Merge branch 'junior' of https://github.com/nomic-ai/gpt4all into junior

README.md

@@ -89,7 +89,7 @@ Please see [GPT4All-J Technical Report](https://static.nomic.ai/gpt4all/2023_GPT
 We have released updated versions of our `GPT4All-J` model and training data. 
 
 - `v1.0`: The original model trained on the v1.0 dataset
-- `v1.1-breezy`: Trained on afiltered dataset where we removed all instances of AI language model
+- `v1.1-breezy`: Trained on a filtered dataset where we removed all instances of AI language model
 - `v1.2-jazzy`: Trained on a filtered dataset where we also removed instances like I'm sorry, I can't answer... and AI language model
 
 The [models](https://huggingface.co/nomic-ai/gpt4all-j) and [data](https://huggingface.co/datasets/nomic-ai/gpt4all-j-prompt-generations) versions can be specified by passing a `revision` argument.

configs/deepspeed/ds_config_gptj.json

@@ -35,14 +35,5 @@
 		  ],
 		  "eps": 1e-08
 		}
-	  },
-	  "scheduler": {
-		"type": "WarmupLR",
-		"params": {
-		  "warmup_min_lr": 0,
-		  "warmup_max_lr": "auto",
-		  "warmup_num_steps": "auto",
-		  "warmup_type": "linear"
-		}
 	  }
 }

configs/train/finetune_gptjr.yaml

@@ -1,9 +1,10 @@
 # model/tokenizer
-model_name: "EleutherAI/gpt-j-6B"
-tokenizer_name: "EleutherAI/gpt-j-6B"
+model_name: "EleutherAI/gpt-j-6b"
+tokenizer_name: "EleutherAI/gpt-j-6b"
 version: null
 gradient_checkpointing: true
 save_name: "nomic-ai/gpt-jr-decay-alpha"
+push_to_hub: false
 encoder_dim: 384
 
 # dataset
@@ -11,7 +12,7 @@ streaming: false
 num_proc: 64
 dataset_path: "/home/paperspace/gpt4all/gpt4all/index/squad_supplemented_train" 
 max_length: 1024
-batch_size: 32
+batch_size: 8
 pct_test: 0.05
 q_column: "question"
 a_column: "answers"
@@ -23,7 +24,7 @@ lr: 1.0e-4
 min_lr: 0 
 weight_decay: 0.0
 eval_every: 50 
-save_every: 500
+save_every: -1
 log_grads_every: 100
 log_lr_every: 10
 output_dir: "ckpts/decay_alpha"
@@ -31,6 +32,8 @@ checkpoint: null
 lora: false
 warmup_steps: 500
 num_epochs: 5 
+debug: false
+scheduler: false
 
 # logging
 wandb: true

gpt4all/models/__init__.py

@@ -1,8 +1,12 @@
-from .configuration_gpt_jr import GPTJRConfig
-from .modeling_gpt_jr import GPTJRForCausalLM
+from .gpt_jr.configuration_gpt_jr import GPTJRConfig
+from .gpt_jr.modeling_gpt_jr import GPTJRForCausalLM
+
+from .pythiaseek import PythiaSeekForCausalLM, PythiaSeekConfig
 
 
 __all__ = [
     "GPTJRConfig",
-    "GPTJRForCausalLM"
+    "GPTJRForCausalLM",
+    "PythiaSeekConfig",
+    "PythiaSeekForCausalLM",
 ]

gpt4all/models/gpt_jr/modeling_gpt_jr.py

@@ -31,7 +31,7 @@ from transformers.modeling_outputs import (
 from transformers.modeling_utils import PreTrainedModel
 from transformers.utils import logging
 from transformers.utils.model_parallel_utils import assert_device_map, get_device_map
-from gpt4all.models.configuration_gpt_jr import GPTJRConfig
+from gpt4all.models.gpt_jr.configuration_gpt_jr import GPTJRConfig
 
 
 logger = logging.get_logger(__name__)
@@ -476,14 +476,13 @@ class GPTJRBlock(nn.Module):
             cross_attn_output[0] 
         )
 
-        # if step is not None:
-        #     alpha = self._update_alpha(step)
-        #     #alpha = alpha.to(cross_attn_ff.device).to(cross_attn_ff.dtype)
-        # else:
-        #     alpha = 0.5 
+        if step is not None:
+            alpha = self._update_alpha(step)
+        else:
+            alpha = 0.5 
+
+        hidden_states = (1 - alpha) * cross_attn_ff +  alpha * self_attention_residual  
 
-        # hidden_states = (1 - alpha) * cross_attn_ff +  alpha * self_attention_residual  
-        hidden_states = cross_attn_ff + self_attention_residual
         if use_cache:
             outputs = (hidden_states,) + outputs
         else:
@@ -491,10 +490,6 @@ class GPTJRBlock(nn.Module):
 
         return outputs  # hidden_states, present, (attentions)
         
-    # def _update_alpha(self, iteration):
-    #     return torch.clamp(torch.tensor([1 / (max(iteration * self.world_size, 1)) ** 0.08]), min=torch.tensor([0.5]), max=torch.tensor([1.0]))
-
-        
     def _update_alpha(self, current_step):
         """
         Computes the learning rate for the current step using a cosine decay schedule.
@@ -801,7 +796,8 @@ class GPTJRForCausalLM(GPTJRPreTrainedModel):
         self.encoder_dim = config.encoder_dim
 
         if self.hidden_size != self.encoder_dim:
-            self.enc_dec_proj = nn.Linear(config.encoder_dim, config.n_embd)
+            self.enc_dec_proj = nn.Sequential(nn.Linear(config.encoder_dim, config.n_embd * 4),
+                                              nn.Linear(config.n_embd * 4, config.n_embd))
 
         # Model parallel
         self.model_parallel = False
@@ -889,7 +885,7 @@ class GPTJRForCausalLM(GPTJRPreTrainedModel):
         return_dict = return_dict if return_dict is not None else self.config.use_return_dict
 
         if self.hidden_size != self.encoder_dim:
-            encoder_hidden_states = encoder_hidden_states.to(self.enc_dec_proj.weight.dtype)
+            encoder_hidden_states = encoder_hidden_states.to(self.enc_dec_proj[0].weight.dtype)
             encoder_hidden_states = self.enc_dec_proj(encoder_hidden_states)
 
         transformer_outputs = self.transformer(

gpt4all/models/gpt_jr/test_gpt_jr.py

@@ -35,7 +35,7 @@ encodings = mean_pooling(encoder(**tokenized), tokenized["attention_mask"])
 
 # make 2 neighbors
 # (bs, knn, encoding_dim)
-encoder_outputs = torch.stack([encodings, encodings]).squeeze().unsqueeze(0)
+encoder_hidden_states = torch.stack([encodings, encodings]).squeeze().unsqueeze(0)
 
 inputs = "What did the fox do?"
 
