.circleci/config.yml

@@ -6,7 +6,7 @@ orbs:
 jobs:
   test:
     docker:
-      - image: python:3.7
+      - image: python:3.8
     steps:
       - checkout
       - restore_cache:

README.md

@@ -1,3 +1,5 @@
+🚨**Update**🚨: As of 2021/10/17, the `jiant` project is no longer being actively maintained. This means there will be no plans to add new models, tasks, or features, or update support to new libraries.
+
 <div align="center">
 
 # `jiant` is an NLP toolkit

jiant/proj/main/components/container_setup.py

@@ -76,7 +76,11 @@ def create_task_dict(task_config_dict: dict, verbose: bool = True) -> Dict[str,
         if not task.name == task_name:
             warnings.warn(
                 "task {} from {} has conflicting names: {}/{}. Using {}".format(
-                    task_name, task_config_path, task_name, task.name, task_name,
+                    task_name,
+                    task_config_path,
+                    task_name,
+                    task.name,
+                    task_name,
                 )
             )
             task.name = task_name
@@ -205,7 +209,8 @@ def create_jiant_task_container(
     task_run_config = TaskRunConfig.from_dict(task_run_config)
 
     num_train_examples_dict = get_num_train_examples(
-        task_cache_dict=task_cache_dict, train_task_list=task_run_config.train_task_list,
+        task_cache_dict=task_cache_dict,
+        train_task_list=task_run_config.train_task_list,
     )
     task_sampler = jiant_task_sampler.create_task_sampler(
         sampler_config=sampler_config,

jiant/proj/main/components/evaluate.py

@@ -10,7 +10,8 @@
 def write_val_results(val_results_dict, metrics_aggregator, output_dir, verbose=True):
     full_results_to_write = {
         "aggregated": jiant_task_sampler.compute_aggregate_major_metrics_from_results_dict(
-            metrics_aggregator=metrics_aggregator, results_dict=val_results_dict,
+            metrics_aggregator=metrics_aggregator,
+            results_dict=val_results_dict,
         ),
     }
     for task_name, task_results in val_results_dict.items():

jiant/proj/main/components/task_sampler.py

@@ -137,7 +137,8 @@ def get_task_p(self, steps=None) -> np.ndarray:
 
         for i, task_name in enumerate(self.task_names):
             p_ls[i] = numexpr.evaluate(
-                self.task_to_unnormalized_prob_funcs_dict[task_name], local_dict={"t": t},
+                self.task_to_unnormalized_prob_funcs_dict[task_name],
+                local_dict={"t": t},
             )
         p_ls /= p_ls.sum()
         return p_ls
@@ -171,7 +172,9 @@ def create_task_sampler(
     elif sampler_type == "ProportionalMultiTaskSampler":
         assert len(sampler_config) == 1
         return ProportionalMultiTaskSampler(
-            task_dict=task_dict, rng=rng, task_to_num_examples_dict=task_to_num_examples_dict,
+            task_dict=task_dict,
+            rng=rng,
+            task_to_num_examples_dict=task_to_num_examples_dict,
         )
     elif sampler_type == "SpecifiedProbMultiTaskSampler":
         assert len(sampler_config) == 2

jiant/proj/main/components/write_configs.py

@@ -20,7 +20,8 @@ def write_configs(config_dict, base_path):
             assert os.path.exists(path)
     for config_key in config_keys:
         py_io.write_json(
-            config_dict[config_key], os.path.join(base_path, f"{config_key}.json"),
+            config_dict[config_key],
+            os.path.join(base_path, f"{config_key}.json"),
         )
     py_io.write_json(config_dict, os.path.join(base_path, "full.json"))
     py_io.write_json(

jiant/proj/main/export_model.py

@@ -14,7 +14,8 @@ class RunConfiguration(zconf.RunConfig):
 
 
 def export_model(
-    hf_pretrained_model_name_or_path: str, output_base_path: str,
+    hf_pretrained_model_name_or_path: str,
+    output_base_path: str,
 ):
     """Retrieve model and tokenizer from Transformers and save all necessary data
     Things saved:

jiant/proj/main/metarunner.py

@@ -25,7 +25,9 @@ class ValState(ExtendedDataClassMixin):
     def new(self):
         # noinspection PyArgumentList
         return self.__class__(
-            score=self.score, metrics=self.metrics, train_state=self.train_state.new(),
+            score=self.score,
+            metrics=self.metrics,
+            train_state=self.train_state.new(),
         )
 
