Upload model

README.md

@@ -0,0 +1,199 @@
+---
+library_name: transformers
+tags: []
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]

config.json

@@ -0,0 +1,39 @@
+{
+  "_name_or_path": "/data/nikit_ws/LOLA-Megatron-DeepSpeed/lola_ws/fine-tune/instruction-ft/output_model",
+  "activation_function": "gelu_fast",
+  "architectures": [
+    "LOLALMHeadModel"
+  ],
+  "attn_pdrop": 0.1,
+  "auto_map": {
+    "AutoConfig": "configuration_lola_gpt2.LOLAConfig",
+    "AutoModelForCausalLM": "modeling_lola_gpt2.LOLALMHeadModel"
+  },
+  "bos_token_id": 0,
+  "embd_pdrop": 0.1,
+  "eos_token_id": 5,
+  "initializer_range": 0.02,
+  "layer_norm_epsilon": 1e-05,
+  "model_type": "lola_v1",
+  "n_embd": 2048,
+  "n_head": 16,
+  "n_inner": 8192,
+  "n_layer": 24,
+  "n_positions": 2048,
+  "num_experts": 16,
+  "reorder_and_upcast_attn": false,
+  "resid_pdrop": 0.1,
+  "scale_attn_by_inverse_layer_idx": false,
+  "scale_attn_weights": true,
+  "summary_activation": null,
+  "summary_first_dropout": 0.1,
+  "summary_proj_to_labels": true,
+  "summary_type": "cls_index",
+  "summary_use_proj": true,
+  "tokenizer_class": "GPT2TokenizerFast",
+  "topk": 1,
+  "torch_dtype": "float32",
+  "transformers_version": "4.39.1",
+  "use_cache": true,
+  "vocab_size": 100000
+}

configuration_lola_gpt2.py

@@ -0,0 +1,80 @@
+from transformers.configuration_utils import PretrainedConfig
+from transformers.utils import logging
+from transformers import GPT2Config
+
+
+logger = logging.get_logger(__name__)
+
+
+class LOLAConfig(PretrainedConfig):
+    """
+    This is the configuration class is a modified copy of https://huggingface.co/openai-community/gpt2 with MoE support.
+    """
+
+    model_type = "lola_v1"
+    keys_to_ignore_at_inference = ["past_key_values"]
+    attribute_map = {
+        "hidden_size": "n_embd",
+        "max_position_embeddings": "n_positions",
+        "num_attention_heads": "n_head",
+        "num_hidden_layers": "n_layer",
+    }
+
+    def __init__(
+        self,
+        vocab_size=100096,
+        n_positions=2048,
+        n_embd=2048,
+        n_layer=24,
+        n_head=16,
+        n_inner=8192,
+        activation_function="gelu_new",
+        resid_pdrop=0.1,
+        embd_pdrop=0.1,
+        attn_pdrop=0.1,
+        layer_norm_epsilon=1e-5,
+        initializer_range=0.02,
+        summary_type="cls_index",
+        summary_use_proj=True,
+        summary_activation=None,
+        summary_proj_to_labels=True,
+        summary_first_dropout=0.1,
+        scale_attn_weights=True,
+        use_cache=True,
+        bos_token_id=100095,
+        eos_token_id=100095,
+        scale_attn_by_inverse_layer_idx=False,
+        reorder_and_upcast_attn=False,
+        num_experts=16,
+        topk=1,
+        **kwargs,
+    ):
+        self.vocab_size = vocab_size
+        self.n_positions = n_positions
+        self.n_embd = n_embd
+        self.n_layer = n_layer
+        self.n_head = n_head
+        self.n_inner = n_inner
+        self.activation_function = activation_function
+        self.resid_pdrop = resid_pdrop
+        self.embd_pdrop = embd_pdrop
+        self.attn_pdrop = attn_pdrop
+        self.layer_norm_epsilon = layer_norm_epsilon
+        self.initializer_range = initializer_range
+        self.summary_type = summary_type
+        self.summary_use_proj = summary_use_proj
+        self.summary_activation = summary_activation
+        self.summary_first_dropout = summary_first_dropout
+        self.summary_proj_to_labels = summary_proj_to_labels
+        self.scale_attn_weights = scale_attn_weights
+        self.use_cache = use_cache
+        self.scale_attn_by_inverse_layer_idx = scale_attn_by_inverse_layer_idx
+        self.reorder_and_upcast_attn = reorder_and_upcast_attn
+        self.num_experts = num_experts
+        self.topk = topk
+
+        self.bos_token_id = bos_token_id
+        self.eos_token_id = eos_token_id
+
+        super().__init__(bos_token_id=bos_token_id, eos_token_id=eos_token_id, **kwargs)
+