@@ -43,7 +43,7 @@ print("Encoded inputs")
 tokenized_input = tokenizer([inputs], padding="max_length", truncation=True, return_tensors="pt")
 
 print("Running model")
-outputs = model(**tokenized_input, encoder_outputs=encoder_outputs)
+outputs = model(**tokenized_input, encoder_hidden_states=encoder_hidden_states)
 
 print(outputs)
 print(outputs[0].shape)

gpt4all/models/pythiaseek/__init__.py

@@ -0,0 +1,2 @@
+from .configuration_pythiaseek import PythiaSeekConfig
+from .modeling_pythiaseek import PythiaSeekForCausalLM

gpt4all/models/pythiaseek/configuration_pythiaseek.py

@@ -0,0 +1,137 @@
+# coding=utf-8
+# Copyright 2022 EleutherAI and The HuggingFace Inc. team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+""" GPTNeoX model configuration"""
+
+from transformers.configuration_utils import PretrainedConfig
+from transformers.utils import logging
+
+
+logger = logging.get_logger(__name__)
+
+GPT_NEOX_PRETRAINED_CONFIG_ARCHIVE_MAP = {
+    "EleutherAI/gpt-neox-20b": "https://huggingface.co/EleutherAI/gpt-neox-20b/resolve/main/config.json",
+    # See all GPTNeoX models at https://huggingface.co/models?filter=gpt_neox
+}
+
+
+class PythiaSeekConfig(PretrainedConfig):
+    r"""
+    This is the configuration class to store the configuration of a [`GPTNeoXModel`]. It is used to instantiate an
+    GPTNeoX model according to the specified arguments, defining the model architecture. Instantiating a configuration
+    with the defaults will yield a similar configuration to that of the GPTNeoX
+    [EleutherAI/gpt-neox-20b](https://huggingface.co/EleutherAI/gpt-neox-20b) architecture.
+
+    Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
+    documentation from [`PretrainedConfig`] for more information.
+
+
+    Args:
+        vocab_size (`int`, *optional*, defaults to 50432):
+            Vocabulary size of the GPTNeoX model. Defines the number of different tokens that can be represented by the
+            `inputs_ids` passed when calling [`GPTNeoXModel`].
+        hidden_size (`int`, *optional*, defaults to 6144):
+            Dimension of the encoder layers and the pooler layer.
+        num_hidden_layers (`int`, *optional*, defaults to 44):
+            Number of hidden layers in the Transformer encoder.
+        num_attention_heads (`int`, *optional*, defaults to 64):
+            Number of attention heads for each attention layer in the Transformer encoder.
+        intermediate_size (`int`, *optional*, defaults to 24576):
+            Dimension of the "intermediate" (i.e., feed-forward) layer in the Transformer encoder.
+        hidden_act (`str` or `function`, *optional*, defaults to `"gelu"`):
+            The non-linear activation function (function or string) in the encoder and pooler. If string, `"gelu"`,
+            `"relu"`, `"selu"` and `"gelu_new"` are supported.
+        rotary_pct (`float`, *optional*, defaults to 0.25):
+            percentage of hidden dimensions to allocate to rotary embeddings
+        rotary_emb_base (`int`, *optional*, defaults to 10000)
+            base for computing rotary embeddings frequency
+        classifier_dropout (`float`, *optional*, defaults to 0.1):
+            Argument used when doing token classification, used in the model [`GPTNeoXForTokenClassification`].
+
+            The dropout ratio for the hidden layer.
+        max_position_embeddings (`int`, *optional*, defaults to 2048):
+            The maximum sequence length that this model might ever be used with. Typically set this to something large
+            just in case (e.g., 512 or 1024 or 2048).
+        initializer_range (`float`, *optional*, defaults to 1e-5):
+            The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
+        layer_norm_eps (`float`, *optional*, defaults to 1e-12):
+            The epsilon used by the layer normalization layers.
+        use_cache (`bool`, *optional*, defaults to `True`):
+            Whether or not the model should return the last key/values attentions (not used by all models). Only
+            relevant if `config.is_decoder=True`.
+        use_parallel_residual (`bool`, *optional*, defaults to `True`):
+            Whether to use a "parallel" formulation in each Transformer layer, which can provide a slight training
+            speedup at large scales (e.g. 20B).
+        Example:
+
+    ```python
+    >>> from transformers import GPTNeoXConfig, GPTNeoXModel
+
+    >>> # Initializing a GPTNeoX gpt-neox-20b style configuration
+    >>> configuration = GPTNeoXConfig()
+
+    >>> # Initializing a model (with random weights) from the gpt-neox-20b style configuration
+    >>> model = GPTNeoXModel(configuration)  # doctest: +SKIP
+
+    >>> # Accessing the model configuration
+    >>> configuration = model.config  # doctest: +SKIP
+    ```"""
+    model_type = "gpt_neox"
+
+    def __init__(
+        self,
+        vocab_size=50432,
+        hidden_size=6144,
+        num_hidden_layers=44,
+        num_attention_heads=64,
+        intermediate_size=24576,
+        hidden_act="gelu",
+        rotary_pct=0.25,
+        rotary_emb_base=10000,
+        classifier_dropout=0.1,
+        max_position_embeddings=2048,
+        initializer_range=0.02,
+        layer_norm_eps=1e-5,
+        use_cache=True,
+        bos_token_id=0,
+        eos_token_id=2,
+        tie_word_embeddings=False,
+        use_parallel_residual=True,
+        encoder_dim=4096,
+        total_alpha_steps=0,
+        initial_alpha=1,
+        final_alpha=.5,
+        **kwargs,
+    ):
+        super().__init__(bos_token_id=bos_token_id, eos_token_id=eos_token_id, **kwargs)
+        self.vocab_size = vocab_size
+        self.max_position_embeddings = max_position_embeddings
+        self.hidden_size = hidden_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.intermediate_size = intermediate_size
+        self.hidden_act = hidden_act
+        self.rotary_pct = rotary_pct
+        self.rotary_emb_base = rotary_emb_base
+        self.classifier_dropout = classifier_dropout
+        self.initializer_range = initializer_range
+        self.layer_norm_eps = layer_norm_eps
+        self.use_cache = use_cache
+        self.tie_word_embeddings = tie_word_embeddings
+        self.use_parallel_residual = use_parallel_residual
+
+        self.encoder_dim = encoder_dim
+        self.total_alpha_steps = total_alpha_steps
+        self.initial_alpha = initial_alpha
+        self.final_alpha = final_alpha

