Upload best model checkpoint

config.json

@@ -0,0 +1,38 @@
+{
+  "activation_function": "gelu_new",
+  "architectures": [
+    "GPT2LMHeadModel"
+  ],
+  "attn_pdrop": 0.1,
+  "bos_token_id": 50256,
+  "dtype": "float32",
+  "embd_pdrop": 0.1,
+  "eos_token_id": 50256,
+  "initializer_range": 0.02,
+  "layer_norm_epsilon": 1e-05,
+  "model_type": "gpt2",
+  "n_ctx": 1024,
+  "n_embd": 768,
+  "n_head": 12,
+  "n_inner": null,
+  "n_layer": 12,
+  "n_positions": 1024,
+  "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
+    }
+  },
+  "transformers_version": "4.57.3",
+  "use_cache": true,
+  "vocab_size": 50257
+}

generation_config.json

@@ -0,0 +1,6 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 50256,
+  "eos_token_id": 50256,
+  "transformers_version": "4.57.3"
+}

metrics.json

@@ -0,0 +1,31 @@
+{
+    "rl_info/A2G": -0.6658737659454346,
+    "rl_info/entropy": 0.7045262455940247,
+    "rl_info/total_token": 650.0,
+    "rl_info/advantage_b4_norm": -2206.338134765625,
+    "rl_info/advantage_after_gnorm": -1.0411098003387451,
+    "rl_info/kl_w_ref": -0.0003157879982609302,
+    "rl_info/clip_ratio": 0.936923086643219,
+    "train/rl_loss": 66.5872802734375,
+    "train/total_loss": 8.323410034179688,
+    "xsum/rouge1": 0.17478446053644453,
+    "xsum/rouge2": 0.0270244843023609,
+    "xsum/rougeL": 0.13926067850589546,
+    "xsum/rougeLsum": 0.14284803751147157,
+    "xsum/bertscore_precision": 0.7250652085152888,
+    "xsum/bertscore_recall": 0.7281463338055265,
+    "xsum/bertscore_f1": 0.7262238959629153,
+    "eval_agg/avg_all_rougef": 0.12097941521404312,
+    "eval_agg/avg_all_bertf": 0.7262238959629153,
+    "eval_agg/avg_all": 0.4236016555884792,
+    "num_rl_rollout": 250,
+    "lm_epoch": 0,
+    "rl_epoch": 0,
+    "step": 2000,
+    "total_data_token": 3685864,
+    "total_rl_token": 8807121,
+    "total_lm_token": 0,
+    "total_token": 8807121,
+    "completed_steps": 2000,
+    "tune_objective": 0.41038940313189515
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5dc913c5fe263bf822ac5cea8b4d84459769d9fc569cda85f78bdccc846ddd7e
+size 497774208

special_tokens_map.json

@@ -0,0 +1,6 @@
+{
+  "bos_token": "<|endoftext|>",
+  "eos_token": "<|endoftext|>",
+  "pad_token": "<|endoftext|>",
+  "unk_token": "<|endoftext|>"
+}

tokenizer_config.json

@@ -0,0 +1,21 @@
+{
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "50256": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "bos_token": "<|endoftext|>",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "extra_special_tokens": {},
+  "model_max_length": 1024,
+  "pad_token": "<|endoftext|>",
+  "tokenizer_class": "GPT2Tokenizer",
+  "unk_token": "<|endoftext|>"
+}

