7126, t_preds_as_label_acc=0.9993, t2s_ce_loss=4.5104, t2s_kd_loss=2.7535, s2t_kd_loss=2.8282, lr=4.9849e-04, projector_lr=4.9849e-04, scale=1.0000
+[2026-05-28 10:41:08] [INFO]  train | epoch 001:    1050 /  1429  global_step=1050, loss=6.6400, nll_loss=3.6438, kd_loss=9.6363, accuracy=0.4293, micro_step_time=0.4112, step_time=0.4135, align_ratio=0.9024, t2s_agreement=0.8026, t2s_agreement_ratio=0.7286, t2s_acc=0.7283, t2s_acc_ratio=0.7283, t_preds_as_label_acc=0.9996, t2s_ce_loss=4.3244, t2s_kd_loss=2.5980, s2t_kd_loss=2.7140, lr=4.9834e-04, projector_lr=4.9834e-04, scale=1.0000
+[2026-05-28 10:41:30] [INFO]  train | epoch 001:    1100 /  1429  global_step=1100, loss=6.8764, nll_loss=3.8478, kd_loss=9.9051, accuracy=0.3950, micro_step_time=0.4410, step_time=0.4434, align_ratio=0.9051, t2s_agreement=0.7995, t2s_agreement_ratio=0.7260, t2s_acc=0.7257, t2s_acc_ratio=0.7257, t_preds_as_label_acc=0.9995, t2s_ce_loss=4.3496, t2s_kd_loss=2.7493, s2t_kd_loss=2.8061, lr=4.9818e-04, projector_lr=4.9818e-04, scale=1.0000
+[2026-05-28 10:41:52] [INFO]  train | epoch 001:    1150 /  1429  global_step=1150, loss=6.8594, nll_loss=3.7844, kd_loss=9.9344, accuracy=0.4014, micro_step_time=0.4327, step_time=0.4350, align_ratio=0.9000, t2s_agreement=0.7971, t2s_agreement_ratio=0.7240, t2s_acc=0.7240, t2s_acc_ratio=0.7240, t_preds_as_label_acc=0.9998, t2s_ce_loss=4.4161, t2s_kd_loss=2.7420, s2t_kd_loss=2.7764, lr=4.9801e-04, projector_lr=4.9801e-04, scale=1.0000
+[2026-05-28 10:42:15] [INFO]  train | epoch 001:    1200 /  1429  global_step=1200, loss=6.7340, nll_loss=3.7207, kd_loss=9.7473, accuracy=0.4217, micro_step_time=0.4503, step_time=0.4526, align_ratio=0.8986, t2s_agreement=0.8079, t2s_agreement_ratio=0.7300, t2s_acc=0.7300, t2s_acc_ratio=0.7300, t_preds_as_label_acc=0.9998, t2s_ce_loss=4.3813, t2s_kd_loss=2.6476, s2t_kd_loss=2.7185, lr=4.9783e-04, projector_lr=4.9783e-04, scale=1.0000
+[2026-05-28 10:42:37] [INFO]  train | epoch 001:    1250 /  1429  global_step=1250, loss=6.9325, nll_loss=3.8341, kd_loss=10.0310, accuracy=0.3956, micro_step_time=0.4429, step_time=0.4452, align_ratio=0.9067, t2s_agreement=0.8030, t2s_agreement_ratio=0.7309, t2s_acc=0.7308, t2s_acc_ratio=0.7308, t_preds_as_label_acc=0.9997, t2s_ce_loss=4.3808, t2s_kd_loss=2.8284, s2t_kd_loss=2.8218, lr=4.9764e-04, projector_lr=4.9764e-04, scale=1.0000