gpt4all/models/pythiaseek/modeling_pythiaseek.py

@@ -0,0 +1,885 @@
+# coding=utf-8
+# Copyright 2022 EleutherAI The HuggingFace Inc. team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+""" PyTorch PythiaSeek model."""
+
+from typing import Optional, Tuple, Union
+
+import math
+import torch
+import torch.utils.checkpoint
+from torch import nn
+from torch.nn import CrossEntropyLoss
+
+from transformers import AutoModel
+from transformers.activations import ACT2FN
+from transformers.modeling_outputs import (
+    BaseModelOutputWithPast,
+    CausalLMOutputWithPast,
+)
+from transformers.modeling_utils import PreTrainedModel
+from transformers.utils import logging
+from transformers.utils.model_parallel_utils import assert_device_map, get_device_map
+from gpt4all.models.pythiaseek import PythiaSeekConfig
+
+
+logger = logging.get_logger(__name__)
+
+
+GPT_NEOX_PRETRAINED_MODEL_ARCHIVE_LIST = [
+    "EleutherAI/gpt-neox-20b",
+    # See all PythiaSeek models at https://huggingface.co/models?filter=gpt_neox
+]
+
+
+class PythiaSeekPreTrainedModel(PreTrainedModel):
+    """
+    An abstract class to handle weights initialization and a simple interface for downloading and loading pretrained
+    models.
+    """
+
+    config_class = PythiaSeekConfig
+    base_model_prefix = "pythiaseek"
+    supports_gradient_checkpointing = True
+    _no_split_modules = ["PythiaSeekLayer"]
+
+    def _init_weights(self, module):
+        """Initialize the weights"""
+        if isinstance(module, nn.Linear):
+            module.weight.data.normal_(mean=0.0, std=self.config.initializer_range)
+            if module.bias is not None:
+                module.bias.data.zero_()
+        elif isinstance(module, nn.Embedding):
+            module.weight.data.normal_(mean=0.0, std=self.config.initializer_range)
+            if module.padding_idx is not None:
+                module.weight.data[module.padding_idx].zero_()
+        elif isinstance(module, nn.LayerNorm):
+            module.bias.data.zero_()
+            module.weight.data.fill_(1.0)
+
+    def _set_gradient_checkpointing(self, module, value=False):
+        if isinstance(module, PythiaSeekModel):
+            module.gradient_checkpointing = value
+
+
+class PythiaSeekAttention(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.num_attention_heads = config.num_attention_heads
+        self.hidden_size = config.hidden_size
+        self.head_size = self.hidden_size // self.num_attention_heads
+        self.rotary_ndims = int(self.head_size * config.rotary_pct)
+        max_positions = config.max_position_embeddings
+        self.register_buffer(
+            "bias",
+            torch.tril(torch.ones((max_positions, max_positions), dtype=torch.bool)).view(
+                1, 1, max_positions, max_positions
+            ),
+        )
+        self.register_buffer("masked_bias", torch.tensor(-1e9))
+        self.rotary_emb = RotaryEmbedding(
+            self.rotary_ndims, config.max_position_embeddings, base=config.rotary_emb_base
+        )
+        self.register_buffer(
+            "norm_factor",
+            torch.sqrt(torch.tensor(self.head_size, dtype=torch.float32)).to(torch.get_default_dtype()),
+            persistent=False,
+        )
+        self.query_key_value = nn.Linear(config.hidden_size, 3 * config.hidden_size)
+        self.dense = nn.Linear(config.hidden_size, config.hidden_size)
+
+    def forward(
+        self,
+        hidden_states: torch.FloatTensor,
+        attention_mask: torch.FloatTensor,
+        position_ids: torch.LongTensor,
+        head_mask: Optional[torch.FloatTensor] = None,
+        layer_past: Optional[Tuple[torch.Tensor]] = None,
+        use_cache: Optional[bool] = False,
+        output_attentions: Optional[bool] = False,
+    ):
+        has_layer_past = layer_past is not None
+
+        # Compute QKV
+        # Attention heads [batch, seq_len, hidden_size]
+        #   --> [batch, seq_len, (np * 3 * head_size)]
+        qkv = self.query_key_value(hidden_states)
+
+        # [batch, seq_len, (num_heads * 3 * head_size)]
+        #   --> [batch, seq_len, num_heads, 3 * head_size]
+        new_qkv_shape = qkv.size()[:-1] + (self.num_attention_heads, 3 * self.head_size)
+        qkv = qkv.view(*new_qkv_shape)
+
+        # [batch, seq_len, num_attention_heads, 3 * head_size] --> 3 [batch, num_attention_heads, seq_len, head_size]
+        query = qkv[..., : self.head_size].permute(0, 2, 1, 3)
+        key = qkv[..., self.head_size : 2 * self.head_size].permute(0, 2, 1, 3)
+        value = qkv[..., 2 * self.head_size :].permute(0, 2, 1, 3)
+
+        # Compute rotary embeddings on rotary_ndims
+        query_rot = query[..., : self.rotary_ndims]
+        query_pass = query[..., self.rotary_ndims :]
+        key_rot = key[..., : self.rotary_ndims]
+        key_pass = key[..., self.rotary_ndims :]
+
+        # Compute token offset for rotary embeddings (when decoding)
+        seq_len = key.shape[-2]
+        if has_layer_past:
+            seq_len += layer_past[0].shape[-2]
+        cos, sin = self.rotary_emb(value, seq_len=seq_len)
+        query, key = apply_rotary_pos_emb(query_rot, key_rot, cos, sin, position_ids)
+        query = torch.cat((query, query_pass), dim=-1)
+        key = torch.cat((key, key_pass), dim=-1)
+
+        # Cache QKV values
+        if has_layer_past:
+            past_key = layer_past[0]
+            past_value = layer_past[1]
+            key = torch.cat((past_key, key), dim=-2)
+            value = torch.cat((past_value, value), dim=-2)
+        present = (key, value) if use_cache else None
+
+        # Compute attention
+        attn_output, attn_weights = self._attn(query, key, value, attention_mask, head_mask)
+
+        # Reshape outputs
+        attn_output = self._merge_heads(attn_output, self.num_attention_heads, self.head_size)
+        attn_output = self.dense(attn_output)
+
+        outputs = (attn_output, present)
+        if output_attentions:
+            outputs += (attn_weights,)
+
+        return outputs
+
+    @classmethod
+    def _split_heads(cls, tensor, num_attention_heads, attn_head_size):
+        """
+        Splits hidden dim into attn_head_size and num_attention_heads
+        """
+        # tensor: [bs, seq_len, hidden_size]
+        new_shape = tensor.size()[:-1] + (num_attention_heads, attn_head_size)
+        # -> [bs, seq_len, num_attention_heads, attn_head_size]
+        tensor = tensor.view(new_shape)
+        # -> [bs, num_attention_heads, seq_len, attn_head_size]
+        tensor = tensor.permute(0, 2, 1, 3)
+        return tensor
+
+    @classmethod
+    def _merge_heads(cls, tensor, num_attention_heads, attn_head_size):
+        """
+        Merges attn_head_size dim and num_attn_heads dim into hidden dim
+        """
+        # tensor [bs, num_attention_heads, seq_len, attn_head_size]
+        tensor = tensor.permute(0, 2, 1, 3).contiguous()
+        # -> [bs, seq_len, num_attention_heads, attn_head_size]
+        tensor = tensor.view(tensor.size(0), tensor.size(1), num_attention_heads * attn_head_size)
+        # -> [bs, seq_len, hidden_size]
+        return tensor
+
+    def _attn(self, query, key, value, attention_mask=None, head_mask=None):
+        # q, k, v: [bs, num_attention_heads, seq_len, attn_head_size]
+        # compute causal mask from causal mask buffer
+        batch_size, num_attention_heads, query_length, attn_head_size = query.size()