train_configs.json

@@ -0,0 +1,112 @@
+{
+    "dataset_name": "nbtpj/summ_ds_train",
+    "dataset_config_name": null,
+    "train_split_name": "sim_with_one_golden__xsum_train",
+    "text_template": "{text}\nTL;DR: {summary}",
+    "label_col": "summary",
+    "freeze_role2": true,
+    "only_train_role1": true,
+    "model_name_or_path": "gpt2",
+    "ref_role1_name_or_path": "nbtpj/summ_gpt2_tldr_xsum",
+    "ref_role2_name_or_path": "gpt2",
+    "pretrained_role2_name_or_path": "gpt2",
+    "config_name": null,
+    "vectorizer_path": "/common/home/users/m/mq.nguyen.2023/testcode/SAC_LM/module9_clmv3/vectorizer/wikitext103_tfidf_full.joblib",
+    "tokenizer_name": null,
+    "use_slow_tokenizer": false,
+    "per_device_train_batch_size": 16,
+    "per_device_query_rollout_batch_size": 32,
+    "per_device_eval_batch_size": 16,
+    "vllm_vram_ratio": 0.3,
+    "learning_rate": 3e-07,
+    "grad_norm": 0.5,
+    "weight_decay": 1e-05,
+    "max_train_steps": 100000,
+    "max_train_rollouts": 100000,
+    "gradient_accumulation_steps": 1,
+    "lr_scheduler_type": "constant",
+    "num_warmup_steps": 200,
+    "seed": 0,
+    "model_type": null,
+    "block_size": 1024,
+    "mini_epoch": 2,
+    "rollout_game": "baseline3v2",
+    "rl_algo": "on_policy",
+    "constraint_type": "kl",
+    "clamp_update": false,
+    "clamp_ref": true,
+    "no_is_correction": false,
+    "rl_w": 1.0,
+    "lm_w": 0.0,
+    "n_generate": 4,
+    "n_augment": 0,
+    "gradient_checkpoint": false,
+    "group_relative_norm": true,
+    "sample_config": {
+        "do_sample": true,
+        "min_new_tokens": 5,
+        "temperature": 1.0
+    },
+    "inference_config": {
+        "do_sample": true,
+        "temperature": 0.0,
+        "min_new_tokens": 15,
+        "max_new_tokens": 40
+    },
+    "rollout_config": {
+        "accuracy_w": 1000.0,
+        "len_pen": 1.0,
+        "accuracy_w2": 0.01,
+        "len_pen2": 1.0,
+        "threshold": 0.002,
+        "similarity_fn": "rouge",
+        "acc_scale": "log"
+    },
+    "ent_coef": 0.0001,
+    "beta_coef": "0.1",
+    "prompt_0": "{text}",
+    "prompt_1": "{text}\nTL;DR: ",
+    "prompt_2": "Given the text: {role1_output}\nReconstruct the summarized text to the detailed:",
+    "prompt_eval": "{text}\nTL;DR:",
+    "epsilon": 0.2,
+    "a2g_norm": true,
+    "vllm_sleep": true,
+    "lora": false,
+    "need_attn_mask": true,
+    "gamma": 0.95,
+    "trust_remote_code": true,
+    "test_glue": false,
+    "test_clm": false,
+    "causal_model": true,
+    "test_gen": true,
+    "log_rollout_txt": true,
+    "trunc_eval": 20000,
+    "buffer_max_size": 20000,
+    "trunc_evals": [
+        "xsum___20000"
+    ],
+    "use_deepspeed": false,
+    "zero_config": 2,
+    "log_interval": "5m",
+    "eval_interval": "2_000",
+    "checkpoint_interval": "2_000",
+    "lm_fraction": -1.0,
+    "push_to_hub": null,
+    "keep_eval_size": false,
+    "mixed_precision": "bf16",
+    "tune_metrics": [
+        "xsum/rouge1___1.0",
+        "xsum/rouge2___2.0",
+        "xsum/bertscore_f1___0.25"
+    ],
+    "base_path": "/common/home/users/m/mq.nguyen.2023/testcode/SAC_LM/module9_clmv3",
+    "script": "/common/home/users/m/mq.nguyen.2023/testcode/SAC_LM/module9_clmv3/execute/role_ablation/role_ablation.py",
+    "train_from_raw": true,
+    "world_size": 1,
+    "cpu_per_worker": 6,
+    "gpu_per_worker": 1,
+    "effective_batch_size": 16,
+    "text_col": "text",
+    "_scenario": "clamp_sft_ref",
+    "_dataset": "xsum"
+}