generation_config.json

@@ -0,0 +1,6 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 100095,
+  "eos_token_id": 100095,
+  "transformers_version": "4.39.1"
+}

model-00001-of-00006.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8e5144c8d82e4165698bea8813c9066134d6c057961f078012b9488c5e7707b1
+size 4998456040

model-00002-of-00006.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3e724984ef8a0e278f869983b4bf037c5ddab4d4f188fb09a3d1b32cd16315bd
+size 4968033584

model-00003-of-00006.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2f28ccb2eac6cb01f6eedea1d1fb5015404d64b08e7a3f864227adcb6a203624
+size 4968033640

model-00004-of-00006.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e8a462d1a2c848ce90488c25270a1ee1400161c83840161e91a6915f84a94065
+size 4968033752

model-00005-of-00006.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:26ac1bb603c614bdd5bc2b3abbbd174a4df3c463cf604fc07069be2acb32527d
+size 4968033752

model-00006-of-00006.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aca6e2e92057414dc81286e2e123f5c4c16c2bbf98a628607b2ac2be235d17b1
+size 4968050328

model.safetensors.index.json

@@ -0,0 +1,1031 @@
+{
+  "metadata": {
+    "total_size": 29838524416
+  },
+  "weight_map": {
+    "transformer.h.0.attn.c_attn.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.0.attn.c_attn.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.0.attn.c_proj.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.0.attn.c_proj.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.0.ln_1.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.0.ln_1.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.0.ln_2.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.0.ln_2.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.0.mlp.c_fc.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.0.mlp.c_fc.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.0.mlp.c_proj.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.0.mlp.c_proj.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.1.attn.c_attn.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.1.attn.c_attn.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.1.attn.c_proj.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.1.attn.c_proj.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.1.ln_1.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.1.ln_1.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.1.ln_2.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.1.ln_2.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.0.c_fc.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.0.c_fc.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.0.c_proj.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.0.c_proj.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.1.c_fc.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.1.c_fc.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.1.c_proj.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.1.c_proj.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.10.c_fc.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.10.c_fc.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.10.c_proj.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.10.c_proj.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.11.c_fc.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.11.c_fc.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.11.c_proj.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.11.c_proj.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.12.c_fc.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.12.c_fc.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.12.c_proj.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.12.c_proj.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.13.c_fc.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.13.c_fc.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.13.c_proj.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.13.c_proj.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.14.c_fc.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.14.c_fc.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.14.c_proj.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.14.c_proj.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.15.c_fc.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.15.c_fc.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.15.c_proj.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.15.c_proj.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.2.c_fc.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.2.c_fc.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.2.c_proj.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.2.c_proj.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.3.c_fc.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.3.c_fc.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.3.c_proj.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.3.c_proj.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.4.c_fc.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.4.c_fc.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.4.c_proj.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.4.c_proj.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.5.c_fc.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.5.c_fc.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.5.c_proj.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.5.c_proj.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.6.c_fc.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.6.c_fc.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.6.c_proj.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.6.c_proj.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.7.c_fc.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.7.c_fc.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.7.c_proj.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.7.c_proj.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.8.c_fc.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.8.c_fc.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.8.c_proj.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.8.c_proj.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.9.c_fc.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.9.c_fc.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.9.c_proj.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.experts.9.c_proj.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.1.moe.gate.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.10.attn.c_attn.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.10.attn.c_attn.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.10.attn.c_proj.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.10.attn.c_proj.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.10.ln_1.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.10.ln_1.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.10.ln_2.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.10.ln_2.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.10.mlp.c_fc.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.10.mlp.c_fc.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.10.mlp.c_proj.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.10.mlp.c_proj.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.11.attn.c_attn.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.11.attn.c_attn.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.11.attn.c_proj.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.11.attn.c_proj.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.11.ln_1.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.11.ln_1.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.11.ln_2.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.11.ln_2.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.0.c_fc.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.0.c_fc.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.0.c_proj.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.0.c_proj.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.1.c_fc.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.1.c_fc.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.1.c_proj.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.1.c_proj.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.10.c_fc.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.10.c_fc.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.10.c_proj.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.10.c_proj.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.11.c_fc.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.11.c_fc.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.11.c_proj.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.11.c_proj.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.12.c_fc.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.12.c_fc.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.12.c_proj.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.12.c_proj.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.13.c_fc.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.11.moe.experts.13.c_fc.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.11.moe.experts.13.c_proj.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.11.moe.experts.13.c_proj.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.11.moe.experts.14.c_fc.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.11.moe.experts.14.c_fc.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.11.moe.experts.14.c_proj.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.11.moe.experts.14.c_proj.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.11.moe.experts.15.c_fc.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.11.moe.experts.15.c_fc.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.11.moe.experts.15.c_proj.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.11.moe.experts.15.c_proj.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.11.moe.experts.2.c_fc.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.2.c_fc.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.2.c_proj.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.2.c_proj.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.3.c_fc.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.3.c_fc.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.3.c_proj.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.3.c_proj.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.4.c_fc.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.4.c_fc.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.4.c_proj.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.4.c_proj.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.5.c_fc.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.5.c_fc.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.5.c_proj.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.5.c_proj.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.6.c_fc.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.6.c_fc.