+        key_length = key.size(-2)
+
+        causal_mask = self.bias[:, :, key_length - query_length : key_length, :key_length]
+
+        query = query.view(batch_size * num_attention_heads, query_length, attn_head_size)
+        key = key.view(batch_size * num_attention_heads, key_length, attn_head_size)
+        attn_scores = torch.zeros(
+            batch_size * num_attention_heads,
+            query_length,
+            key_length,
+            dtype=query.dtype,
+            device=key.device,
+        )
+        attn_scores = torch.baddbmm(
+            attn_scores,
+            query,
+            key.transpose(1, 2),
+            beta=1.0,
+            alpha=(torch.tensor(1.0, dtype=self.norm_factor.dtype, device=self.norm_factor.device) / self.norm_factor),
+        )
+        attn_scores = attn_scores.view(batch_size, num_attention_heads, query_length, key_length)
+
+        mask_value = torch.finfo(attn_scores.dtype).min
+        # Need to be a tensor, otherwise we get error: `RuntimeError: expected scalar type float but found double`.
+        # Need to be on the same device, otherwise `RuntimeError: ..., x and y to be on the same device`
+        mask_value = torch.tensor(mask_value, dtype=attn_scores.dtype).to(attn_scores.device)
+        attn_scores = torch.where(causal_mask, attn_scores, mask_value)
+
+        if attention_mask is not None:
+            # Apply the attention mask
+            attn_scores = attn_scores + attention_mask
+
+        attn_weights = nn.functional.softmax(attn_scores, dim=-1)
+        attn_weights = attn_weights.to(value.dtype)
+
+        # Mask heads if we want to
+        if head_mask is not None:
+            attn_weights = attn_weights * head_mask
+
+        attn_output = torch.matmul(attn_weights, value)
+        return attn_output, attn_weights
+
+class PythiaSeekCrossAttention(PythiaSeekAttention):
+    def __init__(self, config):
+        super().__init__(config)
+
+        max_positions = config.max_position_embeddings
+        self.register_buffer(
+            "bias",
+            torch.tril(torch.ones((max_positions, max_positions), dtype=torch.uint8)).view(
+                1, 1, max_positions, max_positions
+            ),
+        )
+        self.register_buffer("masked_bias", torch.tensor(-1e9))
+
+        self.embed_dim = config.hidden_size
+        self.num_attention_heads = config.num_attention_heads
+        self.head_dim = self.embed_dim // self.num_attention_heads
+        if self.head_dim * self.num_attention_heads != self.embed_dim:
+            raise ValueError(
+                f"embed_dim must be divisible by num_attention_heads (got `embed_dim`: {self.embed_dim} and"
+                f" `num_attention_heads`: {self.num_attention_heads})."
+            )
+        self.scale_attn = torch.sqrt(torch.tensor(self.head_dim, dtype=torch.float32)).to(torch.get_default_dtype())
+
+        self.k_proj = nn.Linear(self.embed_dim, self.embed_dim, bias=False)
+        self.v_proj = nn.Linear(self.embed_dim, self.embed_dim, bias=False)
+        self.q_proj = nn.Linear(self.embed_dim, self.embed_dim, bias=False)
+        self.out_proj = nn.Linear(self.embed_dim, self.embed_dim, bias=False)
+
+        
+    def _split_knn_attn_heads(self, tensor, num_attention_heads, attn_head_size):
+        new_shape = tensor.size()[:-1] + (num_attention_heads, attn_head_size)
+        tensor = tensor.view(new_shape)
+
+        return tensor.permute(0, 1, 3, 2)
+
+        
+    def _merge_heads(self, tensor, num_attention_heads, attn_head_size):
+        """
+        Merges attn_head_size dim and num_attn_heads dim into hidden dim
+        """
+        # tensor -> (bs, seq_len, num_attention_heads, head_dim)
+        tensor = tensor.permute(0, 2, 1, 3).contiguous()
+        new_shape = tensor.size()[:-2] + (num_attention_heads * attn_head_size,)
+        return tensor.view(new_shape)
+
+    def _attn(
+        self,
+        query,
+        key,
+        value,
+        attention_mask=None,
+        head_mask=None,
+    ):
+
+        # query -> (bs, num_attention_heads, seq_len, head_dim)
+        # key -> (bs, num_attention_heads, head_dim, neighbors)
+        # attn_weights -> (bs, num_attention_heads, seq_len, neighbors)
+        attn_weights = torch.matmul(query, key)
+
+        attn_weights = nn.functional.softmax(attn_weights, dim=-1)
+        attn_weights = attn_weights.to(value.dtype)
+
+        # Mask heads if we want to
+        if head_mask is not None:
+            attn_weights = attn_weights * head_mask
+
+        # value -> (bs, num_attention_heads, seq_len, head_dim)
+        # attn_weights -> (bs, num_attention_heads, seq_len, neighbors)
+        # attn_output -> (bs, num_attention_heads, seq_len, head_dim)
+        attn_output = torch.matmul(attn_weights, value)
+
+        return attn_output, attn_weights
+
+            
+    def forward(
+        self,
+        hidden_states: Optional[torch.FloatTensor],
+        encoder_hidden_states: Optional[torch.FloatTensor],
+        attention_mask: Optional[torch.FloatTensor] = None,
+        layer_past: Optional[Tuple[torch.Tensor]] = None,
+        head_mask: Optional[torch.FloatTensor] = None,
+        use_cache: Optional[bool] = False,
+        output_attentions: Optional[bool] = False,
+    ) -> Union[
+        Tuple[torch.Tensor, Tuple[torch.Tensor]],
+        Optional[Tuple[torch.Tensor, Tuple[torch.Tensor], Tuple[torch.Tensor, ...]]],
+    ]:
+
+        query = self.q_proj(hidden_states)
+        # if we are doing cross attention
+        key = self.k_proj(encoder_hidden_states)
+        value = self.v_proj(encoder_hidden_states)
+        # (bs, seq_len, dim) -> (bs, num_attention_heads, seq_len, head_dim)
+        query = self._split_heads(query, self.num_attention_heads, self.head_dim)
+        # (bs, dim) -> (bs, num_attention_heads,  head_dim)
+        key = self._split_knn_attn_heads(key, self.num_attention_heads, self.head_dim)
+        value = self._split_knn_attn_heads(value, self.num_attention_heads, self.head_dim)
+
+
+        value = value.permute(0, 3, 1, 2)
+        key = key.permute(0, 3, 2, 1)
+
+        if layer_past is not None:
+            past_key = layer_past[0]
+            past_value = layer_past[1]
+            key = torch.cat((past_key, key), dim=-2)
+            value = torch.cat((past_value, value), dim=-2)
+
+        if use_cache is True:
+            present = (key, value)
+        else:
+            present = None
+
+        # compute self-attention: V x Softmax(QK^T)
+        attn_output, attn_weights = self._attn(query, key, value, attention_mask, head_mask)
+
+        attn_output = self._merge_heads(attn_output, self.num_attention_heads, self.head_dim)
+        attn_output = self.out_proj(attn_output)
+
+        outputs = (attn_output, present)
+        if output_attentions:
+            outputs += (attn_weights,)
+
+        return outputs  # a, present, (attentions)
+
+
+def attention_mask_func(attention_scores, ltor_mask):
+    attention_scores.masked_fill_(~ltor_mask, torch.finfo(attention_scores.dtype).min)
+    return attention_scores
+
+
+class RotaryEmbedding(torch.nn.Module):
+    def __init__(self, dim, max_position_embeddings, base=10000, device=None):
+        super().__init__()
+        inv_freq = 1.0 / (base ** (torch.arange(0, dim, 2).float().to(device) / dim))
+        self.register_buffer("inv_freq", inv_freq)
+