     def to_dict(self):
@@ -104,7 +106,8 @@ def yield_train_step(self):
             train_iterator = self.runner.run_train_context(verbose=self.verbose)
         else:
             train_iterator = self.runner.resume_train_context(
-                train_state=self.train_state, verbose=self.verbose,
+                train_state=self.train_state,
+                verbose=self.verbose,
             )
         for train_state in train_iterator:
             self.train_state = train_state
@@ -242,7 +245,8 @@ def eval_save(self):
             self.log_writer.write_entry("train_val_best", self.best_val_state.to_dict())
             del self.best_state_dict
             self.best_state_dict = copy_state_dict(
-                state_dict=get_model_for_saving(self.model).state_dict(), target_device=CPU_DEVICE,
+                state_dict=get_model_for_saving(self.model).state_dict(),
+                target_device=CPU_DEVICE,
             )
             if self.save_best_model:
                 self.save_best_model_with_metadata(val_metrics_dict=val_metrics_dict)

jiant/proj/main/modeling/model_setup.py

@@ -105,14 +105,18 @@ def delegate_load(jiant_model, weights_dict: dict, load_mode: str):
     """
     if load_mode == "from_transformers":
         return load_encoder_from_transformers_weights(
-            encoder=jiant_model.encoder, weights_dict=weights_dict,
+            encoder=jiant_model.encoder,
+            weights_dict=weights_dict,
         )
     elif load_mode == "from_transformers_with_mlm":
         remainder = load_encoder_from_transformers_weights(
-            encoder=jiant_model.encoder, weights_dict=weights_dict, return_remainder=True,
+            encoder=jiant_model.encoder,
+            weights_dict=weights_dict,
+            return_remainder=True,
         )
         load_lm_heads_from_transformers_weights(
-            jiant_model=jiant_model, weights_dict=remainder,
+            jiant_model=jiant_model,
+            weights_dict=remainder,
         )
         return
     elif load_mode == "all":
@@ -121,11 +125,15 @@ def delegate_load(jiant_model, weights_dict: dict, load_mode: str):
         return load_encoder_only(jiant_model=jiant_model, weights_dict=weights_dict)
     elif load_mode == "partial_weights":
         return load_partial_heads(
-            jiant_model=jiant_model, weights_dict=weights_dict, allow_missing_head_weights=True,
+            jiant_model=jiant_model,
+            weights_dict=weights_dict,
+            allow_missing_head_weights=True,
         )
     elif load_mode == "partial_heads":
         return load_partial_heads(
-            jiant_model=jiant_model, weights_dict=weights_dict, allow_missing_head_model=True,
+            jiant_model=jiant_model,
+            weights_dict=weights_dict,
+            allow_missing_head_model=True,
         )
     elif load_mode == "partial":
         return load_partial_heads(
@@ -157,7 +165,7 @@ def load_encoder_from_transformers_weights(
     remainder_weights_dict = {}
     load_weights_dict = {}
     model_arch = ModelArchitectures.from_model_type(model_type=encoder.config.model_type)
-    encoder_prefix = model_arch.value + "."
+    encoder_prefix = model_arch.get_encoder_prefix() + "."
     # Encoder
     for k, v in weights_dict.items():
         if k.startswith(encoder_prefix):

jiant/proj/main/modeling/primary.py

@@ -75,7 +75,9 @@ def forward(self, batch: BatchMixin, task: Task, compute_loss: bool = False):
         taskmodel_key = self.task_to_taskmodel_map[task_name]
         taskmodel = self.taskmodels_dict[taskmodel_key]
         return taskmodel(
-            batch=batch, tokenizer=self.tokenizer, compute_loss=compute_loss,
+            batch=batch,
+            tokenizer=self.tokenizer,
+            compute_loss=compute_loss,
         ).to_dict()
 