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.6.c_proj.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.6.c_proj.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.7.c_fc.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.7.c_fc.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.7.c_proj.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.7.c_proj.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.8.c_fc.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.8.c_fc.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.8.c_proj.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.8.c_proj.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.9.c_fc.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.9.c_fc.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.9.c_proj.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.experts.9.c_proj.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.11.moe.gate.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.12.attn.c_attn.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.12.attn.c_attn.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.12.attn.c_proj.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.12.attn.c_proj.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.12.ln_1.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.12.ln_1.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.12.ln_2.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.12.ln_2.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.12.mlp.c_fc.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.12.mlp.c_fc.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.12.mlp.c_proj.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.12.mlp.c_proj.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.13.attn.c_attn.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.13.attn.c_attn.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.13.attn.c_proj.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.13.attn.c_proj.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.13.ln_1.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.13.ln_1.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.13.ln_2.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.13.ln_2.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.0.c_fc.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.0.c_fc.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.0.c_proj.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.0.c_proj.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.1.c_fc.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.1.c_fc.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.1.c_proj.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.1.c_proj.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.10.c_fc.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.10.c_fc.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.10.c_proj.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.10.c_proj.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.11.c_fc.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.11.c_fc.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.11.c_proj.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.11.c_proj.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.12.c_fc.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.12.c_fc.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.12.c_proj.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.12.c_proj.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.13.c_fc.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.13.c_fc.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.13.c_proj.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.13.c_proj.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.14.c_fc.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.14.c_fc.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.14.c_proj.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.14.c_proj.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.15.c_fc.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.15.c_fc.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.15.c_proj.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.15.c_proj.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.2.c_fc.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.2.c_fc.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.2.c_proj.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.2.c_proj.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.3.c_fc.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.3.c_fc.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.3.c_proj.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.3.c_proj.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.4.c_fc.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.4.c_fc.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.4.c_proj.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.4.c_proj.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.5.c_fc.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.5.c_fc.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.5.c_proj.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.5.c_proj.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.6.c_fc.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.6.c_fc.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.6.c_proj.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.6.c_proj.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.7.c_fc.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.7.c_fc.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.7.c_proj.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.7.c_proj.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.8.c_fc.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.8.c_fc.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.8.c_proj.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.8.c_proj.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.9.c_fc.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.9.c_fc.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.9.c_proj.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.experts.9.c_proj.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.13.moe.gate.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.14.attn.c_attn.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.14.attn.c_attn.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.14.attn.c_proj.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.14.attn.c_proj.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.14.ln_1.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.14.ln_1.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.14.ln_2.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.14.ln_2.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.14.mlp.c_fc.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.14.mlp.c_fc.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.14.mlp.c_proj.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.14.mlp.c_proj.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.15.attn.c_attn.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.15.attn.c_attn.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.15.attn.c_proj.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.15.attn.c_proj.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.15.ln_1.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.15.ln_1.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.15.ln_2.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.15.ln_2.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.0.c_fc.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.0.c_fc.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.0.c_proj.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.0.c_proj.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.1.c_fc.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.1.c_fc.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.1.c_proj.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.1.c_proj.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.10.c_fc.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.10.c_fc.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.10.c_proj.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.10.c_proj.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.11.c_fc.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.11.c_fc.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.11.c_proj.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.11.c_proj.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.12.c_fc.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.12.c_fc.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.12.c_proj.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.12.c_proj.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.13.c_fc.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.13.c_fc.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.13.c_proj.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.13.c_proj.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.14.c_fc.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.15.moe.experts.14.c_fc.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.15.moe.experts.14.c_proj.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.15.moe.experts.14.c_proj.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.15.moe.experts.15.c_fc.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.15.moe.experts.15.c_fc.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.15.moe.experts.15.c_proj.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.15.moe.experts.15.c_proj.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.15.moe.experts.2.c_fc.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.2.c_fc.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.2.c_proj.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.2.c_proj.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.3.c_fc.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.3.c_fc.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.3.c_proj.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.3.c_proj.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.4.c_fc.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.4.c_fc.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.4.c_proj.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.4.c_proj.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.5.c_fc.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.5.c_fc.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.5.c_proj.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.5.c_proj.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.6.c_fc.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.6.c_fc.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.6.c_proj.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.6.c_proj.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.7.c_fc.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.7.c_fc.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.7.c_proj.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.7.c_proj.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.8.c_fc.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.8.c_fc.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.8.c_proj.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.8.c_proj.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.9.c_fc.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.9.c_fc.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.9.c_proj.bias": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.experts.9.c_proj.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.15.moe.gate.weight": "model-00004-of-00006.safetensors",
