stage best wce model after hyp search

checkpoint-48000/config.json

@@ -1,46 +0,0 @@
-{
-  "_name_or_path": "Helsinki-NLP/opus-mt-tc-big-en-fr",
-  "activation_dropout": 0.0,
-  "activation_function": "relu",
-  "architectures": [
-    "MarianMTModel"
-  ],
-  "attention_dropout": 0.0,
-  "bad_words_ids": [
-    [
-      53016
-    ]
-  ],
-  "bos_token_id": 0,
-  "classifier_dropout": 0.0,
-  "d_model": 1024,
-  "decoder_attention_heads": 16,
-  "decoder_ffn_dim": 4096,
-  "decoder_layerdrop": 0.0,
-  "decoder_layers": 6,
-  "decoder_start_token_id": 53016,
-  "decoder_vocab_size": 53017,
-  "dropout": 0.1,
-  "encoder_attention_heads": 16,
-  "encoder_ffn_dim": 4096,
-  "encoder_layerdrop": 0.0,
-  "encoder_layers": 6,
-  "eos_token_id": 43311,
-  "forced_eos_token_id": 43311,
-  "init_std": 0.02,
-  "is_encoder_decoder": true,
-  "max_length": 512,
-  "max_position_embeddings": 1024,
-  "model_type": "marian",
-  "normalize_embedding": false,
-  "num_beams": 4,
-  "num_hidden_layers": 6,
-  "pad_token_id": 53016,
-  "scale_embedding": true,
-  "share_encoder_decoder_embeddings": true,
-  "static_position_embeddings": true,
-  "torch_dtype": "float32",
-  "transformers_version": "4.28.0",
-  "use_cache": true,
-  "vocab_size": 53017
-}

checkpoint-48000/generation_config.json

@@ -1,15 +0,0 @@
-{
-  "bad_words_ids": [
-    [
-      53016
-    ]
-  ],
-  "bos_token_id": 0,
-  "decoder_start_token_id": 53016,
-  "eos_token_id": 43311,
-  "forced_eos_token_id": 43311,
-  "max_length": 512,
-  "num_beams": 4,
-  "pad_token_id": 53016,
-  "transformers_version": "4.28.0"
-}

checkpoint-48000/pytorch_model.bin

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:033a7f99e720b4906a371665a4a87d95d5707e9f03a48714c113d2e1db26282d
-size 922885701

checkpoint-48000/special_tokens_map.json

@@ -1,5 +0,0 @@
-{
-  "eos_token": "</s>",
-  "pad_token": "<pad>",
-  "unk_token": "<unk>"
-}

checkpoint-48000/tokenizer_config.json

@@ -1,12 +0,0 @@
-{
-  "clean_up_tokenization_spaces": true,
-  "eos_token": "</s>",
-  "model_max_length": 512,
-  "pad_token": "<pad>",
-  "separate_vocabs": false,
-  "source_lang": "en",
-  "sp_model_kwargs": {},
-  "target_lang": "fr",
-  "tokenizer_class": "MarianTokenizer",
-  "unk_token": "<unk>"
-}

