RuntimeError: Placeholder storage has not been allocated on MPS device!

[kkckk1110]

---

RuntimeError Traceback (most recent call last) Cell In[22], line 1 ----> 1 zeroshot_trainer.evaluate(test_dataset)

File ~/miniforge3/envs/TTM/lib/python3.9/site-packages/transformers/trainer.py:3641, in Trainer.evaluate(self, eval_dataset, ignore_keys, metric_key_prefix) 3638 start_time = time.time() 3640 eval_loop = self.prediction_loop if self.args.use_legacy_prediction_loop else self.evaluation_loop -> 3641 output = eval_loop( 3642 eval_dataloader, 3643 description="Evaluation", 3644 # No point gathering the predictions if there are no metrics, otherwise we defer to 3645 # self.args.prediction_loss_only 3646 prediction_loss_only=True if self.compute_metrics is None else None, 3647 ignore_keys=ignore_keys, 3648 metric_key_prefix=metric_key_prefix, 3649 ) 3651 total_batch_size = self.args.eval_batch_size * self.args.world_size 3652 if f"{metric_key_prefix}_jit_compilation_time" in output.metrics:

File ~/miniforge3/envs/TTM/lib/python3.9/site-packages/transformers/trainer.py:3826, in Trainer.evaluation_loop(self, dataloader, description, prediction_loss_only, ignore_keys, metric_key_prefix) 3823 batch_size = observed_batch_size 3825 # Prediction step -> 3826 losses, logits, labels = self.prediction_step(model, inputs, prediction_loss_only, ignore_keys=ignore_keys) 3827 main_input_name = getattr(self.model, "main_input_name", "input_ids") 3828 inputs_decode = self._prepare_input(inputs[main_input_name]) if args.include_inputs_for_metrics else None

File ~/miniforge3/envs/TTM/lib/python3.9/site-packages/transformers/trainer.py:4040, in Trainer.prediction_step(self, model, inputs, prediction_loss_only, ignore_keys) 4038 if has_labels or loss_without_labels: 4039 with self.compute_loss_context_manager(): -> 4040 loss, outputs = self.compute_loss(model, inputs, return_outputs=True) 4041 loss = loss.mean().detach() 4043 if isinstance(outputs, dict):

File ~/miniforge3/envs/TTM/lib/python3.9/site-packages/transformers/trainer.py:3338, in Trainer.compute_loss(self, model, inputs, return_outputs) 3336 else: 3337 labels = None -> 3338 outputs = model(**inputs) 3339 # Save past state if it exists 3340 # TODO: this needs to be fixed and made cleaner later. 3341 if self.args.past_index >= 0:

File ~/miniforge3/envs/TTM/lib/python3.9/site-packages/torch/nn/modules/module.py:1532, in Module._wrapped_call_impl(self, *args, kwargs) 1530 return self._compiled_call_impl(*args, *kwargs) # type: ignore[misc] 1531 else: -> 1532 return self._call_impl(args, kwargs)

File ~/miniforge3/envs/TTM/lib/python3.9/site-packages/torch/nn/modules/module.py:1541, in Module._call_impl(self, *args, *kwargs) 1536 # If we don't have any hooks, we want to skip the rest of the logic in 1537 # this function, and just call forward. 1538 if not (self._backward_hooks or self._backward_pre_hooks or self._forward_hooks or self._forward_pre_hooks 1539 or _global_backward_pre_hooks or _global_backward_hooks 1540 or _global_forward_hooks or _global_forward_pre_hooks): -> 1541 return forward_call(args, **kwargs) 1543 try: 1544 result = None

File ~/miniforge3/envs/TTM/lib/python3.9/site-packages/tsfm_public/models/tinytimemixer/modeling_tinytimemixer.py:1473, in TinyTimeMixerForPrediction.forward(self, past_values, future_values, observed_mask, output_hidden_states, return_loss, return_dict, freq_token) 1470 return_dict = return_dict if return_dict is not None else self.use_return_dict 1472 # past_values: tensor [batch_size x context_length x num_input_channels] -> 1473 model_output = self.backbone( 1474 past_values, 1475 observed_mask=observed_mask, 1476 output_hidden_states=output_hidden_states, 1477 return_dict=return_dict, 1478 freq_token=freq_token, 1479 ) # model_output: [batch_size x nvars x num_patch x d_model] 1481 if isinstance(model_output, tuple): 1482 model_output = TinyTimeMixerModelOutput(*model_output)

File ~/miniforge3/envs/TTM/lib/python3.9/site-packages/torch/nn/modules/module.py:1532, in Module._wrapped_call_impl(self, *args, kwargs) 1530 return self._compiled_call_impl(*args, *kwargs) # type: ignore[misc] 1531 else: -> 1532 return self._call_impl(args, kwargs)