 
@@ -105,7 +107,9 @@ def wrap_jiant_forward(
     is_multi_gpu = isinstance(jiant_model, nn.DataParallel)
     model_output = construct_output_from_dict(
         jiant_model(
-            batch=batch.to_dict() if is_multi_gpu else batch, task=task, compute_loss=compute_loss,
+            batch=batch.to_dict() if is_multi_gpu else batch,
+            task=task,
+            compute_loss=compute_loss,
         )
     )
     if is_multi_gpu and compute_loss:

jiant/proj/main/modeling/taskmodels.py

@@ -79,12 +79,17 @@ def __init__(self, task, encoder, head: heads.ClassificationHead, **kwargs):
 
     def forward(self, batch, tokenizer, compute_loss: bool = False):
         encoder_output = self.encoder.encode(
-            input_ids=batch.input_ids, segment_ids=batch.segment_ids, input_mask=batch.input_mask,
+            input_ids=batch.input_ids,
+            segment_ids=batch.segment_ids,
+            input_mask=batch.input_mask,
         )
         logits = self.head(pooled=encoder_output.pooled)
         if compute_loss:
             loss_fct = nn.CrossEntropyLoss()
-            loss = loss_fct(logits.view(-1, self.head.num_labels), batch.label_id.view(-1),)
+            loss = loss_fct(
+                logits.view(-1, self.head.num_labels),
+                batch.label_id.view(-1),
+            )
             return LogitsAndLossOutput(logits=logits, loss=loss, other=encoder_output.other)
         else:
             return LogitsOutput(logits=logits, other=encoder_output.other)
@@ -97,7 +102,9 @@ def __init__(self, task, encoder, head: heads.RegressionHead, **kwargs):
 
     def forward(self, batch, tokenizer, compute_loss: bool = False):
         encoder_output = self.encoder.encode(
-            input_ids=batch.input_ids, segment_ids=batch.segment_ids, input_mask=batch.input_mask,
+            input_ids=batch.input_ids,
+            segment_ids=batch.segment_ids,
+            input_mask=batch.input_mask,
         )
         # TODO: Abuse of notation - these aren't really logits  (issue #1187)
         logits = self.head(pooled=encoder_output.pooled)
@@ -133,7 +140,10 @@ def forward(self, batch, tokenizer, compute_loss: bool = False):
             for j in range(len(encoder_output_other_ls[0])):
                 reshaped_outputs.append(
                     [
-                        torch.stack([misc[j][layer_i] for misc in encoder_output_other_ls], dim=1,)
+                        torch.stack(
+                            [misc[j][layer_i] for misc in encoder_output_other_ls],
+                            dim=1,
+                        )
                         for layer_i in range(len(encoder_output_other_ls[0][0]))
                     ]
                 )
@@ -168,12 +178,17 @@ def forward(self, batch, tokenizer, compute_loss: bool = False):
             TYPE: Description
         """
         encoder_output = self.encoder.encode(
-            input_ids=batch.input_ids, segment_ids=batch.segment_ids, input_mask=batch.input_mask,
+            input_ids=batch.input_ids,
+            segment_ids=batch.segment_ids,
+            input_mask=batch.input_mask,
         )
         logits = self.head(unpooled=encoder_output.unpooled, spans=batch.spans)
         if compute_loss:
             loss_fct = nn.CrossEntropyLoss()
-            loss = loss_fct(logits.view(-1, self.head.num_labels), batch.label_id.view(-1),)
+            loss = loss_fct(
+                logits.view(-1, self.head.num_labels),
+                batch.label_id.view(-1),
+            )
             return LogitsAndLossOutput(logits=logits, loss=loss, other=encoder_output.other)
         else:
             return LogitsOutput(logits=logits, other=encoder_output.other)
@@ -190,7 +205,9 @@ def __init__(self, task, encoder, head: heads.TokenClassificationHead, **kwargs)
 
     def forward(self, batch, tokenizer, compute_loss: bool = False):
         encoder_output = self.encoder.encode(
-            input_ids=batch.input_ids, segment_ids=batch.segment_ids, input_mask=batch.input_mask,
+            input_ids=batch.input_ids,
+            segment_ids=batch.segment_ids,
+            input_mask=batch.input_mask,
         )
         logits = self.head(unpooled=encoder_output.unpooled)
         # Ensure logits in valid range is at least self.offset_margin higher than others
@@ -199,7 +216,8 @@ def forward(self, batch, tokenizer, compute_loss: bool = False):
         if compute_loss:
             loss_fct = nn.CrossEntropyLoss()
             loss = loss_fct(
-                logits.transpose(dim0=1, dim1=2).flatten(end_dim=1), batch.gt_span_idxs.flatten(),
+                logits.transpose(dim0=1, dim1=2).flatten(end_dim=1),
+                batch.gt_span_idxs.flatten(),
             )
             return LogitsAndLossOutput(logits=logits, loss=loss, other=encoder_output.other)
         else:
@@ -213,12 +231,17 @@ def __init__(self, task, encoder, head: heads.SpanComparisonHead, **kwargs):
 