checkpoint-48000/trainer_state.json

@@ -1,112 +0,0 @@
-{
-  "best_metric": 43.4061,
-  "best_model_checkpoint": "opus_big_enfr_FT_wce/checkpoint-48000",
-  "epoch": 1.1829944547134936,
-  "global_step": 48000,
-  "is_hyper_param_search": false,
-  "is_local_process_zero": true,
-  "is_world_process_zero": true,
-  "log_history": [
-    {
-      "epoch": 0.2,
-      "learning_rate": 1.9753635243376465e-05,
-      "loss": 0.1935,
-      "step": 8000
-    },
-    {
-      "epoch": 0.2,
-      "eval_bleu": 42.2657,
-      "eval_gen_len": 33.6098,
-      "eval_loss": 0.09848988056182861,
-      "eval_runtime": 153.6166,
-      "eval_samples_per_second": 6.79,
-      "eval_steps_per_second": 0.43,
-      "step": 8000
-    },
-    {
-      "epoch": 0.39,
-      "learning_rate": 1.950723967960567e-05,
-      "loss": 0.162,
-      "step": 16000
-    },
-    {
-      "epoch": 0.39,
-      "eval_bleu": 42.8204,
-      "eval_gen_len": 34.3816,
-      "eval_loss": 0.09720870107412338,
-      "eval_runtime": 158.3909,
-      "eval_samples_per_second": 6.585,
-      "eval_steps_per_second": 0.417,
-      "step": 16000
-    },
-    {
-      "epoch": 0.59,
-      "learning_rate": 1.9260874922982132e-05,
-      "loss": 0.1575,
-      "step": 24000
-    },
-    {
-      "epoch": 0.59,
-      "eval_bleu": 42.6214,
-      "eval_gen_len": 34.3087,
-      "eval_loss": 0.0975893884897232,
-      "eval_runtime": 129.8641,
-      "eval_samples_per_second": 8.031,
-      "eval_steps_per_second": 0.508,
-      "step": 24000
-    },
-    {
-      "epoch": 0.79,
-      "learning_rate": 1.901447935921134e-05,
-      "loss": 0.1533,
-      "step": 32000
-    },
-    {
-      "epoch": 0.79,
-      "eval_bleu": 42.7095,
-      "eval_gen_len": 34.1035,
-      "eval_loss": 0.09651403874158859,
-      "eval_runtime": 109.2582,
-      "eval_samples_per_second": 9.546,
-      "eval_steps_per_second": 0.604,
-      "step": 32000
-    },
-    {
-      "epoch": 0.99,
-      "learning_rate": 1.876814540973506e-05,
-      "loss": 0.1517,
-      "step": 40000
-    },
-    {
-      "epoch": 0.99,
-      "eval_bleu": 42.6552,
-      "eval_gen_len": 33.7267,
-      "eval_loss": 0.0962938442826271,
-      "eval_runtime": 121.4941,
-      "eval_samples_per_second": 8.585,
-      "eval_steps_per_second": 0.543,
-      "step": 40000
-    },
-    {
-      "epoch": 1.18,
-      "learning_rate": 1.852181146025878e-05,
-      "loss": 0.1352,
-      "step": 48000
-    },
-    {
-      "epoch": 1.18,
-      "eval_bleu": 43.4061,
-      "eval_gen_len": 33.8447,
-      "eval_loss": 0.09698126465082169,
-      "eval_runtime": 107.0828,
-      "eval_samples_per_second": 9.74,
-      "eval_steps_per_second": 0.616,
-      "step": 48000
-    }
-  ],
-  "max_steps": 649200,
-  "num_train_epochs": 16,
-  "total_flos": 1.5506756153848627e+17,
-  "trial_name": null,
-  "trial_params": null
-}

checkpoint-48000/training_args.bin

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:7271f16142689129ff74e2fbdfff484e58a6b4d54b9c292458687c830a3e99ee
-size 3771

checkpoint-48000/optimizer.pt → optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:feb944dd3694845a5a827e8355ddb3eb7119afc806e43fb93c003121f511b0c2
-size 1845322757
+oid sha256:c69d37d0a9056d0549681cf95643d15326e624fec81c58ad2615122f6e8121fc
+size 1845323269

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:033a7f99e720b4906a371665a4a87d95d5707e9f03a48714c113d2e1db26282d
+oid sha256:80b8ede54a14e515a139ae4089b6227b9b07e2b448d36c9a694bd7e26883ec78
 size 922885701

checkpoint-48000/rng_state.pth → rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0229a5f2115e7147f52ccec11492de81225a00cd1c6cf89a19ef407cf5ad3904
-size 14575
+oid sha256:0d7abcde0cae38e8dd2f057b9db1039b4cc83f121a6e5e369f0664f3197e38c3
+size 14639

checkpoint-48000/scaler.pt → scaler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2543dae2fad4aef2bd6aa05e0f4050d32e1386326a137c134e69671f720671c0
+oid sha256:622cef27408eec50002067937ac331e9eb28d60bf926f2fd37184f130260b759
 size 557

checkpoint-48000/scheduler.pt → scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0d5fb0ff60520aa2fb2cdc484ebac963ffe6c641d211ecec8e2ae7d9d5e8dd56
+oid sha256:d841877dae97f3484921bd550ae3afec1339fe3ea6e45d722c6eee90ccd95ea9
 size 627