+[2026-05-28 10:42:58] [INFO]  train | epoch 001:    1300 /  1429  global_step=1300, loss=6.6117, nll_loss=3.6930, kd_loss=9.5304, accuracy=0.4185, micro_step_time=0.4263, step_time=0.4286, align_ratio=0.9109, t2s_agreement=0.8158, t2s_agreement_ratio=0.7489, t2s_acc=0.7490, t2s_acc_ratio=0.7490, t_preds_as_label_acc=0.9996, t2s_ce_loss=4.1896, t2s_kd_loss=2.6526, s2t_kd_loss=2.6883, lr=4.9745e-04, projector_lr=4.9745e-04, scale=1.0000
+[2026-05-28 10:43:20] [INFO]  train | epoch 001:    1350 /  1429  global_step=1350, loss=6.6494, nll_loss=3.5892, kd_loss=9.7097, accuracy=0.4374, micro_step_time=0.4295, step_time=0.4318, align_ratio=0.9009, t2s_agreement=0.8143, t2s_agreement_ratio=0.7298, t2s_acc=0.7295, t2s_acc_ratio=0.7295, t_preds_as_label_acc=0.9996, t2s_ce_loss=4.4780, t2s_kd_loss=2.5836, s2t_kd_loss=2.6481, lr=4.9725e-04, projector_lr=4.9725e-04, scale=1.0000
+[2026-05-28 10:43:43] [INFO]  train | epoch 001:    1400 /  1429  global_step=1400, loss=6.8974, nll_loss=3.8426, kd_loss=9.9522, accuracy=0.4030, micro_step_time=0.4542, step_time=0.4564, align_ratio=0.9057, t2s_agreement=0.8056, t2s_agreement_ratio=0.7328, t2s_acc=0.7327, t2s_acc_ratio=0.7327, t_preds_as_label_acc=0.9997, t2s_ce_loss=4.4144, t2s_kd_loss=2.7716, s2t_kd_loss=2.7662, lr=4.9705e-04, projector_lr=4.9705e-04, scale=1.0000
+[2026-05-28 10:43:55] [INFO]  End of epoch 1
+[2026-05-28 10:43:55] [INFO]  train | epoch 001 | loss 6.5449 | nll_loss 3.6624 | kd_loss 9.4274
+[2026-05-28 10:43:55] [INFO]  Evaluating before saving model...
+[2026-05-28 10:43:55] [INFO]  Evaluating on dev set with 1 GPU(s)
+[transformers] Passing `generation_config` together with generation-related arguments=({'max_new_tokens'}) is deprecated and will be removed in future versions. Please pass either a `generation_config` object OR all generation parameters explicitly, but not both.
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[2026-05-28 10:46:22] [INFO]  eval_results in run@1: {'exact_match': 2.7, 'rougeL': 20.9859}
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[2026-05-28 10:48:25] [INFO]  eval_results in run@2: {'exact_match': 3.2, 'rougeL': 21.792}
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[2026-05-28 10:50:34] [INFO]  eval_results in run@3: {'exact_match': 2.8, 'rougeL': 21.4007}
+[2026-05-28 10:50:34] [INFO]  dev | {'loss': 3.596543, 'token_num': 75795, 'token_acc': 0.423181, 'top1_prob': 0.632258} | {'exact_match': 2.9, 'rougeL': 21.3929}
+[2026-05-28 10:50:34] [INFO]  Saving tokenizer...
+[2026-05-28 10:50:34] [INFO]  Saving model...
