{ "nbformat": 4, "nbformat_minor": 0, "metadata": { "colab": { "name": "Untitled330.ipynb", "provenance": [], "collapsed_sections": [] }, "kernelspec": { "name": "python3", "display_name": "Python 3" }, "language_info": { "name": "python" } }, "cells": [ { "cell_type": "code", "metadata": { "id": "Ii2x731Ta8fu" }, "source": [ "%%capture\n", "!pip install transformers\n", "!pip install datasets\n", "!pip install --upgrade git+https://github.com/google/flax.git" ], "execution_count": 1, "outputs": [] }, { "cell_type": "code", "metadata": { "id": "_9NMPFKua9hr" }, "source": [ "import jax\n", "from transformers.modeling_flax_utils import FlaxPreTrainedModel\n", "import flax.linen as nn\n", "import jax.numpy as jnp\n", "from transformers import GPT2Config\n", "#from transformers import FlaxGPT2PreTrainedModel\n", "from transformers import FlaxGPT2Model\n", "import jax.numpy as jnp\n", "from transformers import GPT2Tokenizer\n", "tokenizer = GPT2Tokenizer.from_pretrained(\"gpt2\",pad_token='<|endoftext|>') \n", "from typing import Any, Optional, Tuple\n", "from flax.core.frozen_dict import FrozenDict, unfreeze\n", "from transformers import file_utils\n", "from transformers.file_utils import add_start_docstrings, add_start_docstrings_to_model_forward\n", "from transformers.models.gpt2.modeling_flax_gpt2 import FlaxGPT2BlockCollection\n", "from transformers.modeling_flax_outputs import FlaxBaseModelOutput" ], "execution_count": null, "outputs": [] }, { "cell_type": "code", "metadata": { "id": "dqkcoBOccszd" }, "source": [ "GPT2_START_DOCSTRING = r\"\"\"\n", " This model inherits from :class:`~transformers.FlaxPreTrainedModel`. Check the superclass documentation for the\n", " generic methods the library implements for all its model (such as downloading or saving, resizing the input\n", " embeddings, pruning heads etc.)\n", " This model is also a Flax Linen `flax.nn.Module\n", " `__ subclass. Use it as a regular Flax\n", " Module and refer to the Flax documentation for all matter related to general usage and behavior.\n", " Finally, this model supports inherent JAX features such as:\n", " - `Just-In-Time (JIT) compilation `__\n", " - `Automatic Differentiation `__\n", " - `Vectorization `__\n", " - `Parallelization `__\n", " Parameters:\n", " config (:class:`~transformers.GPT2Config`): Model configuration class with all the parameters of the model.\n", " Initializing with a config file does not load the weights associated with the model, only the\n", " configuration. Check out the :meth:`~transformers.FlaxPreTrainedModel.from_pretrained` method to load the\n", " model weights.\n", "\"\"\"\n", "\n", "GPT2_INPUTS_DOCSTRING = r\"\"\"\n", " Args:\n", " input_ids (:obj:`numpy.ndarray` of shape :obj:`(batch_size,input_ids_length)`):\n", " :obj:`input_ids_length` = ``sequence_length``. Indices of input sequence tokens in the vocabulary.\n", " Indices can be obtained using :class:`~transformers.GPT2Tokenizer`. See\n", " :meth:`transformers.PreTrainedTokenizer.encode` and :meth:`transformers.PreTrainedTokenizer.__call__` for\n", " details.\n", " `What are input IDs? <../glossary.html#input-ids>`__\n", " attention_mask (:obj:`numpy.ndarray` of shape :obj:`(batch_size, sequence_length)`, `optional`):\n", " Mask to avoid performing attention on padding token indices. Mask values selected in ``[0, 1]``:\n", " - 1 for tokens that are **not masked**,\n", " - 0 for tokens that are **masked**.\n", " `What are attention masks? <../glossary.html#attention-mask>`__\n", " position_ids (:obj:`numpy.ndarray` of shape :obj:`(batch_size, sequence_length)`, `optional`):\n", " Indices of positions of each input sequence tokens in the position embeddings. Selected in the range ``[0,\n", " config.max_position_embeddings - 1]``.\n", " past_key_values (:obj:`Dict[str, np.ndarray]`, `optional`, returned by ``init_cache`` or when passing previous ``past_key_values``):\n", " Dictionary of pre-computed hidden-states (key and values in the attention blocks) that can be used for fast\n", " auto-regressive decoding. Pre-computed key and value hidden-states are of shape `[batch_size, max_length]`.\n", " output_attentions (:obj:`bool`, `optional`):\n", " Whether or not to return the attentions tensors of all attention layers. See ``attentions`` under returned\n", " tensors for more detail.\n", " output_hidden_states (:obj:`bool`, `optional`):\n", " Whether or not to return the hidden states of all layers. See ``hidden_states`` under returned tensors for\n", " more detail.\n", " return_dict (:obj:`bool`, `optional`):\n", " Whether or not to return a :class:`~transformers.file_utils.ModelOutput` instead of a plain tuple.\n", "\"\"\"" ], "execution_count": 3, "outputs": [] }, { "cell_type": "code", "metadata": { "id": "NX-Z5iCMbKL5" }, "source": [ "class FlaxGGGPreTrainedModel(FlaxPreTrainedModel):\n", " \"\"\"\n", " An abstract class to handle weights initialization and a simple interface for downloading and loading pretrained\n", " models.