+    "transformer.h.16.attn.c_attn.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.16.attn.c_attn.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.16.attn.c_proj.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.16.attn.c_proj.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.16.ln_1.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.16.ln_1.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.16.ln_2.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.16.ln_2.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.16.mlp.c_fc.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.16.mlp.c_fc.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.16.mlp.c_proj.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.16.mlp.c_proj.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.17.attn.c_attn.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.17.attn.c_attn.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.17.attn.c_proj.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.17.attn.c_proj.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.17.ln_1.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.17.ln_1.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.17.ln_2.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.17.ln_2.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.0.c_fc.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.0.c_fc.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.0.c_proj.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.0.c_proj.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.1.c_fc.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.1.c_fc.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.1.c_proj.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.1.c_proj.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.10.c_fc.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.10.c_fc.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.10.c_proj.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.10.c_proj.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.11.c_fc.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.11.c_fc.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.11.c_proj.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.11.c_proj.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.12.c_fc.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.12.c_fc.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.12.c_proj.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.12.c_proj.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.13.c_fc.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.13.c_fc.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.13.c_proj.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.13.c_proj.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.14.c_fc.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.14.c_fc.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.14.c_proj.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.14.c_proj.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.15.c_fc.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.15.c_fc.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.15.c_proj.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.15.c_proj.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.2.c_fc.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.2.c_fc.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.2.c_proj.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.2.c_proj.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.3.c_fc.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.3.c_fc.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.3.c_proj.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.3.c_proj.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.4.c_fc.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.4.c_fc.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.4.c_proj.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.4.c_proj.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.5.c_fc.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.5.c_fc.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.5.c_proj.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.5.c_proj.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.6.c_fc.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.6.c_fc.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.6.c_proj.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.6.c_proj.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.7.c_fc.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.7.c_fc.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.7.c_proj.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.7.c_proj.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.8.c_fc.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.8.c_fc.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.8.c_proj.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.8.c_proj.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.9.c_fc.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.9.c_fc.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.9.c_proj.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.experts.9.c_proj.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.17.moe.gate.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.18.attn.c_attn.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.18.attn.c_attn.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.18.attn.c_proj.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.18.attn.c_proj.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.18.ln_1.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.18.ln_1.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.18.ln_2.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.18.ln_2.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.18.mlp.c_fc.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.18.mlp.c_fc.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.18.mlp.c_proj.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.18.mlp.c_proj.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.19.attn.c_attn.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.19.attn.c_attn.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.19.attn.c_proj.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.19.attn.c_proj.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.19.ln_1.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.19.ln_1.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.19.ln_2.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.19.ln_2.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.0.c_fc.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.0.c_fc.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.0.c_proj.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.0.c_proj.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.1.c_fc.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.1.c_fc.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.1.c_proj.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.1.c_proj.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.10.c_fc.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.10.c_fc.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.10.c_proj.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.10.c_proj.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.11.c_fc.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.11.c_fc.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.11.c_proj.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.11.c_proj.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.12.c_fc.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.12.c_fc.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.12.c_proj.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.12.c_proj.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.13.c_fc.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.13.c_fc.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.13.c_proj.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.13.c_proj.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.14.c_fc.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.14.c_fc.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.14.c_proj.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.14.c_proj.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.15.c_fc.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.19.moe.experts.15.c_fc.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.19.moe.experts.15.c_proj.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.19.moe.experts.15.c_proj.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.19.moe.experts.2.c_fc.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.2.c_fc.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.2.c_proj.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.2.c_proj.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.3.c_fc.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.3.c_fc.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.3.c_proj.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.3.c_proj.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.4.c_fc.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.4.c_fc.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.4.c_proj.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.4.c_proj.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.5.c_fc.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.5.c_fc.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.5.c_proj.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.5.c_proj.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.6.c_fc.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.6.c_fc.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.6.c_proj.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.6.c_proj.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.7.c_fc.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.7.c_fc.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.7.c_proj.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.7.c_proj.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.8.c_fc.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.8.c_fc.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.8.c_proj.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.8.c_proj.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.9.c_fc.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.9.c_fc.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.9.c_proj.bias": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.experts.9.c_proj.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.19.moe.gate.weight": "model-00005-of-00006.safetensors",