+
+[2026-05-28 10:50:46] [INFO]  Model has been saved to ./outputs/gpt2/gpt2-medium/dskd_v2_eta/forward_kl-bf16__teacher_qwen1.5__kd^rate0.5__kd^temp2.0__epoch20__bsz8x1x1x1_8__lr0.0005__proj^lr0.001/epoch1_step1430_loss3.5965_rougel21.3929
+[2026-05-28 10:50:46] [INFO]  Start iterations of epoch 2
+[2026-05-28 10:50:56] [INFO]  train | epoch 002:      20 /  1429  global_step=1450, loss=6.0438, nll_loss=3.4318, kd_loss=8.6558, accuracy=0.4401, micro_step_time=0.4333, step_time=0.4357, align_ratio=0.8957, t2s_agreement=0.8430, t2s_agreement_ratio=0.7676, t2s_acc=0.7677, t2s_acc_ratio=0.7677, t_preds_as_label_acc=0.9997, t2s_ce_loss=3.4432, t2s_kd_loss=2.6017, s2t_kd_loss=2.6109, lr=4.9683e-04, projector_lr=4.9683e-04, scale=1.0000
+[2026-05-28 10:51:19] [INFO]  train | epoch 002:      70 /  1429  global_step=1500, loss=4.4686, nll_loss=2.7859, kd_loss=6.1513, accuracy=0.5042, micro_step_time=0.4573, step_time=0.4598, align_ratio=0.9042, t2s_agreement=0.9029, t2s_agreement_ratio=0.8195, t2s_acc=0.8193, t2s_acc_ratio=0.8193, t_preds_as_label_acc=0.9996, t2s_ce_loss=1.5470, t2s_kd_loss=2.3007, s2t_kd_loss=2.3036, lr=4.9661e-04, projector_lr=4.9661e-04, scale=1.0000
+[2026-05-28 10:51:40] [INFO]  train | epoch 002:     120 /  1429  global_step=1550, loss=4.5086, nll_loss=2.8773, kd_loss=6.1400, accuracy=0.4956, micro_step_time=0.4346, step_time=0.4370, align_ratio=0.8998, t2s_agreement=0.9083, t2s_agreement_ratio=0.8179, t2s_acc=0.8178, t2s_acc_ratio=0.8178, t_preds_as_label_acc=0.9997, t2s_ce_loss=1.4316, t2s_kd_loss=2.3587, s2t_kd_loss=2.3497, lr=4.9638e-04, projector_lr=4.9638e-04, scale=1.0000
+[2026-05-28 10:52:03] [INFO]  train | epoch 002:     170 /  1429  global_step=1600, loss=4.7026, nll_loss=2.9674, kd_loss=6.4377, accuracy=0.4800, micro_step_time=0.4432, step_time=0.4456, align_ratio=0.9037, t2s_agreement=0.9020, t2s_agreement_ratio=0.8236, t2s_acc=0.8237, t2s_acc_ratio=0.8237, t_preds_as_label_acc=0.9996, t2s_ce_loss=1.6217, t2s_kd_loss=2.4064, s2t_kd_loss=2.4096, lr=4.9614e-04, projector_lr=4.9614e-04, scale=1.0000
+[2026-05-28 10:52:25] [INFO]  train | epoch 002:     220 /  1429  global_step=1650, loss=4.6409, nll_loss=2.8855, kd_loss=6.3963, accuracy=0.4939, micro_step_time=0.4517, step_time=0.4542, align_ratio=0.8979, t2s_agreement=0.8972, t2s_agreement_ratio=0.8055, t2s_acc=0.8052, t2s_acc_ratio=0.8052, t_preds_as_label_acc=0.9994, t2s_ce_loss=1.6592, t2s_kd_loss=2.3616, s2t_kd_loss=2.3756, lr=4.9590e-04, projector_lr=4.9590e-04, scale=1.0000
+[2026-05-28 10:52:49] [INFO]  train | epoch 002:     270 /  1429  global_step=1700, loss=4.8101, nll_loss=2.9848, kd_loss=6.6353, accuracy=0.4850, micro_step_time=0.4655, step_time=0.4680, align_ratio=0.8960, t2s_agreement=0.8943, t2s_agreement_ratio=0.8104, t2s_acc=0.8103, t2s_acc_ratio=0.8103, t_preds_as_label_acc=0.9999, t2s_ce_loss=1.8540, t2s_kd_loss=2.3868, s2t_kd_loss=2.3945, lr=4.9565e-04, projector_lr=4.9565e-04, scale=1.0000