+        # Build here to make `torch.jit.trace` work.
+        self.max_seq_len_cached = max_position_embeddings
+        t = torch.arange(self.max_seq_len_cached, device=self.inv_freq.device, dtype=self.inv_freq.dtype)
+        freqs = torch.einsum("i,j->ij", t, self.inv_freq)
+        # Different from paper, but it uses a different permutation in order to obtain the same calculation
+        emb = torch.cat((freqs, freqs), dim=-1)
+        self.cos_cached = emb.cos()[None, None, :, :]
+        self.sin_cached = emb.sin()[None, None, :, :]
+
+    def forward(self, x, seq_len=None):
+        # x: [bs, num_attention_heads, seq_len, head_size]
+        # This `if` block is unlikely to be run after we build sin/cos in `__init__`. Keep the logic here just in case.
+        if seq_len > self.max_seq_len_cached:
+            self.max_seq_len_cached = seq_len
+            t = torch.arange(self.max_seq_len_cached, device=x.device, dtype=self.inv_freq.dtype)
+            freqs = torch.einsum("i,j->ij", t, self.inv_freq)
+            # Different from paper, but it uses a different permutation in order to obtain the same calculation
+            emb = torch.cat((freqs, freqs), dim=-1).to(x.device)
+            self.cos_cached = emb.cos()[None, None, :, :]
+            self.sin_cached = emb.sin()[None, None, :, :]
+        return self.cos_cached[:seq_len, ...].to(x.device), self.sin_cached[:seq_len, ...].to(x.device)
+
+
+def rotate_half(x):
+    """Rotates half the hidden dims of the input."""
+    x1 = x[..., : x.shape[-1] // 2]
+    x2 = x[..., x.shape[-1] // 2 :]
+    return torch.cat((-x2, x1), dim=-1)
+
+
+def apply_rotary_pos_emb(q, k, cos, sin, position_ids):
+    gather_indices = position_ids[:, None, :, None]  # [bs, 1, seq_len, 1]
+    gather_indices = gather_indices.repeat(1, cos.shape[1], 1, cos.shape[3])
+    cos = torch.gather(cos.repeat(gather_indices.shape[0], 1, 1, 1), 2, gather_indices)
+    sin = torch.gather(sin.repeat(gather_indices.shape[0], 1, 1, 1), 2, gather_indices)
+    q_embed = (q * cos) + (rotate_half(q) * sin)
+    k_embed = (k * cos) + (rotate_half(k) * sin)
+    return q_embed, k_embed
+
+
+class PythiaSeekMLP(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.dense_h_to_4h = nn.Linear(config.hidden_size, config.intermediate_size)
+        self.dense_4h_to_h = nn.Linear(config.intermediate_size, config.hidden_size)
+        self.act = ACT2FN[config.hidden_act]
+
+    def forward(self, hidden_states):
+        hidden_states = self.dense_h_to_4h(hidden_states)
+        hidden_states = self.act(hidden_states)
+        hidden_states = self.dense_4h_to_h(hidden_states)
+        return hidden_states
+
+
+class PythiaSeekLayer(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.use_parallel_residual = config.use_parallel_residual
+        self.input_layernorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
+        self.post_attention_layernorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
+        self.attention = PythiaSeekAttention(config)
+        self.mlp = PythiaSeekMLP(config)
+
+        self.cross_attn_ln = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
+        self.cross_attn = PythiaSeekCrossAttention(config)
+        self.cross_attn_mlp = PythiaSeekMLP(config)
+        
+        self.total_alpha_steps = config.total_alpha_steps
+        self.initial_alpha = config.initial_alpha
+        self.final_alpha = config.final_alpha
+
+    def forward(
+        self,
+        hidden_states: Optional[torch.FloatTensor],
+        encoder_hidden_states: Optional[torch.FloatTensor],
+        attention_mask: Optional[torch.FloatTensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        head_mask: Optional[torch.FloatTensor] = None,
+        use_cache: Optional[bool] = False,
+        layer_past: Optional[Tuple[torch.Tensor]] = None,
+        output_attentions: Optional[bool] = False,
+        step: Optional[int] = None,
+    ):
+        ln_hidden_states = self.input_layernorm(hidden_states)
+        attention_layer_outputs = self.attention(
+            ln_hidden_states,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            layer_past=layer_past,
+            head_mask=head_mask,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+        )
+        attn_output = attention_layer_outputs[0]  # output_attn: attn_output, present, (attn_weights)
+        outputs = attention_layer_outputs[1:]
+
+        if self.use_parallel_residual:
+            # pseudocode:
+            # x = x + attn(ln1(x)) + mlp(ln2(x))
+            mlp_output = self.mlp(self.post_attention_layernorm(hidden_states))
+            self_attention_residual = mlp_output + attn_output + hidden_states
+        else:
+            # pseudocode:
+            # x = x + attn(ln1(x))
+            # x = x + mlp(ln2(x))
+            attn_output = attn_output + hidden_states
+            mlp_output = self.mlp(self.post_attention_layernorm(attn_output))
+            self_attention_residual = mlp_output + attn_output
+
+        # encoder_hidden_states -> (bs, knn, encoder_dim)
+        if encoder_hidden_states.dtype != ln_hidden_states.dtype:
+            encoder_hidden_states = encoder_hidden_states.to(ln_hidden_states.dtype)
+
+        encoder_normed = self.cross_attn_ln(encoder_hidden_states)
+
+        # cross_attn_outputs -> (bs, seq_len, dim)
+        cross_attn_output = self.cross_attn(
+            ln_hidden_states,
+            encoder_hidden_states=encoder_normed,
+            attention_mask=attention_mask,
+            head_mask=head_mask,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+        )
+
+        cross_attn_ff = self.cross_attn_mlp(
+            cross_attn_output[0] 
+        )
+
+        if step is not None:
+            alpha = self._update_alpha(step)
+        else:
+            alpha = 0.5 
+
+        hidden_states = (1 - alpha) * cross_attn_ff +  alpha * self_attention_residual  
+
+        if use_cache:
+            outputs = (hidden_states,) + outputs
+        else:
+            outputs = (hidden_states,) + outputs[1:]
+
+        return outputs  # hidden_states, present, (attentions)
+        
+    def _update_alpha(self, current_step):
+        """
+        Computes the learning rate for the current step using a cosine decay schedule.
+
+        Args:
+            initial_lr (float): The initial learning rate.
+            final_lr (float): The final learning rate.
+            total_steps (int): The total number of steps in the schedule.
+            current_step (int): The current step.
+
+        Returns:
+            float: The learning rate for the current step.
+        """
+        initial_alpha = 1
+        final_alpha = .5
+        if current_step >= self.total_alpha_steps:
+            return final_alpha
+
+        # Compute the cosine decay factor
+        cosine_decay = 0.5 * (1 + math.cos(math.pi * current_step / self.total_alpha_steps))
+
+        # Compute the current learning rate
+        alpha = final_alpha + (initial_alpha - final_alpha) * cosine_decay
+
+        return alpha
+
+
+
+class PythiaSeekModel(PythiaSeekPreTrainedModel):
+    def __init__(self, config):
+        super().__init__(config)
+        self.config = config
+
+        self.embed_in = nn.Embedding(config.vocab_size, config.hidden_size)
+        self.layers = nn.ModuleList([PythiaSeekLayer(config) for _ in range(config.num_hidden_layers)])