+    "transformer.h.2.attn.c_attn.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.2.attn.c_attn.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.2.attn.c_proj.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.2.attn.c_proj.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.2.ln_1.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.2.ln_1.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.2.ln_2.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.2.ln_2.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.2.mlp.c_fc.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.2.mlp.c_fc.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.2.mlp.c_proj.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.2.mlp.c_proj.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.20.attn.c_attn.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.20.attn.c_attn.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.20.attn.c_proj.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.20.attn.c_proj.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.20.ln_1.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.20.ln_1.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.20.ln_2.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.20.ln_2.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.20.mlp.c_fc.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.20.mlp.c_fc.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.20.mlp.c_proj.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.20.mlp.c_proj.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.21.attn.c_attn.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.21.attn.c_attn.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.21.attn.c_proj.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.21.attn.c_proj.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.21.ln_1.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.21.ln_1.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.21.ln_2.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.21.ln_2.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.0.c_fc.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.0.c_fc.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.0.c_proj.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.0.c_proj.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.1.c_fc.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.1.c_fc.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.1.c_proj.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.1.c_proj.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.10.c_fc.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.10.c_fc.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.10.c_proj.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.10.c_proj.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.11.c_fc.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.11.c_fc.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.11.c_proj.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.11.c_proj.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.12.c_fc.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.12.c_fc.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.12.c_proj.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.12.c_proj.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.13.c_fc.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.13.c_fc.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.13.c_proj.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.13.c_proj.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.14.c_fc.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.14.c_fc.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.14.c_proj.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.14.c_proj.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.15.c_fc.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.15.c_fc.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.15.c_proj.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.15.c_proj.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.2.c_fc.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.2.c_fc.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.2.c_proj.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.2.c_proj.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.3.c_fc.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.3.c_fc.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.3.c_proj.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.3.c_proj.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.4.c_fc.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.4.c_fc.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.4.c_proj.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.4.c_proj.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.5.c_fc.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.5.c_fc.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.5.c_proj.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.5.c_proj.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.6.c_fc.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.6.c_fc.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.6.c_proj.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.6.c_proj.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.7.c_fc.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.7.c_fc.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.7.c_proj.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.7.c_proj.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.8.c_fc.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.8.c_fc.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.8.c_proj.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.8.c_proj.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.9.c_fc.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.9.c_fc.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.9.c_proj.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.experts.9.c_proj.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.21.moe.gate.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.22.attn.c_attn.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.22.attn.c_attn.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.22.attn.c_proj.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.22.attn.c_proj.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.22.ln_1.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.22.ln_1.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.22.ln_2.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.22.ln_2.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.22.mlp.c_fc.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.22.mlp.c_fc.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.22.mlp.c_proj.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.22.mlp.c_proj.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.23.attn.c_attn.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.23.attn.c_attn.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.23.attn.c_proj.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.23.attn.c_proj.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.23.ln_1.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.23.ln_1.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.23.ln_2.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.23.ln_2.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.0.c_fc.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.0.c_fc.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.0.c_proj.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.0.c_proj.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.1.c_fc.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.1.c_fc.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.1.c_proj.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.1.c_proj.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.10.c_fc.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.10.c_fc.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.10.c_proj.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.10.c_proj.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.11.c_fc.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.11.c_fc.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.11.c_proj.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.11.c_proj.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.12.c_fc.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.12.c_fc.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.12.c_proj.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.12.c_proj.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.13.c_fc.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.13.c_fc.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.13.c_proj.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.13.c_proj.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.14.c_fc.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.14.c_fc.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.14.c_proj.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.14.c_proj.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.15.c_fc.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.15.c_fc.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.15.c_proj.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.15.c_proj.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.2.c_fc.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.2.c_fc.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.2.c_proj.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.2.c_proj.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.3.c_fc.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.3.c_fc.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.3.c_proj.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.3.c_proj.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.4.c_fc.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.4.c_fc.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.4.c_proj.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.4.c_proj.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.5.c_fc.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.5.c_fc.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.5.c_proj.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.5.c_proj.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.6.c_fc.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.6.c_fc.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.6.c_proj.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.6.c_proj.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.7.c_fc.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.7.c_fc.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.7.c_proj.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.7.c_proj.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.8.c_fc.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.8.c_fc.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.8.c_proj.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.8.c_proj.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.9.c_fc.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.9.c_fc.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.9.c_proj.bias": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.experts.9.c_proj.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.23.moe.gate.weight": "model-00006-of-00006.safetensors",