+[2026-05-28 10:53:12] [INFO]  train | epoch 002:     320 /  1429  global_step=1750, loss=4.9264, nll_loss=3.0439, kd_loss=6.8088, accuracy=0.4778, micro_step_time=0.4518, step_time=0.4542, align_ratio=0.8884, t2s_agreement=0.8935, t2s_agreement_ratio=0.8013, t2s_acc=0.8011, t2s_acc_ratio=0.8011, t_preds_as_label_acc=0.9994, t2s_ce_loss=1.9246, t2s_kd_loss=2.4548, s2t_kd_loss=2.4294, lr=4.9539e-04, projector_lr=4.9539e-04, scale=1.0000
+[2026-05-28 10:53:35] [INFO]  train | epoch 002:     370 /  1429  global_step=1800, loss=4.8748, nll_loss=2.9848, kd_loss=6.7649, accuracy=0.4833, micro_step_time=0.4711, step_time=0.4735, align_ratio=0.8938, t2s_agreement=0.8914, t2s_agreement_ratio=0.8018, t2s_acc=0.8016, t2s_acc_ratio=0.8016, t_preds_as_label_acc=0.9996, t2s_ce_loss=1.9875, t2s_kd_loss=2.3798, s2t_kd_loss=2.3975, lr=4.9512e-04, projector_lr=4.9512e-04, scale=1.0000
+[2026-05-28 10:54:00] [INFO]  train | epoch 002:     420 /  1429  global_step=1850, loss=4.8761, nll_loss=2.9907, kd_loss=6.7615, accuracy=0.4797, micro_step_time=0.4834, step_time=0.4858, align_ratio=0.9018, t2s_agreement=0.8921, t2s_agreement_ratio=0.8109, t2s_acc=0.8107, t2s_acc_ratio=0.8107, t_preds_as_label_acc=0.9996, t2s_ce_loss=1.9300, t2s_kd_loss=2.4376, s2t_kd_loss=2.3940, lr=4.9485e-04, projector_lr=4.9485e-04, scale=1.0000
+[2026-05-28 10:54:22] [INFO]  train | epoch 002:     470 /  1429  global_step=1900, loss=4.9311, nll_loss=2.9828, kd_loss=6.8794, accuracy=0.4838, micro_step_time=0.4365, step_time=0.4389, align_ratio=0.8990, t2s_agreement=0.8845, t2s_agreement_ratio=0.7991, t2s_acc=0.7985, t2s_acc_ratio=0.7985, t_preds_as_label_acc=0.9991, t2s_ce_loss=2.1373, t2s_kd_loss=2.3567, s2t_kd_loss=2.3854, lr=4.9457e-04, projector_lr=4.9457e-04, scale=1.0000
+[2026-05-28 10:54:43] [INFO]  train | epoch 002:     520 /  1429  global_step=1950, loss=4.8309, nll_loss=2.9759, kd_loss=6.6859, accuracy=0.4872, micro_step_time=0.4370, step_time=0.4395, align_ratio=0.8932, t2s_agreement=0.8909, t2s_agreement_ratio=0.8037, t2s_acc=0.8036, t2s_acc_ratio=0.8036, t_preds_as_label_acc=0.9997, t2s_ce_loss=1.9532, t2s_kd_loss=2.3688, s2t_kd_loss=2.3639, lr=4.9428e-04, projector_lr=4.9428e-04, scale=1.0000
+[2026-05-28 10:55:06] [INFO]  train | epoch 002:     570 /  1429  global_step=2000, loss=4.9772, nll_loss=3.0522, kd_loss=6.9023, accuracy=0.4721, micro_step_time=0.4430, step_time=0.4454, align_ratio=0.9023, t2s_agreement=0.8895, t2s_agreement_ratio=0.8067, t2s_acc=0.8065, t2s_acc_ratio=0.8065, t_preds_as_label_acc=0.9998, t2s_ce_loss=2.0693, t2s_kd_loss=2.4226, s2t_kd_loss=2.4103, lr=4.9398e-04, projector_lr=4.9398e-04, scale=1.0000