+        self.final_layer_norm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
+
+        self.gradient_checkpointing = False
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def get_input_embeddings(self):
+        return self.embed_in
+
+    def set_input_embeddings(self, value):
+        self.embed_in = value
+
+    def forward(
+        self,
+        input_ids: Optional[torch.LongTensor] = None,
+        encoder_hidden_states: Optional[torch.FloatTensor] = None,
+        attention_mask: Optional[torch.FloatTensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        head_mask: Optional[torch.FloatTensor] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        past_key_values: Optional[Tuple[Tuple[torch.FloatTensor]]] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+        step: Optional[int] = None
+    ) -> Union[Tuple, BaseModelOutputWithPast]:
+        r"""
+        past_key_values (`tuple(tuple(torch.FloatTensor))` of length `config.n_layers` with each tuple having 4 tensors of shape `(batch_size, num_heads, sequence_length - 1, embed_size_per_head)`):
+            Contains precomputed key and value hidden states of the attention blocks. Can be used to speed up decoding.
+            If `past_key_values` are used, the user can optionally input only the last `decoder_input_ids` (those that
+            don't have their past key value states given to this model) of shape `(batch_size, 1)` instead of all
+            `decoder_input_ids` of shape `(batch_size, sequence_length)`.
+        use_cache (`bool`, *optional*):
+            If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding (see
+            `past_key_values`).
+        """
+        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
+        output_hidden_states = (
+            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
+        )
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+        use_cache = use_cache if use_cache is not None else self.config.use_cache
+
+        if input_ids is not None and inputs_embeds is not None:
+            raise ValueError("You cannot specify both input_ids and inputs_embeds at the same time")
+        elif input_ids is not None:
+            input_shape = input_ids.size()
+        elif inputs_embeds is not None:
+            input_shape = inputs_embeds.size()[:-1]
+        else:
+            raise ValueError("You have to specify either input_ids or inputs_embeds")
+
+        batch_size, seq_length = input_shape
+
+        if past_key_values is None:
+            past_length = 0
+            past_key_values = tuple([None] * self.config.num_hidden_layers)
+        else:
+            past_length = past_key_values[0][0].size(-2)
+
+        if position_ids is None:
+            device = input_ids.device if input_ids is not None else inputs_embeds.device
+            position_ids = torch.arange(past_length, seq_length + past_length, dtype=torch.long, device=device)
+            position_ids = position_ids.unsqueeze(0).view(-1, seq_length)
+        else:
+            position_ids = position_ids.view(-1, seq_length).long()
+
+        # Attention mask.
+        if attention_mask is not None:
+            assert batch_size > 0, "batch_size has to be defined and > 0"
+            attention_mask = attention_mask.view(batch_size, -1)
+            # We create a 3D attention mask from a 2D tensor mask.
+            # Sizes are [batch_size, 1, 1, to_seq_length]
+            # So we can broadcast to [batch_size, num_heads, from_seq_length, to_seq_length]
+            # this attention mask is more simple than the triangular masking of causal attention
+            # used in OpenAI GPT, we just need to prepare the broadcast dimension here.
+            attention_mask = attention_mask[:, None, None, :]
+
+            # Since attention_mask is 1.0 for positions we want to attend and 0.0 for
+            # masked positions, this operation will create a tensor which is 0.0 for
+            # positions we want to attend and the dtype's smallest value for masked positions.
+            # Since we are adding it to the raw scores before the softmax, this is
+            # effectively the same as removing these entirely.
+            attention_mask = attention_mask.to(dtype=self.dtype)  # fp16 compatibility
+            attention_mask = (1.0 - attention_mask) * torch.finfo(self.dtype).min
+
+        # Prepare head mask if needed
+        # 1.0 in head_mask indicate we keep the head
+        # attention_probs has shape bsz x n_heads x N x N
+        # input head_mask has shape [num_heads] or [num_hidden_layers x num_heads]
+        # and head_mask is converted to shape [num_hidden_layers x batch x num_heads x seq_length x seq_length]
+        head_mask = self.get_head_mask(head_mask, self.config.num_hidden_layers)
+
+        if inputs_embeds is None:
+            inputs_embeds = self.embed_in(input_ids)
+
+        hidden_states = inputs_embeds
+
+        if self.gradient_checkpointing and self.training:
+            if use_cache:
+                logger.warning(
+                    "`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`..."
+                )
+                use_cache = False
+
+        presents = () if use_cache else None
+        all_attentions = () if output_attentions else None
+        all_hidden_states = () if output_hidden_states else None
+        for i, (layer, layer_past) in enumerate(zip(self.layers, past_key_values)):
+            if output_hidden_states:
+                all_hidden_states = all_hidden_states + (hidden_states,)
+
+            if self.gradient_checkpointing and self.training:
+
+                def create_custom_forward(module):
+                    def custom_forward(*inputs):
+                        # None for layer_past
+                        return module(*inputs, use_cache, None, output_attentions, step)
+
+                    return custom_forward
+
+                outputs = torch.utils.checkpoint.checkpoint(
+                    create_custom_forward(layer),
+                    hidden_states,
+                    encoder_hidden_states,
+                    attention_mask,
+                    position_ids,
+                    head_mask[i],
+                )
+            else:
+                outputs = layer(
+                    hidden_states=hidden_states,
+                    encoder_hidden_states=encoder_hidden_states,
+                    attention_mask=attention_mask,
+                    position_ids=position_ids,
+                    head_mask=head_mask[i],
+                    layer_past=layer_past,
+                    use_cache=use_cache,
+                    output_attentions=output_attentions,
+                    step=step,
+                )
+            hidden_states = outputs[0]
+            if use_cache is True:
+                presents = presents + (outputs[1],)
+            if output_attentions:
+                all_attentions = all_attentions + (outputs[2 if use_cache else 1],)
+
+        hidden_states = self.final_layer_norm(hidden_states)
+        # Add last hidden state
+        if output_hidden_states:
+            all_hidden_states = all_hidden_states + (hidden_states,)
+
+        if not return_dict:
+            return tuple(v for v in [hidden_states, presents, all_hidden_states, all_attentions] if v is not None)
+
+        return BaseModelOutputWithPast(
+            last_hidden_state=hidden_states,