+    "transformer.h.3.attn.c_attn.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.3.attn.c_attn.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.3.attn.c_proj.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.3.attn.c_proj.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.3.ln_1.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.3.ln_1.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.3.ln_2.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.3.ln_2.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.0.c_fc.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.0.c_fc.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.0.c_proj.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.0.c_proj.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.1.c_fc.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.1.c_fc.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.1.c_proj.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.1.c_proj.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.10.c_fc.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.10.c_fc.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.10.c_proj.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.10.c_proj.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.11.c_fc.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.3.moe.experts.11.c_fc.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.3.moe.experts.11.c_proj.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.3.moe.experts.11.c_proj.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.3.moe.experts.12.c_fc.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.3.moe.experts.12.c_fc.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.3.moe.experts.12.c_proj.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.3.moe.experts.12.c_proj.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.3.moe.experts.13.c_fc.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.3.moe.experts.13.c_fc.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.3.moe.experts.13.c_proj.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.3.moe.experts.13.c_proj.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.3.moe.experts.14.c_fc.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.3.moe.experts.14.c_fc.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.3.moe.experts.14.c_proj.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.3.moe.experts.14.c_proj.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.3.moe.experts.15.c_fc.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.3.moe.experts.15.c_fc.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.3.moe.experts.15.c_proj.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.3.moe.experts.15.c_proj.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.3.moe.experts.2.c_fc.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.2.c_fc.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.2.c_proj.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.2.c_proj.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.3.c_fc.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.3.c_fc.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.3.c_proj.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.3.c_proj.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.4.c_fc.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.4.c_fc.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.4.c_proj.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.4.c_proj.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.5.c_fc.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.5.c_fc.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.5.c_proj.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.5.c_proj.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.6.c_fc.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.6.c_fc.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.6.c_proj.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.6.c_proj.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.7.c_fc.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.7.c_fc.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.7.c_proj.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.7.c_proj.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.8.c_fc.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.8.c_fc.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.8.c_proj.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.8.c_proj.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.9.c_fc.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.9.c_fc.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.9.c_proj.bias": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.experts.9.c_proj.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.3.moe.gate.weight": "model-00001-of-00006.safetensors",
+    "transformer.h.4.attn.c_attn.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.4.attn.c_attn.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.4.attn.c_proj.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.4.attn.c_proj.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.4.ln_1.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.4.ln_1.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.4.ln_2.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.4.ln_2.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.4.mlp.c_fc.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.4.mlp.c_fc.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.4.mlp.c_proj.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.4.mlp.c_proj.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.5.attn.c_attn.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.5.attn.c_attn.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.5.attn.c_proj.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.5.attn.c_proj.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.5.ln_1.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.5.ln_1.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.5.ln_2.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.5.ln_2.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.0.c_fc.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.0.c_fc.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.0.c_proj.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.0.c_proj.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.1.c_fc.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.1.c_fc.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.1.c_proj.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.1.c_proj.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.10.c_fc.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.10.c_fc.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.10.c_proj.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.10.c_proj.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.11.c_fc.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.11.c_fc.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.11.c_proj.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.11.c_proj.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.12.c_fc.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.12.c_fc.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.12.c_proj.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.12.c_proj.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.13.c_fc.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.13.c_fc.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.13.c_proj.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.13.c_proj.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.14.c_fc.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.14.c_fc.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.14.c_proj.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.14.c_proj.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.15.c_fc.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.15.c_fc.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.15.c_proj.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.15.c_proj.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.2.c_fc.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.2.c_fc.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.2.c_proj.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.2.c_proj.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.3.c_fc.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.3.c_fc.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.3.c_proj.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.3.c_proj.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.4.c_fc.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.4.c_fc.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.4.c_proj.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.4.c_proj.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.5.c_fc.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.5.c_fc.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.5.c_proj.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.5.c_proj.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.6.c_fc.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.6.c_fc.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.6.c_proj.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.6.c_proj.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.7.c_fc.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.7.c_fc.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.7.c_proj.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.7.c_proj.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.8.c_fc.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.8.c_fc.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.8.c_proj.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.8.c_proj.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.9.c_fc.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.9.c_fc.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.9.c_proj.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.experts.9.c_proj.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.5.moe.gate.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.6.attn.c_attn.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.6.attn.c_attn.