+[2026-05-28 10:55:28] [INFO]  train | epoch 002:     620 /  1429  global_step=2050, loss=4.9762, nll_loss=3.0195, kd_loss=6.9328, accuracy=0.4806, micro_step_time=0.4423, step_time=0.4446, align_ratio=0.9004, t2s_agreement=0.8864, t2s_agreement_ratio=0.8061, t2s_acc=0.8061, t2s_acc_ratio=0.8061, t_preds_as_label_acc=0.9999, t2s_ce_loss=2.1496, t2s_kd_loss=2.3935, s2t_kd_loss=2.3897, lr=4.9368e-04, projector_lr=4.9368e-04, scale=1.0000
+[2026-05-28 10:55:50] [INFO]  train | epoch 002:     670 /  1429  global_step=2100, loss=4.9890, nll_loss=3.0476, kd_loss=6.9304, accuracy=0.4745, micro_step_time=0.4426, step_time=0.4450, align_ratio=0.9035, t2s_agreement=0.8820, t2s_agreement_ratio=0.8053, t2s_acc=0.8052, t2s_acc_ratio=0.8052, t_preds_as_label_acc=0.9995, t2s_ce_loss=2.1666, t2s_kd_loss=2.3852, s2t_kd_loss=2.3785, lr=4.9337e-04, projector_lr=4.9337e-04, scale=1.0000
+[2026-05-28 10:56:13] [INFO]  train | epoch 002:     720 /  1429  global_step=2150, loss=4.9352, nll_loss=2.9921, kd_loss=6.8783, accuracy=0.4799, micro_step_time=0.4615, step_time=0.4640, align_ratio=0.9030, t2s_agreement=0.8912, t2s_agreement_ratio=0.8065, t2s_acc=0.8061, t2s_acc_ratio=0.8061, t_preds_as_label_acc=0.9991, t2s_ce_loss=2.1175, t2s_kd_loss=2.3903, s2t_kd_loss=2.3705, lr=4.9305e-04, projector_lr=4.9305e-04, scale=1.0000
+[2026-05-28 10:56:37] [INFO]  train | epoch 002:     770 /  1429  global_step=2200, loss=5.0018, nll_loss=3.0242, kd_loss=6.9794, accuracy=0.4745, micro_step_time=0.4667, step_time=0.4692, align_ratio=0.9048, t2s_agreement=0.8841, t2s_agreement_ratio=0.8074, t2s_acc=0.8073, t2s_acc_ratio=0.8073, t_preds_as_label_acc=0.9998, t2s_ce_loss=2.2377, t2s_kd_loss=2.3699, s2t_kd_loss=2.3718, lr=4.9273e-04, projector_lr=4.9273e-04, scale=1.0000
+[2026-05-28 10:57:00] [INFO]  train | epoch 002:     820 /  1429  global_step=2250, loss=5.0836, nll_loss=3.0118, kd_loss=7.1555, accuracy=0.4762, micro_step_time=0.4618, step_time=0.4642, align_ratio=0.8984, t2s_agreement=0.8809, t2s_agreement_ratio=0.7999, t2s_acc=0.8000, t2s_acc_ratio=0.8000, t_preds_as_label_acc=0.9998, t2s_ce_loss=2.4227, t2s_kd_loss=2.3837, s2t_kd_loss=2.3491, lr=4.9239e-04, projector_lr=4.9239e-04, scale=1.0000
+[2026-05-28 10:57:24] [INFO]  train | epoch 002:     870 /  1429  global_step=2300, loss=5.1423, nll_loss=3.0634, kd_loss=7.2212, accuracy=0.4719, micro_step_time=0.4820, step_time=0.4844, align_ratio=0.9007, t2s_agreement=0.8792, t2s_agreement_ratio=0.7970, t2s_acc=0.7969, t2s_acc_ratio=0.7969, t_preds_as_label_acc=0.9995, t2s_ce_loss=2.4638, t2s_kd_loss=2.3864, s2t_kd_loss=2.3709, lr=4.9205e-04, projector_lr=4.9205e-04, scale=1.0000
+[2026-05-28 10:57:48] [INFO]  train | epoch 002:     920 /  1429  global_step=2350, loss=5.3226, nll_loss=3.1412, kd_loss=7.5040, accuracy=0.4529, micro_step_time=0.4707, step_time=0.4732, align_ratio=0.9027, t2s_agreement=0.8754, t2s_agreement_ratio=0.8011, t2s_acc=0.8010, t2s_acc_ratio=0.8010, t_preds_as_label_acc=0.9996, t2s_ce_loss=2.5444, t2s_kd_loss=2.5051, s2t_kd_loss=2.4545, lr=4.9171e-04, projector_lr=4.9171e-04, scale=1.0000
