rdiehlmartinez commited on
Commit
a378726
·
1 Parent(s): 6c5fe27

enabling gradient information to be saved out correctly

Browse files
pythia_training_metrics.py → pythia-training-metrics.py RENAMED
@@ -120,19 +120,21 @@ class PythiaTrainingMetrics(datasets.GeneratorBasedBuilder):
120
  elif self.config_name == "gradients_mini":
121
  for gradient_step in get_gradient_step(checkpoint_step)[:2]:
122
  model_size_to_fp[model_size].append(f"{directory_path}/checkpoint_gradients_mini_{gradient_step}.pickle")
 
 
123
 
124
  downloaded_files = dl_manager.download_and_extract(model_size_to_fp)
125
 
126
  return [
127
  datasets.SplitGenerator(
128
- name=datasets.Split.TRAIN,
129
  gen_kwargs={
130
  "filepaths": downloaded_fps
131
  }
132
- ) for downloaded_fps in downloaded_files.values()
133
  ]
134
 
135
- def _generate_examples(self, filepaths):
136
 
137
  # the filepaths should be a list of filepaths
138
  if isinstance(filepaths, str):
@@ -149,11 +151,12 @@ class PythiaTrainingMetrics(datasets.GeneratorBasedBuilder):
149
 
150
  if self.config.name in ["activations", "weights"]:
151
  for layer_name, layer_data in data.items():
152
- for data in layer_data:
153
- yield global_idx, {"checkpoint_step": checkpoint_step, "layer_name": layer_name, "data": data}
154
- global_idx += 1
155
  elif self.config.name in ["gradients", "gradients_mini"]:
 
 
 
156
  for layer_name, layer_data in data.items():
157
- for gradient_step, gradient in layer_data.items():
158
- yield global_idx, {"checkpoint_step": checkpoint_step, "layer_name": layer_name, "gradient_step": gradient_step, "gradient": gradient}
159
- global_idx += 1
 
120
  elif self.config_name == "gradients_mini":
121
  for gradient_step in get_gradient_step(checkpoint_step)[:2]:
122
  model_size_to_fp[model_size].append(f"{directory_path}/checkpoint_gradients_mini_{gradient_step}.pickle")
123
+ else:
124
+ raise Exception("Invalid config name")
125
 
126
  downloaded_files = dl_manager.download_and_extract(model_size_to_fp)
127
 
128
  return [
129
  datasets.SplitGenerator(
130
+ name=model_size_name,
131
  gen_kwargs={
132
  "filepaths": downloaded_fps
133
  }
134
+ ) for model_size_name, downloaded_fps in downloaded_files.items()
135
  ]
136
 
137
+ def _generate_examples(self, filepaths, **kwargs):
138
 
139
  # the filepaths should be a list of filepaths
140
  if isinstance(filepaths, str):
 
151
 
152
  if self.config.name in ["activations", "weights"]:
153
  for layer_name, layer_data in data.items():
154
+ yield global_idx, {"checkpoint_step": checkpoint_step, "layer_name": layer_name, "data": data}
155
+ global_idx += 1
 
156
  elif self.config.name in ["gradients", "gradients_mini"]:
157
+
158
+ gradient_step = int(filepath.split('/')[-1].split("_")[-1].split(".")[0])
159
+
160
  for layer_name, layer_data in data.items():
161
+ yield global_idx, {"checkpoint_step": checkpoint_step, "layer_name": layer_name, "gradient_step": gradient_step, "gradient": layer_data}
162
+ global_idx += 1