\n", " \"\"\"\n", "\n", " config_class = GPT2Config\n", " base_model_prefix = \"transformer\"\n", " module_class: nn.Module = None\n", "\n", " def __init__(\n", " self,\n", " config: GPT2Config,\n", " input_shape: Tuple = (1,1),\n", " seed: int = 0,\n", " dtype: jnp.dtype = jnp.float32,\n", " **kwargs,\n", " ):\n", " \n", " module = self.module_class(config=config, dtype=dtype, **kwargs)\n", " super().__init__(config, module, input_shape=input_shape, seed=seed, dtype=dtype)\n", "\n", " def init_weights(self, rng: jax.random.PRNGKey, input_shape: Tuple) -> FrozenDict:\n", " # init input tensors\n", " input_ids = jnp.zeros(input_shape, dtype=\"i4\")\n", " attention_mask = jnp.ones_like(input_ids)\n", " \n", " position_ids = jnp.broadcast_to(jnp.arange(jnp.atleast_2d(input_ids).shape[-1]), input_shape)\n", "\n", " params_rng, dropout_rng = jax.random.split(rng)\n", " rngs = {\"params\": params_rng, \"dropout\": dropout_rng}\n", "\n", " return self.module.init(rngs, input_ids, attention_mask, position_ids, return_dict=False)[\"params\"]\n", "\n", " def init_cache(self, batch_size, max_length):\n", " r\"\"\"\n", " Args:\n", " batch_size (:obj:`int`):\n", " batch_size used for fast auto-regressive decoding. Defines the batch size of the initialized cache.\n", " max_length (:obj:`int`):\n", " maximum possible length for auto-regressive decoding. Defines the sequence length of the initialized\n", " cache.\n", " \"\"\"\n", " # init input variables to retrieve cache\n", " input_ids = jnp.ones((batch_size, max_length))\n", " attention_mask = jnp.ones_like(input_ids)\n", " position_ids = jnp.broadcast_to(jnp.arange(jnp.atleast_2d(input_ids).shape[-1]), input_ids.shape)\n", "\n", " init_variables = self.module.init(\n", " jax.random.PRNGKey(0), input_ids, attention_mask, position_ids, return_dict=False, init_cache=True\n", " )\n", " return init_variables[\"cache\"]\n", "\n", " @add_start_docstrings_to_model_forward(GPT2_INPUTS_DOCSTRING)\n", " def __call__(\n", " self,\n", " input_ids,\n", " attention_mask=None,\n", " position_ids=None,\n", " params: dict = None,\n", " past_key_values: dict = None,\n", " dropout_rng: jax.random.PRNGKey = None,\n", " train: bool = False,\n", " output_attentions: Optional[bool] = None,\n", " output_hidden_states: Optional[bool] = None,\n", " return_dict: Optional[bool] = None,\n", " ):\n", " output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions\n", " output_hidden_states = (\n", " output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states\n", " )\n", " return_dict = return_dict if return_dict is not None else self.config.return_dict\n", " print(input_ids.shape)\n", "\n", " # batch_size, num_choices,sequence_length = input_ids.shape\n", "\n", " if position_ids is None:\n", " if past_key_values is not None:\n", " raise ValueError(\"Make sure to provide `position_ids` when passing `past_key_values`.\")\n", " \n", " position_ids=jnp.broadcast_to(jnp.arange(jnp.atleast_2d(input_ids).shape[-1]), input_ids.shape)\n", "\n", " # position_ids = jnp.broadcast_to(jnp.arange(sequence_length)[None, :], (batch_size, sequence_length))\n", "\n", " if attention_mask is None:\n", " attention_mask = jnp.ones((input_ids))\n", " print('attn not')\n", "\n", " # Handle any PRNG if needed\n", " rngs = {}\n", " if dropout_rng is not None:\n", " rngs[\"dropout\"] = dropout_rng\n", "\n", " inputs = {\"params\": params or self.params}\n", "\n", " # if past_key_values are passed then cache is already initialized a private flag init_cache has to be passed down to ensure cache is used. It has to be made sure that cache is marked as mutable so that it can be changed by FlaxGPT2Attention module\n", " if past_key_values:\n", " inputs[\"cache\"] = past_key_values\n", " mutable = [\"cache\"]\n", " else:\n", " mutable = False\n", "\n", " outputs = self.module.apply(\n", " inputs,\n", " jnp.array(input_ids, dtype=\"i4\"),\n", " jnp.array(attention_mask, dtype=\"i4\"),\n", " jnp.array(position_ids, dtype=\"i4\"),\n", " not train,\n", " False,\n", " output_attentions,\n", " output_hidden_states,\n", " return_dict,\n", " rngs=rngs,\n", " mutable=mutable,\n", " )\n", " print('cache')\n", "\n", " # add updated cache to model output\n", " if past_key_values is not None and return_dict:\n", " outputs, past_key_values = outputs\n", " outputs[\"past_key_values\"] = unfreeze(past_key_values[\"cache\"])\n", " return outputs\n", " elif past_key_values is not None and not return_dict:\n", " outputs, past_key_values = outputs\n", " outputs = outputs[:1] + (unfreeze(past_key_values[\"cache\"]),) + outputs[1:]\n", "\n", " return outputs" ], "execution_count": 6, "outputs": [] }, { "cell_type": "code", "metadata": { "id": "4vRAWll2bwQQ" }, "source": [ "class FlaxGGGModule(nn.Module):\n", " config: GPT2Config\n", " dtype: jnp.dtype = jnp.float32\n", "\n", " def setup(self):\n", " self.embed_dim = self.config.hidden_size\n", "\n", " self.wte = nn.Embed(\n", " self.config.vocab_size,\n", " self.embed_dim,\n", " embedding_init=jax.nn.initializers.normal(stddev=self.config.initializer_range),\n", " dtype=self.dtype,\n", " )\n", " self.wpe = nn.Embed(\n", " self.config.max_position_embeddings,\n", " self.embed_dim,\n", " embedding_init=jax.nn.initializers.normal(stddev=self.config.initializer_range),\n", " dtype=self.dtype,\n", " )\n", " self.dropout = nn.Dropout(rate=self.config.embd_pdrop)\n", " self.h = FlaxGPT2BlockCollection(self.config, dtype=self.dtype)\n", " self.ln_f = nn.LayerNorm(epsilon=self.config.layer_norm_epsilon, dtype=self.dtype)\n", "\n", " def __call__(\n", " self,\n", " input_ids,\n", " attention_mask,\n", " position_ids,\n", " deterministic=True,\n", " init_cache: bool = False,\n", " output_attentions: bool = False,\n", " output_hidden_states: bool = False,\n", " return_dict: bool = True,\n", " ):\n", " input_embeds = self.wte(input_ids.astype(\"i4\"))\n", " position_embeds = self.wpe(position_ids.astype(\"i4\"))\n", " \n", "\n", " hidden_states = input_embeds + position_embeds\n", " hidden_states = self.dropout(hidden_states, deterministic=deterministic)\n", " outputs = self.h(\n", " hidden_states,\n", " attention_mask,\n", " deterministic=deterministic,\n", " init_cache=init_cache,\n", " output_attentions=output_attentions,\n", " output_hidden_states=output_hidden_states,\n", " return_dict=return_dict,\n", " )\n", "\n", " hidden_states = outputs[0]\n", " hidden_states = self.ln_f(hidden_states)\n", " print('ggg')\n", " if not return_dict:\n", " return (hidden_states,) + outputs[1:]\n", "\n", " return FlaxBaseModelOutput(\n", " last_hidden_state=hidden_states,\n", " hidden_states=outputs.hidden_states,\n", " attentions=outputs.attentions,)\n", "class FlaxNewModel(FlaxGGGPreTrainedModel):\n", " module_class = FlaxGGGModule" ], "execution_count": 7, "outputs": [] }, { "cell_type": "code", "metadata": { "id": "_ljSn6GdedtI" }, "source": [ "class FlaxGPT2ForMultipleChoiceModule(nn.Module):\n", " config:GPT2Config\n", " dtype: jnp.dtype = jnp.float32\n", " def setup(self):\n", " self.gpt2 = FlaxNewModel(config=self.config, dtype=self.dtype)\n", " self.dropout = nn.Dropout(rate=0.2)\n", " self.classifier = nn.Dense(4, dtype=self.dtype)\n", "\n", " def __call__(self,input_ids,attention_mask,position_ids,return_dict=True,deterministic=True,*args):\n", " batch_size = input_ids.shape[0]\n", " rng=jax.random.PRNGKey(0)\n", " _, dropout_rng = jax.random.split(rng)\n", " print('abc')\n", "\n", " outputs=self.gpt2(input_ids, attention_mask,position_ids,return_dict=return_dict)\n", " \n", "\n", " hidden_states = outputs[0]\n", "\n", " \n", " hidden_states= jnp.mean(hidden_states, axis=1)\n", "\n", " print(hidden_states.shape)\n", " \n", " \n", " hidden_states=hidden_states.reshape(batch_size,-1) #(32,8,768)->(32,8*768)\n", "\n", " dropout_output = self.dropout(hidden_states,deterministic=deterministic,rng=dropout_rng)\n", "\n", " print(dropout_output.shape)\n", " \n", "\n", " logits = self.classifier(dropout_output)\n", " print('bnv')\n", " reshaped_logits = logits.reshape(-1, 4) \n", " #(32,4)\n", " if not return_dict:\n", " return (reshaped_logits,) + outputs[2:]\n", " return reshaped_logits" ], "execution_count": 8, "outputs": [] }, { "cell_type": "code", "metadata": { "id": "M4UPf3Waexq0" }, "source": [ "class FlaxGPT2ForMultipleChoice(FlaxNewModel):\n", " module_class = FlaxGPT2ForMultipleChoiceModule" ], "execution_count": 9, "outputs": [] }, { "cell_type": "code", "metadata": { "id": "roQ3vls4e4TH" }, "source": [ "model = FlaxGPT2ForMultipleChoice.from_pretrained('gpt2')" ], "execution_count": null, "outputs": [] }, { "cell_type": "code", "metadata": { "id": "E9qOSaaie417" }, "source": [ "input_ids=jnp.ones((1,2,11))\n", "attention_mask=jnp.ones((1,2,11))" ], "execution_count": 12, "outputs": [] }, { "cell_type": "code", "metadata": { "colab": { "base_uri": "https://localhost:8080/", "height": 409 }, "id": "am7hYv8auWVy", "outputId": "0c8192ca-a0ab-432e-d483-46f8a2cc2576" }, "source": [ "out1 = model(input_ids, attention_mask)" ], "execution_count": 13, "outputs": [ { "output_type": "stream", "text": [ "(1, 2, 11)\n", "attn not\n", "ggg\n", "abc\n", "(1, 2, 11)\n", "attn not\n" ], "name": "stdout" }, { "output_type": "error", "ename": "ValueError", "evalue": "ignored", "traceback": [ "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m", "\u001b[0;31mValueError\u001b[0m Traceback (most recent call last)", "\u001b[0;32m\u001b[0m in \u001b[0;36m\u001b[0;34m()\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mout1\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mmodel\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0minput_ids\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mattention_mask\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m", "\u001b[0;32m\u001b[0m in \u001b[0;36m__call__\u001b[0;34m(self, input_ids, attention_mask, position_ids, params, past_key_values, dropout_rng, train, output_attentions, output_hidden_states, return_dict)\u001b[0m\n\u001b[1;32m 112\u001b[0m \u001b[0mreturn_dict\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 113\u001b[0m \u001b[0mrngs\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mrngs\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 