+[2026-05-28 10:58:09] [INFO]  train | epoch 002:     970 /  1429  global_step=2400, loss=5.1305, nll_loss=2.9722, kd_loss=7.2888, accuracy=0.4931, micro_step_time=0.4276, step_time=0.4301, align_ratio=0.8929, t2s_agreement=0.8751, t2s_agreement_ratio=0.7827, t2s_acc=0.7827, t2s_acc_ratio=0.7827, t_preds_as_label_acc=0.9995, t2s_ce_loss=2.6644, t2s_kd_loss=2.2966, s2t_kd_loss=2.3278, lr=4.9135e-04, projector_lr=4.9135e-04, scale=1.0000
+[2026-05-28 10:58:32] [INFO]  train | epoch 002:    1020 /  1429  global_step=2450, loss=5.2676, nll_loss=3.0355, kd_loss=7.4997, accuracy=0.4809, micro_step_time=0.4425, step_time=0.4449, align_ratio=0.9029, t2s_agreement=0.8741, t2s_agreement_ratio=0.7927, t2s_acc=0.7924, t2s_acc_ratio=0.7924, t_preds_as_label_acc=0.9995, t2s_ce_loss=2.7404, t2s_kd_loss=2.3806, s2t_kd_loss=2.3787, lr=4.9099e-04, projector_lr=4.9099e-04, scale=1.0000
+[2026-05-28 10:58:53] [INFO]  train | epoch 002:    1070 /  1429  global_step=2500, loss=5.2331, nll_loss=3.0306, kd_loss=7.4356, accuracy=0.4817, micro_step_time=0.4248, step_time=0.4270, align_ratio=0.9023, t2s_agreement=0.8777, t2s_agreement_ratio=0.7997, t2s_acc=0.7995, t2s_acc_ratio=0.7995, t_preds_as_label_acc=0.9997, t2s_ce_loss=2.6374, t2s_kd_loss=2.4008, s2t_kd_loss=2.3974, lr=4.9062e-04, projector_lr=4.9062e-04, scale=1.0000
+[2026-05-28 10:59:16] [INFO]  train | epoch 002:    1120 /  1429  global_step=2550, loss=5.1095, nll_loss=3.0035, kd_loss=7.2154, accuracy=0.4795, micro_step_time=0.4500, step_time=0.4523, align_ratio=0.9030, t2s_agreement=0.8789, t2s_agreement_ratio=0.7954, t2s_acc=0.7950, t2s_acc_ratio=0.7950, t_preds_as_label_acc=0.9995, t2s_ce_loss=2.5539, t2s_kd_loss=2.3317, s2t_kd_loss=2.3298, lr=4.9024e-04, projector_lr=4.9024e-04, scale=1.0000
+[2026-05-28 10:59:38] [INFO]  train | epoch 002:    1170 /  1429  global_step=2600, loss=5.3187, nll_loss=3.0869, kd_loss=7.5505, accuracy=0.4709, micro_step_time=0.4379, step_time=0.4402, align_ratio=0.9035, t2s_agreement=0.8741, t2s_agreement_ratio=0.7908, t2s_acc=0.7907, t2s_acc_ratio=0.7907, t_preds_as_label_acc=0.9997, t2s_ce_loss=2.7752, t2s_kd_loss=2.3921, s2t_kd_loss=2.3832, lr=4.8986e-04, projector_lr=4.8986e-04, scale=1.0000
+[2026-05-28 10:59:59] [INFO]  train | epoch 002:    1220 /  1429  global_step=2650, loss=5.1238, nll_loss=2.9515, kd_loss=7.2961, accuracy=0.4874, micro_step_time=0.4271, step_time=0.4294, align_ratio=0.9009, t2s_agreement=0.8765, t2s_agreement_ratio=0.7921, t2s_acc=0.7918, t2s_acc_ratio=0.7918, t_preds_as_label_acc=0.9995, t2s_ce_loss=2.6850, t2s_kd_loss=2.2965, s2t_kd_loss=2.3145, lr=4.8947e-04, projector_lr=4.8947e-04, scale=1.0000
