Revert "[s2s] --eval_max_generate_length (huggingface#7018)"

This reverts commit 6090cc9.

examples/seq2seq/distil_marian_enro_teacher.sh

@@ -16,5 +16,5 @@ python distillation.py \
   --train_batch_size=$BS --eval_batch_size=$BS \
   --tokenizer_name Helsinki-NLP/opus-mt-en-ro \
   --warmup_steps 500 --logger_name wandb \
-  --fp16_opt_level O1 --task translation --normalize_hidden --num_sanity_val_steps=0 \
+  --fp16_opt_level O1 --task translation --normalize_hidden \
   "$@"

examples/seq2seq/distil_marian_no_teacher.sh

@@ -13,5 +13,5 @@ python distillation.py \
   --train_batch_size=$BS --eval_batch_size=$BS \
   --tokenizer_name $m --model_name_or_path $m \
   --warmup_steps 500 --sortish_sampler --logger_name wandb \
-  --gpus 1 --fp16_opt_level=O1 --task translation --num_sanity_val_steps=0 \
+  --gpus 1 --fp16_opt_level=O1 --task translation \
   "$@"

examples/seq2seq/finetune.py

@@ -11,6 +11,7 @@
 import numpy as np
 import pytorch_lightning as pl
 import torch
+from packaging import version
 from torch.utils.data import DataLoader
 
 from lightning_base import BaseTransformer, add_generic_args, generic_train
@@ -93,9 +94,6 @@ def __init__(self, hparams, **kwargs):
             "val": self.hparams.val_max_target_length,
             "test": self.hparams.test_max_target_length,
         }
-        if self.hparams.sortish_sampler and self.hparams.gpus > 1:
-            self.hparams.sortish_sampler = False
-            warnings.warn("ignoring sortish_sampler as it is unsupported on multiple GPUs")
         assert self.target_lens["train"] <= self.target_lens["val"], f"target_lens: {self.target_lens}"
         assert self.target_lens["train"] <= self.target_lens["test"], f"target_lens: {self.target_lens}"
 
@@ -116,10 +114,6 @@ def __init__(self, hparams, **kwargs):
         )
         self.eval_beams = self.model.config.num_beams if self.hparams.eval_beams is None else self.hparams.eval_beams
         assert self.eval_beams >= 1, f"got self.eval_beams={self.eval_beams}. Need an integer > 1"
-        if self.hparams.eval_max_gen_length is not None:
-            self.eval_max_length = self.hparams.eval_max_gen_length
-        else:
-            self.eval_max_length = self.model.config.max_length
         self.val_metric = self.default_val_metric if self.hparams.val_metric is None else self.hparams.val_metric
 
     def freeze_embeds(self):
@@ -215,15 +209,12 @@ def calc_generative_metrics(self, preds, target) -> Dict:
 
     def _generative_step(self, batch: dict) -> dict:
         t0 = time.time()
-
-        # parser.add_argument('--eval_max_gen_length', type=int, default=None, help='never generate more than n tokens')
         generated_ids = self.model.generate(
             batch["input_ids"],
             attention_mask=batch["attention_mask"],
             use_cache=True,
             decoder_start_token_id=self.decoder_start_token_id,
             num_beams=self.eval_beams,
-            max_length=self.eval_max_length,
         )
         gen_time = (time.time() - t0) / batch["input_ids"].shape[0]
         preds: List[str] = self.ids_to_clean_text(generated_ids)
@@ -257,7 +248,7 @@ def get_dataloader(self, type_path: str, batch_size: int, shuffle: bool = False)
         dataset = self.get_dataset(type_path)
         sampler = None
         if self.hparams.sortish_sampler and type_path == "train":
-            assert self.hparams.gpus <= 1  # this should never break because of the assertion in __init__
+            assert self.hparams.gpus <= 1  # TODO: assert earlier
             sampler = dataset.make_sortish_sampler(batch_size)
             shuffle = False
 
@@ -330,7 +321,6 @@ def add_model_specific_args(parser, root_dir):
         parser.add_argument(
             "--val_metric", type=str, default=None, required=False, choices=["bleu", "rouge2", "loss", None]
         )
-        parser.add_argument("--eval_max_gen_length", type=int, default=None, help="never generate more than n tokens")
         parser.add_argument("--save_top_k", type=int, default=1, required=False, help="How many checkpoints to save")
         parser.add_argument(
             "--early_stopping_patience",
@@ -366,6 +356,8 @@ def main(args, model=None) -> SummarizationModule:
             model: SummarizationModule = SummarizationModule(args)
         else:
             model: SummarizationModule = TranslationModule(args)
+    if version.parse(torch.__version__) == version.parse("1.6") and args.fp16:
+        warnings.warn("FP16 only seems to work with torch 1.5+apex")
     dataset = Path(args.data_dir).name
     if (
         args.logger_name == "default"

examples/seq2seq/test_seq2seq_examples.py

@@ -34,7 +34,6 @@
     "supervise_forward": True,
     "normalize_hidden": True,
     "label_smoothing": 0.2,
-    "eval_max_gen_length": None,
     "eval_beams": 1,
     "val_metric": "loss",
     "save_top_k": 1,