114\u001b[0;31m \u001b[0mmutable\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mmutable\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 115\u001b[0m )\n\u001b[1;32m 116\u001b[0m \u001b[0mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'cache'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/flax/linen/module.py\u001b[0m in \u001b[0;36mapply\u001b[0;34m(self, variables, rngs, method, mutable, capture_intermediates, *args, **kwargs)\u001b[0m\n\u001b[1;32m 964\u001b[0m \u001b[0mmethod\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 965\u001b[0m \u001b[0mmutable\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mmutable\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mcapture_intermediates\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mcapture_intermediates\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 966\u001b[0;31m )(variables, *args, **kwargs, rngs=rngs)\n\u001b[0m\u001b[1;32m 967\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 968\u001b[0m def init_with_output(self,\n", "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/flax/core/scope.py\u001b[0m in \u001b[0;36mwrapper\u001b[0;34m(variables, rngs, *args, **kwargs)\u001b[0m\n\u001b[1;32m 685\u001b[0m **kwargs) -> Union[Any, Tuple[Any, VariableDict]]:\n\u001b[1;32m 686\u001b[0m \u001b[0;32mwith\u001b[0m \u001b[0mbind\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mvariables\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mrngs\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mrngs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmutable\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mmutable\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtemporary\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0mroot\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 687\u001b[0;31m \u001b[0my\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mfn\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mroot\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 688\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0mmutable\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0;32mFalse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 689\u001b[0m \u001b[0;32mreturn\u001b[0m \u001b[0my\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mroot\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmutable_variables\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/flax/linen/module.py\u001b[0m in \u001b[0;36mscope_fn\u001b[0;34m(scope, *args, **kwargs)\u001b[0m\n\u001b[1;32m 1214\u001b[0m \u001b[0m_context\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcapture_stack\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mappend\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mcapture_intermediates\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 1215\u001b[0m \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1216\u001b[0;31m \u001b[0;32mreturn\u001b[0m \u001b[0mfn\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmodule\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mclone\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mparent\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mscope\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 1217\u001b[0m \u001b[0;32mfinally\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 1218\u001b[0m \u001b[0m_context\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcapture_stack\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mpop\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/flax/linen/module.py\u001b[0m in \u001b[0;36mwrapped_module_method\u001b[0;34m(*args, **kwargs)\u001b[0m\n\u001b[1;32m 281\u001b[0m \u001b[0m_context\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmodule_stack\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mappend\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 282\u001b[0m \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 283\u001b[0;31m \u001b[0my\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mfun\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 284\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0m_context\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcapture_stack\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 285\u001b[0m \u001b[0mfilter_fn\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0m_context\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcapture_stack\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m-\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", "\u001b[0;32m\u001b[0m in \u001b[0;36m__call__\u001b[0;34m(self, input_ids, attention_mask, position_ids, return_dict, deterministic, *args)\u001b[0m\n\u001b[1;32m 13\u001b[0m \u001b[0mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'abc'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 14\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 15\u001b[0;31m \u001b[0moutputs\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mgpt2\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0minput_ids\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mattention_mask\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mposition_ids\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mreturn_dict\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mreturn_dict\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 16\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 17\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n", "\u001b[0;32m\u001b[0m in \u001b[0;36m__call__\u001b[0;34m(self, input_ids, attention_mask, position_ids, params, past_key_values, dropout_rng, train, output_attentions, output_hidden_states, return_dict)\u001b[0m\n\u001b[1;32m 112\u001b[0m \u001b[0mreturn_dict\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 113\u001b[0m \u001b[0mrngs\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mrngs\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 114\u001b[0;31m \u001b[0mmutable\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mmutable\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 115\u001b[0m )\n\u001b[1;32m 116\u001b[0m \u001b[0mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'cache'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/flax/linen/module.py\u001b[0m in \u001b[0;36mapply\u001b[0;34m(self, variables, rngs, method, mutable, capture_intermediates, *args, **kwargs)\u001b[0m\n\u001b[1;32m 964\u001b[0m \u001b[0mmethod\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 965\u001b[0m \u001b[0mmutable\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mmutable\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mcapture_intermediates\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mcapture_intermediates\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 966\u001b[0;31m )(variables, *args, **kwargs, rngs=rngs)\n\u001b[0m\u001b[1;32m 967\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 968\u001b[0m def init_with_output(self,\n", "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/flax/core/scope.py\u001b[0m in \u001b[0;36mwrapper\u001b[0;34m(variables, rngs, *args, **kwargs)\u001b[0m\n\u001b[1;32m 685\u001b[0m **kwargs) -> Union[Any, Tuple[Any, VariableDict]]:\n\u001b[1;32m 686\u001b[0m \u001b[0;32mwith\u001b[0m \u001b[0mbind\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mvariables\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mrngs\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mrngs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmutable\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mmutable\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtemporary\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0mroot\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 687\u001b[0;31m \u001b[0my\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mfn\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mroot\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 688\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0mmutable\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0;32mFalse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 689\u001b[0m \u001b[0;32mreturn\u001b[0m \u001b[0my\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mroot\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmutable_variables\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/flax/linen/module.py\u001b[0m in \u001b[0;36mscope_fn\u001b[0;34m(scope, *args, **kwargs)\u001b[0m\n\u001b[1;32m 1214\u001b[0m \u001b[0m_context\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcapture_stack\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mappend\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mcapture_intermediates\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 1215\u001b[0m \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1216\u001b[0;31m \u001b[0;32mreturn\u001b[0m \u001b[0mfn\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmodule\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mclone\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mparent\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mscope\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 1217\u001b[0m \u001b[0;32mfinally\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 1218\u001b[0m \u001b[0m_context\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcapture_stack\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mpop\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/flax/linen/module.py\u001b[0m in \u001b[0;36mwrapped_module_method\u001b[0;34m(*args, **kwargs)\u001b[0m\n\u001b[1;32m 281\u001b[0m \u001b[0m_context\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmodule_stack\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mappend\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 282\u001b[0m \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 283\u001b[0;31m \u001b[0my\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mfun\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 