+            past_key_values=presents,
+            hidden_states=all_hidden_states,
+            attentions=all_attentions,
+        )
+
+
+class PythiaSeekForCausalLM(PythiaSeekPreTrainedModel):
+    _keys_to_ignore_on_load_missing = [r"position_ids", r"predictions.decoder.bias"]
+
+    def __init__(self, config):
+        super().__init__(config)
+
+        self.pythiaseek = PythiaSeekModel(config)
+        self.embed_out = nn.Linear(config.hidden_size, config.vocab_size, bias=False)
+
+        self.hidden_size = config.hidden_size
+        self.encoder_dim = config.encoder_dim
+
+        if self.hidden_size != self.encoder_dim:
+            self.enc_dec_proj = nn.Sequential(nn.Linear(config.encoder_dim, config.hidden_size * 4),
+                                              nn.Linear(config.hidden_size * 4, config.hidden_size))
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def get_output_embeddings(self):
+        return self.embed_out
+
+    def set_output_embeddings(self, new_embeddings):
+        self.embed_out = new_embeddings
+
+    def forward(
+        self,
+        input_ids: torch.LongTensor,
+        encoder_hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.FloatTensor] = None,
+        past_key_values: Optional[Tuple[Tuple[torch.Tensor]]] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        head_mask: Optional[torch.FloatTensor] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        labels: Optional[torch.LongTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+        step: Optional[int] = None
+    ) -> Union[Tuple, CausalLMOutputWithPast]:
+        r"""
+        past_key_values (`tuple(tuple(torch.FloatTensor))`, *optional*, returned when `use_cache=True` is passed or when `config.use_cache=True`):
+            Tuple of `tuple(torch.FloatTensor)` of length `config.n_layers`, with each tuple having 2 tensors of shape
+            `(batch_size, num_heads, sequence_length, embed_size_per_head)`) and 2 additional tensors of shape
+            `(batch_size, num_heads, encoder_sequence_length, embed_size_per_head)`. The two additional tensors are
+            only required when the model is used as a decoder in a Sequence to Sequence model.
+
+            Contains pre-computed hidden-states (key and values in the self-attention blocks that can be used (see
+            `past_key_values` input) to speed up sequential decoding.
+
+            If `past_key_values` are used, the user can optionally input only the last `decoder_input_ids` (those that
+            don't have their past key value states given to this model) of shape `(batch_size, 1)` instead of all
+            `decoder_input_ids` of shape `(batch_size, sequence_length)`.
+        labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
+            Labels for computing the left-to-right language modeling loss (next word prediction). Indices should be in
+            `[-100, 0, ..., config.vocab_size]` (see `input_ids` docstring) Tokens with indices set to `-100` are
+            ignored (masked), the loss is only computed for the tokens with labels n `[0, ..., config.vocab_size]`.
+        use_cache (`bool`, *optional*):
+            If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding (see
+            `past_key_values`).
+
+        Returns:
+
+        Example:
+
+        ```python
+        >>> from transformers import AutoTokenizer, PythiaSeekForCausalLM, PythiaSeekConfig
+        >>> import torch
+
+        >>> tokenizer = AutoTokenizer.from_pretrained("EleutherAI/gpt-neox-20b")
+        >>> config = PythiaSeekConfig.from_pretrained("EleutherAI/gpt-neox-20b")
+        >>> config.is_decoder = True
+        >>> model = PythiaSeekForCausalLM.from_pretrained("EleutherAI/gpt-neox-20b", config=config)
+
+        >>> inputs = tokenizer("Hello, my dog is cute", return_tensors="pt")
+        >>> outputs = model(**inputs)
+
+        >>> prediction_logits = outputs.logits
+        ```"""
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        if self.hidden_size != self.encoder_dim:
+            encoder_hidden_states = encoder_hidden_states.to(self.enc_dec_proj[0].weight.dtype)
+            encoder_hidden_states = self.enc_dec_proj(encoder_hidden_states)
+
+        outputs = self.pythiaseek(
+            input_ids,
+            encoder_hidden_states=encoder_hidden_states,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            head_mask=head_mask,
+            inputs_embeds=inputs_embeds,
+            past_key_values=past_key_values,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+            step=step,
+        )
+
+        hidden_states = outputs[0]
+        lm_logits = self.embed_out(hidden_states)
+
+        lm_loss = None
+        if labels is not None:
+            # move labels to correct device to enable model parallelism
+            labels = labels.to(lm_logits.device)
+            # we are doing next-token prediction; shift prediction scores and input ids by one
+            shift_logits = lm_logits[:, :-1, :].contiguous()
+            labels = labels[:, 1:].contiguous()
+            loss_fct = CrossEntropyLoss()
+            lm_loss = loss_fct(shift_logits.view(-1, shift_logits.size(-1)), labels.view(-1))
+
+        if not return_dict:
+            output = (lm_logits,) + outputs[1:]
+            return ((lm_loss,) + output) if lm_loss is not None else output
+
+        return CausalLMOutputWithPast(
+            loss=lm_loss,
+            logits=lm_logits,
+            past_key_values=outputs.past_key_values,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )
+
+    def prepare_inputs_for_generation(
+        self, input_ids, past_key_values=None, attention_mask=None, inputs_embeds=None, **kwargs
+    ):
+        input_shape = input_ids.shape
+
+        # cut decoder_input_ids if past is used
+        if past_key_values and past_key_values[0] is not None:
+            input_ids = input_ids[:, -1:]
+
+        position_ids = kwargs.get("position_ids", None)
+        if attention_mask is not None and position_ids is None:
+            # create position_ids on the fly for batch generation
+            position_ids = attention_mask.long().cumsum(-1) - 1
+            position_ids.masked_fill_(attention_mask == 0, 1)
+            if past_key_values:
+                position_ids = position_ids[:, -1].unsqueeze(-1)
+
+        # if model is used as a decoder in encoder-decoder model, the decoder attention mask is created on the fly
+        if attention_mask is None:
+            attention_mask = input_ids.new_ones(input_shape)
+
+        # if `inputs_embeds` are passed, we only want to use them in the 1st generation step
+        if inputs_embeds is not None and past_key_values is None:
+            model_inputs = {"inputs_embeds": inputs_embeds}
+        else:
+            model_inputs = {"input_ids": input_ids}
+
+        model_inputs.update(
+            {
+                "attention_mask": attention_mask,
+                "past_key_values": past_key_values,
+                "position_ids": position_ids,
+            }
+        )
+
+        return model_inputs
+
+    def _reorder_cache(self, past_key_values, beam_idx):
+        reordered_past = ()
+        for layer_past in past_key_values:
+            reordered_past += (
+                tuple(past_state.index_select(0, beam_idx) for past_state in layer_past[:2]) + layer_past[2:],
+            )
+        return reordered_past