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.6.attn.c_proj.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.6.attn.c_proj.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.6.ln_1.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.6.ln_1.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.6.ln_2.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.6.ln_2.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.6.mlp.c_fc.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.6.mlp.c_fc.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.6.mlp.c_proj.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.6.mlp.c_proj.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.7.attn.c_attn.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.7.attn.c_attn.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.7.attn.c_proj.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.7.attn.c_proj.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.7.ln_1.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.7.ln_1.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.7.ln_2.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.7.ln_2.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.0.c_fc.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.0.c_fc.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.0.c_proj.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.0.c_proj.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.1.c_fc.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.1.c_fc.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.1.c_proj.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.1.c_proj.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.10.c_fc.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.10.c_fc.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.10.c_proj.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.10.c_proj.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.11.c_fc.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.11.c_fc.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.11.c_proj.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.11.c_proj.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.12.c_fc.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.7.moe.experts.12.c_fc.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.7.moe.experts.12.c_proj.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.7.moe.experts.12.c_proj.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.7.moe.experts.13.c_fc.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.7.moe.experts.13.c_fc.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.7.moe.experts.13.c_proj.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.7.moe.experts.13.c_proj.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.7.moe.experts.14.c_fc.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.7.moe.experts.14.c_fc.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.7.moe.experts.14.c_proj.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.7.moe.experts.14.c_proj.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.7.moe.experts.15.c_fc.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.7.moe.experts.15.c_fc.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.7.moe.experts.15.c_proj.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.7.moe.experts.15.c_proj.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.7.moe.experts.2.c_fc.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.2.c_fc.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.2.c_proj.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.2.c_proj.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.3.c_fc.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.3.c_fc.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.3.c_proj.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.3.c_proj.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.4.c_fc.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.4.c_fc.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.4.c_proj.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.4.c_proj.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.5.c_fc.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.5.c_fc.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.5.c_proj.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.5.c_proj.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.6.c_fc.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.6.c_fc.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.6.c_proj.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.6.c_proj.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.7.c_fc.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.7.c_fc.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.7.c_proj.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.7.c_proj.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.8.c_fc.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.8.c_fc.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.8.c_proj.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.8.c_proj.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.9.c_fc.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.9.c_fc.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.9.c_proj.bias": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.experts.9.c_proj.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.7.moe.gate.weight": "model-00002-of-00006.safetensors",
+    "transformer.h.8.attn.c_attn.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.8.attn.c_attn.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.8.attn.c_proj.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.8.attn.c_proj.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.8.ln_1.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.8.ln_1.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.8.ln_2.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.8.ln_2.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.8.mlp.c_fc.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.8.mlp.c_fc.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.8.mlp.c_proj.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.8.mlp.c_proj.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.9.attn.c_attn.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.9.attn.c_attn.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.9.attn.c_proj.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.9.attn.c_proj.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.9.ln_1.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.9.ln_1.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.9.ln_2.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.9.ln_2.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.0.c_fc.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.0.c_fc.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.0.c_proj.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.0.c_proj.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.1.c_fc.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.1.c_fc.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.1.c_proj.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.1.c_proj.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.10.c_fc.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.10.c_fc.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.10.c_proj.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.10.c_proj.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.11.c_fc.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.11.c_fc.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.11.c_proj.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.11.c_proj.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.12.c_fc.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.12.c_fc.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.12.c_proj.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.12.c_proj.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.13.c_fc.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.13.c_fc.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.13.c_proj.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.13.c_proj.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.14.c_fc.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.14.c_fc.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.14.c_proj.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.14.c_proj.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.15.c_fc.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.15.c_fc.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.15.c_proj.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.15.c_proj.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.2.c_fc.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.2.c_fc.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.2.c_proj.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.2.c_proj.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.3.c_fc.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.3.c_fc.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.3.c_proj.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.3.c_proj.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.4.c_fc.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.4.c_fc.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.4.c_proj.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.4.c_proj.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.5.c_fc.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.5.c_fc.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.5.c_proj.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.5.c_proj.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.6.c_fc.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.6.c_fc.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.6.c_proj.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.6.c_proj.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.7.c_fc.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.7.c_fc.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.7.c_proj.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.7.c_proj.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.8.c_fc.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.8.c_fc.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.8.c_proj.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.8.c_proj.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.9.c_fc.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.9.c_fc.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.9.c_proj.bias": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.experts.9.c_proj.weight": "model-00003-of-00006.safetensors",
+    "transformer.h.9.moe.gate.weight": "model-00003-of-00006.safetensors",
+    "transformer.ln_f.bias": "model-00006-of-00006.safetensors",
+    "transformer.ln_f.weight": "model-00006-of-00006.safetensors",
+    "transformer.wpe.weight": "model-00001-of-00006.safetensors",
+    "transformer.wte.weight": "model-00001-of-00006.safetensors"
+  }
+}