284\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0m_context\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcapture_stack\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 285\u001b[0m \u001b[0mfilter_fn\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0m_context\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcapture_stack\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m-\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", "\u001b[0;32m\u001b[0m in \u001b[0;36m__call__\u001b[0;34m(self, input_ids, attention_mask, position_ids, deterministic, init_cache, output_attentions, output_hidden_states, return_dict)\u001b[0m\n\u001b[1;32m 47\u001b[0m \u001b[0moutput_attentions\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0moutput_attentions\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 48\u001b[0m \u001b[0moutput_hidden_states\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0moutput_hidden_states\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 49\u001b[0;31m \u001b[0mreturn_dict\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mreturn_dict\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 50\u001b[0m )\n\u001b[1;32m 51\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n", "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/flax/linen/module.py\u001b[0m in \u001b[0;36mwrapped_module_method\u001b[0;34m(*args, **kwargs)\u001b[0m\n\u001b[1;32m 281\u001b[0m \u001b[0m_context\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmodule_stack\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mappend\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 282\u001b[0m \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 283\u001b[0;31m \u001b[0my\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mfun\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 284\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0m_context\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcapture_stack\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 285\u001b[0m \u001b[0mfilter_fn\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0m_context\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcapture_stack\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m-\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/transformers/models/gpt2/modeling_flax_gpt2.py\u001b[0m in \u001b[0;36m__call__\u001b[0;34m(self, hidden_states, attention_mask, deterministic, init_cache, output_attentions, output_hidden_states, return_dict)\u001b[0m\n\u001b[1;32m 452\u001b[0m \u001b[0mdeterministic\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mdeterministic\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 453\u001b[0m \u001b[0minit_cache\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0minit_cache\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 454\u001b[0;31m \u001b[0moutput_attentions\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0moutput_attentions\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 455\u001b[0m )\n\u001b[1;32m 456\u001b[0m \u001b[0mhidden_states\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mlayer_outputs\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/flax/linen/module.py\u001b[0m in \u001b[0;36mwrapped_module_method\u001b[0;34m(*args, **kwargs)\u001b[0m\n\u001b[1;32m 281\u001b[0m \u001b[0m_context\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmodule_stack\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mappend\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 282\u001b[0m \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 283\u001b[0;31m \u001b[0my\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mfun\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 284\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0m_context\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcapture_stack\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 285\u001b[0m \u001b[0mfilter_fn\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0m_context\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcapture_stack\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m-\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/transformers/models/gpt2/modeling_flax_gpt2.py\u001b[0m in \u001b[0;36m__call__\u001b[0;34m(self, hidden_states, attention_mask, deterministic, init_cache, output_attentions)\u001b[0m\n\u001b[1;32m 285\u001b[0m \u001b[0mdeterministic\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mdeterministic\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 286\u001b[0m \u001b[0minit_cache\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0minit_cache\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 287\u001b[0;31m \u001b[0moutput_attentions\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0moutput_attentions\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 288\u001b[0m )\n\u001b[1;32m 289\u001b[0m \u001b[0;31m# residual connection\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/flax/linen/module.py\u001b[0m