     def forward(self, batch, tokenizer, compute_loss: bool = False):
         encoder_output = self.encoder.encode(
-            input_ids=batch.input_ids, segment_ids=batch.segment_ids, input_mask=batch.input_mask,
+            input_ids=batch.input_ids,
+            segment_ids=batch.segment_ids,
+            input_mask=batch.input_mask,
         )
         logits = self.head(unpooled=encoder_output.unpooled, spans=batch.spans)
         if compute_loss:
             loss_fct = nn.BCEWithLogitsLoss()
-            loss = loss_fct(logits.view(-1, self.head.num_labels), batch.label_ids.float(),)
+            loss = loss_fct(
+                logits.view(-1, self.head.num_labels),
+                batch.label_ids.float(),
+            )
             return LogitsAndLossOutput(logits=logits, loss=loss, other=encoder_output.other)
         else:
             return LogitsOutput(logits=logits, other=encoder_output.other)
@@ -233,7 +256,9 @@ def __init__(self, task, encoder, head: heads.TokenClassificationHead, **kwargs)
 
     def forward(self, batch, tokenizer, compute_loss: bool = False):
         encoder_output = self.encoder.encode(
-            input_ids=batch.input_ids, segment_ids=batch.segment_ids, input_mask=batch.input_mask,
+            input_ids=batch.input_ids,
+            segment_ids=batch.segment_ids,
+            input_mask=batch.input_mask,
         )
         logits = self.head(unpooled=encoder_output.unpooled)
         if compute_loss:
@@ -254,7 +279,9 @@ def __init__(self, task, encoder, head: heads.QAHead, **kwargs):
 
     def forward(self, batch, tokenizer, compute_loss: bool = False):
         encoder_output = self.encoder.encode(
-            input_ids=batch.input_ids, segment_ids=batch.segment_ids, input_mask=batch.input_mask,
+            input_ids=batch.input_ids,
+            segment_ids=batch.segment_ids,
+            input_mask=batch.input_mask,
         )
         logits = self.head(unpooled=encoder_output.unpooled)
         if compute_loss:
@@ -300,7 +327,9 @@ def __init__(self, task, encoder, head: heads.AbstractPoolerHead, **kwargs):
 
     def forward(self, batch, tokenizer, compute_loss: bool = False):
         encoder_output = self.encoder.encode(
-            input_ids=batch.input_ids, segment_ids=batch.segment_ids, input_mask=batch.input_mask,
+            input_ids=batch.input_ids,
+            segment_ids=batch.segment_ids,
+            input_mask=batch.input_mask,
         )
 
         # A tuple of layers of hidden states

jiant/proj/main/preprocessing.py

@@ -51,7 +51,9 @@ def smart_truncate(dataset: torch_utils.ListDataset, max_seq_length: int, verbos
     for datum in maybe_tqdm(dataset.data, desc="Smart truncate data", verbose=verbose):
         new_datum_ls.append(
             smart_truncate_datum(
-                datum=datum, max_seq_length=max_seq_length, max_valid_length=max_valid_length,
+                datum=datum,
+                max_seq_length=max_seq_length,
+                max_valid_length=max_valid_length,
             )
         )
     new_dataset = torch_utils.ListDataset(new_datum_ls)
@@ -70,7 +72,9 @@ def smart_truncate_cache(
         for datum in maybe_tqdm(chunk, desc="Smart truncate chunk-datum", verbose=verbose):
             new_chunk.append(
                 smart_truncate_datum(
-                    datum=datum, max_seq_length=max_seq_length, max_valid_length=max_valid_length,
+                    datum=datum,
+                    max_seq_length=max_seq_length,
+                    max_valid_length=max_valid_length,
                 )
             )
         torch.save(new_chunk, cache.get_chunk_path(chunk_i))