trainer_state.json

@@ -0,0 +1,176 @@
+{
+  "best_metric": 43.5535,
+  "best_model_checkpoint": "big_wce_weights_1.25_train_0.75/checkpoint-80000",
+  "epoch": 1.9716574245224892,
+  "global_step": 80000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9753635243376465e-05,
+      "loss": 0.1866,
+      "step": 8000
+    },
+    {
+      "epoch": 0.2,
+      "eval_bleu": 42.0127,
+      "eval_gen_len": 33.6165,
+      "eval_loss": 0.09847372770309448,
+      "eval_runtime": 164.5065,
+      "eval_samples_per_second": 6.34,
+      "eval_steps_per_second": 0.401,
+      "step": 8000
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.950723967960567e-05,
+      "loss": 0.1554,
+      "step": 16000
+    },
+    {
+      "epoch": 0.39,
+      "eval_bleu": 42.646,
+      "eval_gen_len": 34.6596,
+      "eval_loss": 0.09764392673969269,
+      "eval_runtime": 134.5362,
+      "eval_samples_per_second": 7.753,
+      "eval_steps_per_second": 0.491,
+      "step": 16000
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9260844115834877e-05,
+      "loss": 0.1511,
+      "step": 24000
+    },
+    {
+      "epoch": 0.59,
+      "eval_bleu": 42.5994,
+      "eval_gen_len": 34.0719,
+      "eval_loss": 0.09760789573192596,
+      "eval_runtime": 181.8789,
+      "eval_samples_per_second": 5.735,
+      "eval_steps_per_second": 0.363,
+      "step": 24000
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.90145101663586e-05,
+      "loss": 0.147,
+      "step": 32000
+    },
+    {
+      "epoch": 0.79,
+      "eval_bleu": 42.7909,
+      "eval_gen_len": 33.908,
+      "eval_loss": 0.09649083018302917,
+      "eval_runtime": 134.5535,
+      "eval_samples_per_second": 7.752,
+      "eval_steps_per_second": 0.491,
+      "step": 32000
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.876814540973506e-05,
+      "loss": 0.1455,
+      "step": 40000
+    },
+    {
+      "epoch": 0.99,
+      "eval_bleu": 42.558,
+      "eval_gen_len": 34.4334,
+      "eval_loss": 0.09574901312589645,
+      "eval_runtime": 126.7021,
+      "eval_samples_per_second": 8.232,
+      "eval_steps_per_second": 0.521,
+      "step": 40000
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.8521749845964266e-05,
+      "loss": 0.1297,
+      "step": 48000
+    },
+    {
+      "epoch": 1.18,
+      "eval_bleu": 42.9867,
+      "eval_gen_len": 34.3883,
+      "eval_loss": 0.09689216315746307,
+      "eval_runtime": 112.9116,
+      "eval_samples_per_second": 9.237,
+      "eval_steps_per_second": 0.585,
+      "step": 48000
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.827538508934073e-05,
+      "loss": 0.13,
+      "step": 56000
+    },
+    {
+      "epoch": 1.38,
+      "eval_bleu": 42.8217,
+      "eval_gen_len": 33.4756,
+      "eval_loss": 0.09708236902952194,
+      "eval_runtime": 102.1103,
+      "eval_samples_per_second": 10.214,
+      "eval_steps_per_second": 0.646,
+      "step": 56000
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 1.802905113986445e-05,
+      "loss": 0.1306,
+      "step": 64000
+    },
+    {
+      "epoch": 1.58,
+      "eval_bleu": 42.8216,
+      "eval_gen_len": 33.9636,
+      "eval_loss": 0.09652863442897797,
+      "eval_runtime": 110.6087,
+      "eval_samples_per_second": 9.43,
+      "eval_steps_per_second": 0.597,
+      "step": 64000
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 1.778268638324091e-05,
+      "loss": 0.1301,
+      "step": 72000
+    },
+    {
+      "epoch": 1.77,
+      "eval_bleu": 42.4795,
+      "eval_gen_len": 35.419,
+      "eval_loss": 0.09661434590816498,
+      "eval_runtime": 132.5432,
+      "eval_samples_per_second": 7.869,
+      "eval_steps_per_second": 0.498,
+      "step": 72000
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.7536321626617377e-05,
+      "loss": 0.1298,
+      "step": 80000
+    },
+    {
+      "epoch": 1.97,
+      "eval_bleu": 43.5535,
+      "eval_gen_len": 33.8552,
+      "eval_loss": 0.09589231759309769,
+      "eval_runtime": 104.1898,
+      "eval_samples_per_second": 10.011,
+      "eval_steps_per_second": 0.633,
+      "step": 80000
+    }
+  ],
+  "max_steps": 649200,
+  "num_train_epochs": 16,
+  "total_flos": 2.5844916958632346e+17,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7271f16142689129ff74e2fbdfff484e58a6b4d54b9c292458687c830a3e99ee
+oid sha256:60b57e3f0ce3836c3e1aabeea3aed565e990fa304449c386084acd628d7e88dd
 size 3771