+[2026-05-28 11:00:20] [INFO]  train | epoch 002:    1270 /  1429  global_step=2700, loss=5.1706, nll_loss=2.9830, kd_loss=7.3583, accuracy=0.4849, micro_step_time=0.4155, step_time=0.4179, align_ratio=0.9037, t2s_agreement=0.8811, t2s_agreement_ratio=0.8037, t2s_acc=0.8033, t2s_acc_ratio=0.8033, t_preds_as_label_acc=0.9993, t2s_ce_loss=2.6623, t2s_kd_loss=2.3509, s2t_kd_loss=2.3452, lr=4.8907e-04, projector_lr=4.8907e-04, scale=1.0000
+[2026-05-28 11:00:41] [INFO]  train | epoch 002:    1320 /  1429  global_step=2750, loss=5.3197, nll_loss=3.0337, kd_loss=7.6058, accuracy=0.4780, micro_step_time=0.4244, step_time=0.4267, align_ratio=0.8969, t2s_agreement=0.8747, t2s_agreement_ratio=0.7878, t2s_acc=0.7876, t2s_acc_ratio=0.7876, t_preds_as_label_acc=0.9997, t2s_ce_loss=2.8757, t2s_kd_loss=2.3725, s2t_kd_loss=2.3576, lr=4.8867e-04, projector_lr=4.8867e-04, scale=1.0000
+[2026-05-28 11:01:03] [INFO]  train | epoch 002:    1370 /  1429  global_step=2800, loss=5.5716, nll_loss=3.2020, kd_loss=7.9412, accuracy=0.4565, micro_step_time=0.4323, step_time=0.4345, align_ratio=0.8988, t2s_agreement=0.8700, t2s_agreement_ratio=0.7890, t2s_acc=0.7890, t2s_acc_ratio=0.7890, t_preds_as_label_acc=0.9997, t2s_ce_loss=2.9875, t2s_kd_loss=2.4911, s2t_kd_loss=2.4627, lr=4.8825e-04, projector_lr=4.8825e-04, scale=1.0000
+[2026-05-28 11:01:26] [INFO]  train | epoch 002:    1420 /  1429  global_step=2850, loss=5.6913, nll_loss=3.1559, kd_loss=8.2267, accuracy=0.4593, micro_step_time=0.4622, step_time=0.4645, align_ratio=0.9060, t2s_agreement=0.8632, t2s_agreement_ratio=0.7834, t2s_acc=0.7833, t2s_acc_ratio=0.7833, t_preds_as_label_acc=0.9996, t2s_ce_loss=3.3403, t2s_kd_loss=2.4640, s2t_kd_loss=2.4225, lr=4.8783e-04, projector_lr=

opt/opt-2.7b/mta_dskd_v2_eta/forward_kl-bf16__teacher_qwen2.5__kd^rate0.5__kd^temp2.0__epoch15__bsz8x1x1x1_8__lr0.001__proj^lr0.001/train.log

@@ -2184,3 +2184,12 @@ torch.distributed.elastic.multiprocessing.api.SignalException: Process 3552093 g
 [transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
 [transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
 [transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[transformers] Both `max_new_tokens` (=256) and `max_length`(=512) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[2026-05-27 08:21:30] [INFO]  eval_results in run@3: {'exact_match': 5.5, 'rougeL': 30.9738}
+[2026-05-27 08:21:30] [INFO]  dev | {'loss': 3.307975, 'token_num': 75795, 'token_acc': 0.528702, 'top1_prob': 0.790553} | {'exact_match': 5.5667, 'rougeL': 30.9747}
+[2026-05-27 08:21:30] [INFO]  Saving tokenizer...
+[2026-05-27 08:21:30] [INFO]  Saving model...
+[2026-05-27 08:21:30] [INFO]  Model has been saved to ./outputs/opt/opt-2.7b/mta_dskd_v2_eta/forward_kl-bf16__teacher_qwen2.5__kd^rate0.5__kd^temp2.0__epoch15__bsz8x1x1x1_8__lr0.001__proj^lr0.001/epoch15_step21450_loss3.3080_rougel30.9747
+[2026-05-27 08:21:30] [INFO]  Done training in 16:53:43