gpt4all/models/pythiaseek/test_pythiaseek.py

@@ -0,0 +1,55 @@
+import torch
+from gpt4all.models import PythiaSeekConfig, PythiaSeekForCausalLM
+from transformers import AutoTokenizer, AutoModel
+
+# seed torch 
+
+torch.manual_seed(0)
+
+config = PythiaSeekConfig(encoder_dim=384, 
+                          hidden_size=1024, 
+                          intermediate_size=4096,
+                          num_hidden_layers=4,
+                          num_attention_heads=4)
+print("loaded config")
+
+print("loading model")
+model = PythiaSeekForCausalLM(config)
+print("loaded model")
+
+
+tokenizer =  AutoTokenizer.from_pretrained("EleutherAI/pythia-1b")
+tokenizer.pad_token = tokenizer.eos_token
+tokenizer.model_max_length = 1024
+
+encoder_tokenizer = AutoTokenizer.from_pretrained('sentence-transformers/all-MiniLM-L6-v2')
+encoder = AutoModel.from_pretrained('sentence-transformers/all-MiniLM-L6-v2')
+
+
+def mean_pooling(model_output, attention_mask):
+    token_embeddings = model_output[0] #First element of model_output contains all token embeddings
+    input_mask_expanded = attention_mask.unsqueeze(-1).expand(token_embeddings.size()).float()
+    return torch.sum(token_embeddings * input_mask_expanded, 1) / torch.clamp(input_mask_expanded.sum(1), min=1e-9)
+
+text = "The quick brown fox jumps over the lazy dog."
+print("Encoded knn")
+tokenized = encoder_tokenizer(text, return_tensors="pt")
+
+# bs, seq_len, dim
+encodings = mean_pooling(encoder(**tokenized), tokenized["attention_mask"])
+
+# make 2 neighbors
+# (bs, knn, encoding_dim)
+encoder_hidden_states = torch.stack([encodings, encodings]).squeeze().unsqueeze(0)
+
+inputs = "What did the fox do?"
+
+print("Encoded inputs")
+tokenized_input = tokenizer([inputs], padding="max_length", truncation=True, return_tensors="pt")
+
+print("Running model")
+outputs = model(**tokenized_input, encoder_hidden_states=encoder_hidden_states)
+
+print(outputs)
+print(outputs[0].shape)
+

gpt4all/train/train_retrieval.py

@@ -75,13 +75,15 @@ def train(accelerator, config):
                                                     revision=config['version'] if 'version' in config else None,
                                                     use_cache=False if checkpoint else True,
                                                     encoder_dim=config["encoder_dim"],
-                                                    total_alpha_steps=total_num_steps
+                                                    # hardcoded!! TODO: change to config
+                                                    total_alpha_steps=10000,
                                                     ) 
     else:
         model = AutoModelForCausalLM.from_pretrained(config["model_name"], 
                                                      use_cache=False if checkpoint else True,
                                                      trust_remote_code=True) 
 
+    accelerator.print(f"Training a {model.num_parameters():,} parameter model")
     if checkpoint:
         model.gradient_checkpointing_enable()
 
@@ -106,28 +108,35 @@ def train(accelerator, config):
 
 
     # Creates Dummy Scheduler if `scheduler` was spcified in the config file else creates `args.lr_scheduler_type` Scheduler
-    if (
-        accelerator.state.deepspeed_plugin is None
-        or "scheduler" not in accelerator.state.deepspeed_plugin.deepspeed_config
-    ):
-        scheduler = get_scheduler(
-            name="cosine",
-            optimizer=optimizer,
-            num_warmup_steps=config["warmup_steps"] * accelerator.num_processes,
-            num_training_steps=total_num_steps,
+    if config["scheduler"] or "scheduler" in accelerator.state.deepspeed_plugin.deepspeed_config:
+        if (
+            accelerator.state.deepspeed_plugin is None
+            or "scheduler" not in accelerator.state.deepspeed_plugin.deepspeed_config
+        ):
+            scheduler = get_scheduler(
+                name="cosine",
+                optimizer=optimizer,
+                num_warmup_steps=config["warmup_steps"] * accelerator.num_processes,
+                num_training_steps=total_num_steps,
+            )
+        else:
+            scheduler = DummyScheduler(
+                optimizer, total_num_steps=total_num_steps, warmup_num_steps=config["warmup_steps"]
+            )
+        model, optimizer, train_dataloader, val_dataloader, scheduler = accelerator.prepare(
+                model, optimizer, train_dataloader, val_dataloader, scheduler
         )
+        scheduler = True
     else:
-        scheduler = DummyScheduler(
-            optimizer, total_num_steps=total_num_steps, warmup_num_steps=config["warmup_steps"]
+        model, optimizer, train_dataloader, val_dataloader = accelerator.prepare(
+                model, optimizer, train_dataloader, val_dataloader
         )
-
-    model, optimizer, train_dataloader, val_dataloader, scheduler = accelerator.prepare(
-            model, optimizer, train_dataloader, val_dataloader, scheduler
-    )
+        scheduler = False
 
 
-    # setup for saving training states in case preemption
-    accelerator.register_for_checkpointing(scheduler)
+    #  setup for saving training states in case preemption
+    if scheduler:
+        accelerator.register_for_checkpointing(scheduler)
 
     if config["checkpoint"]:
         accelerator.load_state(config["checkpoint"])
@@ -154,6 +163,22 @@ def train(accelerator, config):
                             encoder_hidden_states=batch["encoder_hidden_states"],
                             step=step)
             loss = outputs.loss
+            accelerator.print(f"Loss: {loss:.4f}")
+
+            if config["debug"]:
+                logits = outputs.logits
+                pred_tokens = torch.argmax(logits, dim=-1)
+                labels = batch["labels"].clone()
+
+                for row in range(labels.shape[0]):
+                    curr_label = labels[row]
+                    mask = curr_label != -100
+                    decoded_true = tokenizer.decode(curr_label[mask], skip_special_tokens=True)
+                    decoded_pred = tokenizer.decode(pred_tokens[row][mask], skip_special_tokens=True)
+
+                    accelerator.print(f"Predicted tokens: {decoded_pred}")
+                    accelerator.print(f"True tokens: {decoded_true}")
+
 
             # gather loss before backprop in case of gradient accumulation
             loss_values = accelerator.gather_for_metrics({"loss": loss.detach().float()})
@@ -167,15 +192,17 @@ def train(accelerator, config):
             if config["wandb"]:
                 if step > 0 and step % (config["log_lr_every"] ) == 0:
                     curr_step = step + epoch * len(train_dataloader)
-                    accelerator.log({"lr": scheduler.get_last_lr()[0]}, step=curr_step)
+                    lr = optimizer.param_groups[0]["lr"]
+                    accelerator.log({"lr": lr}, step=curr_step)
 
             if (step + 1) % gradient_accumulation_steps == 0 or step == len(train_dataloader) - 1:
                 optimizer.step()
-                scheduler.step()
+                if scheduler:
+                    scheduler.step()
                 optimizer.zero_grad()
 
 
-            if step > 0 and step % config["save_every"] == 0:
+            if step > 0 and config["save_every"] > 0 and step % config["save_every"] == 0:
                 curr_step = step + epoch * len(train_dataloader)
                 accelerator.save_state(f"{config['output_dir']}/step_{curr_step}")
 
@@ -194,17 +221,17 @@ def train(accelerator, config):
                     curr_step = step + epoch * len(train_dataloader)
                     accelerator.log({**log_train, **log_val}, step=curr_step)
 
-                accelerator.print(f"Current LR: {scheduler.get_last_lr()[0]}")
+                accelerator.print(f"Current LR: {optimizer.param_groups[0]['lr']}")
                 accelerator.print(format_metrics(log_train, "train", f" step {step} "))
                 accelerator.print(format_metrics(log_val, "val", f" step {step} "))
 
                 train_loss.reset()
 
         accelerator.print(f"Epoch {epoch} finished")
-        accelerator.print(f"Pushing to HF hub")
         accelerator.wait_for_everyone()
         unwrapped_model = accelerator.unwrap_model(model)
         if config["push_to_hub"]:
+            accelerator.print(f"Pushing to HF hub")
             try:
                 if accelerator.is_main_process:
                     unwrapped_model.push_to_hub(config["save_name"] + f"-epoch_{epoch}", private=True)