in \u001b[0;36mwrapped_module_method\u001b[0;34m(*args, **kwargs)\u001b[0m\n\u001b[1;32m 281\u001b[0m \u001b[0m_context\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmodule_stack\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mappend\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 282\u001b[0m \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 283\u001b[0;31m \u001b[0my\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mfun\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 284\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0m_context\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcapture_stack\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 285\u001b[0m \u001b[0mfilter_fn\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0m_context\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcapture_stack\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m-\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/transformers/models/gpt2/modeling_flax_gpt2.py\u001b[0m in \u001b[0;36m__call__\u001b[0;34m(self, hidden_states, attention_mask, deterministic, init_cache, output_attentions)\u001b[0m\n\u001b[1;32m 177\u001b[0m ):\n\u001b[1;32m 178\u001b[0m \u001b[0mqkv_out\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mc_attn\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mhidden_states\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 179\u001b[0;31m \u001b[0mquery\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mkey\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mvalue\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mjnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msplit\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mqkv_out\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;36m3\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0maxis\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m2\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 180\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 181\u001b[0m \u001b[0mquery\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_split_heads\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mquery\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/jax/_src/numpy/lax_numpy.py\u001b[0m in \u001b[0;36msplit\u001b[0;34m(ary, indices_or_sections, axis)\u001b[0m\n\u001b[1;32m 1806\u001b[0m \u001b[0;34m@\u001b[0m\u001b[0m_wraps\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msplit\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 1807\u001b[0m \u001b[0;32mdef\u001b[0m \u001b[0msplit\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mary\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mindices_or_sections\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0maxis\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mint\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;36m0\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1808\u001b[0;31m \u001b[0;32mreturn\u001b[0m \u001b[0m_split\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"split\"\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mary\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mindices_or_sections\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0maxis\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0maxis\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 1809\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 1810\u001b[0m \u001b[0;32mdef\u001b[0m \u001b[0m_split_on_axis\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnp_fun\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0maxis\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/jax/_src/numpy/lax_numpy.py\u001b[0m in \u001b[0;36m_split\u001b[0;34m(op, ary, indices_or_sections, axis)\u001b[0m\n\u001b[1;32m 1798\u001b[0m + ((r + 1) * (part_size + 1) - 1)])\n\u001b[1;32m 1799\u001b[0m \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1800\u001b[0;31m \u001b[0;32mraise\u001b[0m \u001b[0mValueError\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"array split does not result in an equal division\"\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 1801\u001b[0m \u001b[0mstarts\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mends\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m*\u001b[0m \u001b[0mndim\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mary\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mshape\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mary\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 1802\u001b[0m \u001b[0m_subval\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;32mlambda\u001b[0m \u001b[0mx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mi\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mv\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0msubvals\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mv\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", "\u001b[0;31mValueError\u001b[0m: array split does not result in an equal division" ] } ] } ] }