File ~/miniforge3/envs/TTM/lib/python3.9/site-packages/torch/nn/modules/module.py:1541, in Module._call_impl(self, *args, *kwargs) 1536 # If we don't have any hooks, we want to skip the rest of the logic in 1537 # this function, and just call forward. 1538 if not (self._backward_hooks or self._backward_pre_hooks or self._forward_hooks or self._forward_pre_hooks 1539 or _global_backward_pre_hooks or _global_backward_hooks 1540 or _global_forward_hooks or _global_forward_pre_hooks): -> 1541 return forward_call(args, **kwargs) 1543 try: 1544 result = None

File ~/miniforge3/envs/TTM/lib/python3.9/site-packages/tsfm_public/models/tinytimemixer/modeling_tinytimemixer.py:1279, in TinyTimeMixerModel.forward(self, past_values, observed_mask, output_hidden_states, return_dict, freq_token) 1275 patched_x = self.patching(scaled_past_values) # [batch_size x num_input_channels x num_patch x patch_length 1277 enc_input = patched_x -> 1279 encoder_output = self.encoder( 1280 enc_input, 1281 output_hidden_states=output_hidden_states, 1282 return_dict=return_dict, 1283 freq_token=freq_token, 1284 ) 1286 if isinstance(encoder_output, tuple): 1287 encoder_output = TinyTimeMixerEncoderOutput(*encoder_output)

File ~/miniforge3/envs/TTM/lib/python3.9/site-packages/torch/nn/modules/module.py:1532, in Module._wrapped_call_impl(self, *args, kwargs) 1530 return self._compiled_call_impl(*args, *kwargs) # type: ignore[misc] 1531 else: -> 1532 return self._call_impl(args, kwargs)

File ~/miniforge3/envs/TTM/lib/python3.9/site-packages/torch/nn/modules/module.py:1541, in Module._call_impl(self, *args, *kwargs) 1536 # If we don't have any hooks, we want to skip the rest of the logic in 1537 # this function, and just call forward. 1538 if not (self._backward_hooks or self._backward_pre_hooks or self._forward_hooks or self._forward_pre_hooks 1539 or _global_backward_pre_hooks or _global_backward_hooks 1540 or _global_forward_hooks or _global_forward_pre_hooks): -> 1541 return forward_call(args, **kwargs) 1543 try: 1544 result = None

File ~/miniforge3/envs/TTM/lib/python3.9/site-packages/tsfm_public/models/tinytimemixer/modeling_tinytimemixer.py:1158, in TinyTimeMixerEncoder.forward(self, past_values, output_hidden_states, return_dict, freq_token) 1155 return_dict = return_dict if return_dict is not None else self.use_return_dict 1157 # flatten [bs x num_patch x d_model]. common_channel/mix_channel: [bs x n_vars x num_patch x d_model] -> 1158 patches = self.patcher(past_values) 1160 if self.resolution_prefix_tuning: 1161 if freq_token is not None:

File ~/miniforge3/envs/TTM/lib/python3.9/site-packages/torch/nn/modules/module.py:1532, in Module._wrapped_call_impl(self, *args, kwargs) 1530 return self._compiled_call_impl(*args, *kwargs) # type: ignore[misc] 1531 else: -> 1532 return self._call_impl(args, kwargs)

File ~/miniforge3/envs/TTM/lib/python3.9/site-packages/torch/nn/modules/module.py:1541, in Module._call_impl(self, *args, *kwargs) 1536 # If we don't have any hooks, we want to skip the rest of the logic in 1537 # this function, and just call forward. 1538 if not (self._backward_hooks or self._backward_pre_hooks or self._forward_hooks or self._forward_pre_hooks 1539 or _global_backward_pre_hooks or _global_backward_hooks 1540 or _global_forward_hooks or _global_forward_pre_hooks): -> 1541 return forward_call(args, **kwargs) 1543 try: 1544 result = None

File ~/miniforge3/envs/TTM/lib/python3.9/site-packages/torch/nn/modules/linear.py:116, in Linear.forward(self, input) 115 def forward(self, input: Tensor) -> Tensor: --> 116 return F.linear(input, self.weight, self.bias)

RuntimeError: Placeholder storage has not been allocated on MPS device!

[wgifford]

@kkckk1110 Thanks for the stack trace! Can you provide a little more detail around what notebook / python code you were trying to run?

[kkckk1110]

Of course! The code is zeroshot_trainer.evaluate(test_dataset) which is in the ttm_tutorial.ipynb