Re. fine-tune-llms-in-2024-with-trl.ipynb

[andysingal]

from datasets import load_dataset 
from random import randint


# Load our test dataset
eval_dataset = load_dataset("json", data_files="test_dataset.json", split="train")
rand_idx = randint(0, len(eval_dataset))

# Test on sample 
prompt = pipe.tokenizer.apply_chat_template(eval_dataset[rand_idx]["messages"][:2], tokenize=False, add_generation_prompt=True)
outputs = pipe(prompt, max_new_tokens=256, do_sample=False, temperature=0.1, top_k=50, top_p=0.1, eos_token_id=pipe.tokenizer.eos_token_id, pad_token_id=pipe.tokenizer.pad_token_id)

print(f"Query:\n{eval_dataset[rand_idx]['messages'][1]['content']}")
print(f"Original Answer:\n{eval_dataset[rand_idx]['messages'][2]['content']}")
print(f"Generated Answer:\n{outputs[0]['generated_text'][len(prompt):].strip()}")

gave error:

---------------------------------------------------------------------------
NotImplementedError                       Traceback (most recent call last)
Cell In[14], line 11
      9 # Test on sample 
     10 prompt = pipe.tokenizer.apply_chat_template(eval_dataset[rand_idx]["messages"][:2], tokenize=False, add_generation_prompt=True)
---> 11 outputs = pipe(prompt, max_new_tokens=256, do_sample=False, temperature=0.1, top_k=50, top_p=0.1, eos_token_id=pipe.tokenizer.eos_token_id, pad_token_id=pipe.tokenizer.pad_token_id)
     13 print(f"Query:\n{eval_dataset[rand_idx]['messages'][1]['content']}")
     14 print(f"Original Answer:\n{eval_dataset[rand_idx]['messages'][2]['content']}")

File /usr/local/lib/python3.10/dist-packages/transformers/pipelines/text_generation.py:208, in TextGenerationPipeline.__call__(self, text_inputs, **kwargs)
    167 def __call__(self, text_inputs, **kwargs):
    168     """
    169     Complete the prompt(s) given as inputs.
    170 
   (...)
    206           ids of the generated text.
    207     """
--> 208     return super().__call__(text_inputs, **kwargs)

File /usr/local/lib/python3.10/dist-packages/transformers/pipelines/base.py:1140, in Pipeline.__call__(self, inputs, num_workers, batch_size, *args, **kwargs)
   1132     return next(
   1133         iter(
   1134             self.get_iterator(
   (...)
   1137         )
   1138     )
   1139 else:
-> 1140     return self.run_single(inputs, preprocess_params, forward_params, postprocess_params)

File /usr/local/lib/python3.10/dist-packages/transformers/pipelines/base.py:1147, in Pipeline.run_single(self, inputs, preprocess_params, forward_params, postprocess_params)
   1145 def run_single(self, inputs, preprocess_params, forward_params, postprocess_params):
   1146     model_inputs = self.preprocess(inputs, **preprocess_params)
-> 1147     model_outputs = self.forward(model_inputs, **forward_params)
   1148     outputs = self.postprocess(model_outputs, **postprocess_params)
   1149     return outputs

File /usr/local/lib/python3.10/dist-packages/transformers/pipelines/base.py:1046, in Pipeline.forward(self, model_inputs, **forward_params)
   1044     with inference_context():
   1045         model_inputs = self._ensure_tensor_on_device(model_inputs, device=self.device)
-> 1046         model_outputs = self._forward(model_inputs, **forward_params)
   1047         model_outputs = self._ensure_tensor_on_device(model_outputs, device=torch.device("cpu"))
   1048 else:

File /usr/local/lib/python3.10/dist-packages/transformers/pipelines/text_generation.py:271, in TextGenerationPipeline._forward(self, model_inputs, **generate_kwargs)
    268         generate_kwargs["min_length"] += prefix_length
    270 # BS x SL
--> 271 generated_sequence = self.model.generate(input_ids=input_ids, attention_mask=attention_mask, **generate_kwargs)
    272 out_b = generated_sequence.shape[0]
    273 if self.framework == "pt":

File /usr/local/lib/python3.10/dist-packages/peft/peft_model.py:1140, in PeftModelForCausalLM.generate(self, **kwargs)
   1138     self.base_model.generation_config = self.generation_config
   1139 try:
-> 1140     outputs = self.base_model.generate(**kwargs)
   1141 except:
   1142     self.base_model.prepare_inputs_for_generation = self.base_model_prepare_inputs_for_generation

File /usr/local/lib/python3.10/dist-packages/torch/utils/_contextlib.py:115, in context_decorator.<locals>.decorate_context(*args, **kwargs)
    112 @functools.wraps(func)
    113 def decorate_context(*args, **kwargs):
    114     with ctx_factory():
--> 115         return func(*args, **kwargs)

File /usr/local/lib/python3.10/dist-packages/transformers/generation/utils.py:1718, in GenerationMixin.generate(self, inputs, generation_config, logits_processor, stopping_criteria, prefix_allowed_tokens_fn, synced_gpus, assistant_model, streamer, negative_prompt_ids, negative_prompt_attention_mask, **kwargs)
   1701     return self.assisted_decoding(
   1702         input_ids,
   1703         assistant_model=assistant_model,
   (...)
   1714         **model_kwargs,
   1715     )
   1716 if generation_mode == GenerationMode.GREEDY_SEARCH:
   1717     # 11. run greedy search
-> 1718     return self.greedy_search(
   1719         input_ids,
   1720         logits_processor=logits_processor,
   1721         stopping_criteria=stopping_criteria,
   1722         pad_token_id=generation_config.pad_token_id,
   1723         eos_token_id=generation_config.eos_token_id,
   1724         output_scores=generation_config.output_scores,
   1725         return_dict_in_generate=generation_config.return_dict_in_generate,
   1726         synced_gpus=synced_gpus,
   1727         streamer=streamer,
   1728         **model_kwargs,
   1729     )
   1731 elif generation_mode == GenerationMode.CONTRASTIVE_SEARCH:
   1732     if not model_kwargs["use_cache"]:

File /usr/local/lib/python3.10/dist-packages/transformers/generation/utils.py:2579, in GenerationMixin.greedy_search(self, input_ids, logits_processor, stopping_criteria, max_length, pad_token_id, eos_token_id, output_attentions, output_hidden_states, output_scores, return_dict_in_generate, synced_gpus, streamer, **model_kwargs)
   2576 model_inputs = self.prepare_inputs_for_generation(input_ids, **model_kwargs)
   2578 # forward pass to get next token
-> 2579 outputs = self(
   2580     **model_inputs,
   2581     return_dict=True,
   2582     output_attentions=output_attentions,
   2583     output_hidden_states=output_hidden_states,
   2584 )
   2586 if synced_gpus and this_peer_finished:
   2587     continue  # don't waste resources running the code we don't need

File /usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py:1518, in Module._wrapped_call_impl(self, *args, **kwargs)
   1516     return self._compiled_call_impl(*args, **kwargs)  # type: ignore[misc]
   1517 else:
-> 1518     return self._call_impl(*args, **kwargs)

File /usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py:1527, in Module._call_impl(self, *args, **kwargs)
   1522 # If we don't have any hooks, we want to skip the rest of the logic in
   1523 # this function, and just call forward.
   1524 if not (self._backward_hooks or self._backward_pre_hooks or self._forward_hooks or self._forward_pre_hooks
   1525         or _global_backward_pre_hooks or _global_backward_hooks
   1526         or _global_forward_hooks or _global_forward_pre_hooks):
-> 1527     return forward_call(*args, **kwargs)
   1529 try:
   1530     result = None

File /usr/local/lib/python3.10/dist-packages/accelerate/hooks.py:165, in add_hook_to_module.<locals>.new_forward(module, *args, **kwargs)
    163         output = module._old_forward(*args, **kwargs)
    164 else:
--> 165     output = module._old_forward(*args, **kwargs)
    166 return module._hf_hook.post_forward(module, output)

File /usr/local/lib/python3.10/dist-packages/transformers/models/llama/modeling_llama.py:1199, in LlamaForCausalLM.forward(self, input_ids, attention_mask, position_ids, past_key_values, inputs_embeds, labels, use_cache, output_attentions, output_hidden_states, return_dict)
   1197     logits = torch.cat(logits, dim=-1)
   1198 else:
-> 1199     logits = self.lm_head(hidden_states)
   1200 logits = logits.float()
   1202 loss = None

File /usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py:1518, in Module._wrapped_call_impl(self, *args, **kwargs)
   1516     return self._compiled_call_impl(*args, **kwargs)  # type: ignore[misc]
   1517 else:
-> 1518     return self._call_impl(*args, **kwargs)

File /usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py:1527, in Module._call_impl(self, *args, **kwargs)
   1522 # If we don't have any hooks, we want to skip the rest of the logic in
   1523 # this function, and just call forward.
   1524 if not (self._backward_hooks or self._backward_pre_hooks or self._forward_hooks or self._forward_pre_hooks
   1525         or _global_backward_pre_hooks or _global_backward_hooks
   1526         or _global_forward_hooks or _global_forward_pre_hooks):
-> 1527     return forward_call(*args, **kwargs)
   1529 try:
   1530     result = None

File /usr/local/lib/python3.10/dist-packages/accelerate/hooks.py:160, in add_hook_to_module.<locals>.new_forward(module, *args, **kwargs)
    159 def new_forward(module, *args, **kwargs):
--> 160     args, kwargs = module._hf_hook.pre_forward(module, *args, **kwargs)
    161     if module._hf_hook.no_grad:
    162         with torch.no_grad():

File /usr/local/lib/python3.10/dist-packages/accelerate/hooks.py:293, in AlignDevicesHook.pre_forward(self, module, *args, **kwargs)
    291             if self.weights_map[name].dtype == torch.int8:
    292                 fp16_statistics = self.weights_map[name.replace("weight", "SCB")]
--> 293         set_module_tensor_to_device(
    294             module, name, self.execution_device, value=self.weights_map[name], fp16_statistics=fp16_statistics
    295         )
    297 return send_to_device(args, self.execution_device), send_to_device(
    298     kwargs, self.execution_device, skip_keys=self.skip_keys
    299 )

File /usr/local/lib/python3.10/dist-packages/accelerate/utils/modeling.py:347, in set_module_tensor_to_device(module, tensor_name, device, value, dtype, fp16_statistics)
    345             module._parameters[tensor_name] = param_cls(new_value, requires_grad=old_value.requires_grad)
    346 elif isinstance(value, torch.Tensor):
--> 347     new_value = value.to(device)
    348 else:
    349     new_value = torch.tensor(value, device=device)

NotImplementedError: Cannot copy out of meta tensor; no data!

[philschmid]

Try again with restarting the Kernel it seems you GPU is already busy