modeling_lola_gpt2.py

@@ -0,0 +1,667 @@
+# This script provides an implementation of GPT2 based mixture-of-experts model.
+# Most of its functionality is copied from existing GPT2 implementation on huggingface: https://huggingface.co/docs/transformers/v4.20.1/en/model_doc/gpt2
+# MoE layers are inspired by Mixtral: https://huggingface.co/docs/transformers/v4.39.1/en/model_doc/mixtral
+# There are however, slight differences in this implementation to adapt it to behave like DeepSpeed Megatron's GPT2 MoE: https://github.com/microsoft/Megatron-DeepSpeed/blob/main/examples_deepspeed/MoE/ds_pretrain_gpt_1.3B_MoE128.sh
+# Please note: Most of the the features from DeepSpeed Megatron's GPT MoE are **not** implemented here.
+
+import warnings
+from typing import Optional, Tuple, Union
+
+from .configuration_lola_gpt2 import LOLAConfig
+import torch
+import torch.utils.checkpoint
+from torch import nn
+import torch.nn.functional as F
+from torch.nn import CrossEntropyLoss
+
+from transformers.modeling_outputs import (
+    BaseModelOutputWithPastAndCrossAttentions,
+    SequenceClassifierOutputWithPast,
+    QuestionAnsweringModelOutput
+)
+from transformers.modeling_utils import SequenceSummary
+from transformers.pytorch_utils import Conv1D
+from transformers.utils import (
+    logging
+)
+from transformers.utils.model_parallel_utils import assert_device_map, get_device_map
+
+from transformers.models.gpt2.modeling_gpt2 import GPT2Attention, GPT2MLP, GPT2Block, GPT2PreTrainedModel
+from transformers.models.gpt2.modeling_gpt2 import GPT2LMHeadModel, GPT2DoubleHeadsModel, GPT2ForSequenceClassification, GPT2ForTokenClassification
+
+
+logger = logging.get_logger(__name__)
+
+# LOLA
+class LOLAModel(GPT2PreTrainedModel):
+    
+    config_class = LOLAConfig
+    
+    def __init__(self, config):
+        super().__init__(config)
+
+        self.embed_dim = config.hidden_size
+
+        self.wte = nn.Embedding(config.vocab_size, self.embed_dim)
+        self.wpe = nn.Embedding(config.max_position_embeddings, self.embed_dim)
+
+        self.drop = nn.Dropout(config.embd_pdrop)
+        self.h = nn.ModuleList([
+            GPT2Block(config, layer_idx=i) if i % 2 == 0 else LOLABlock(config, layer_idx=i) for i in range(config.num_hidden_layers)
+        ])
+        self.ln_f = nn.LayerNorm(self.embed_dim, eps=config.layer_norm_epsilon)
+
+        # Model parallel
+        self.model_parallel = False
+        self.device_map = None
+        self.gradient_checkpointing = False
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    
+    def parallelize(self, device_map=None):
+        # Check validity of device_map
+        warnings.warn(
+            "`GPT2Model.parallelize` is deprecated and will be removed in v5 of Transformers, you should load your"
+            " model with `device_map='balanced'` in the call to `from_pretrained`. You can also provide your own"
+            " `device_map` but it needs to be a dictionary module_name to device, so for instance {'h.0': 0, 'h.1': 1,"
+            " ...}",
+            FutureWarning,
+        )
+        self.device_map = (
+            get_device_map(len(self.h), range(torch.cuda.device_count())) if device_map is None else device_map
+        )
+        assert_device_map(self.device_map, len(self.h))
+        self.model_parallel = True
+        self.first_device = "cpu" if "cpu" in self.device_map.keys() else "cuda:" + str(min(self.device_map.keys()))
+        self.last_device = "cuda:" + str(max(self.device_map.keys()))
+        self.wte = self.wte.to(self.first_device)
+        self.wpe = self.wpe.to(self.first_device)
+        # Load onto devices
+        for k, v in self.device_map.items():
+            for block in v:
+                cuda_device = "cuda:" + str(k)
+                self.h[block] = self.h[block].to(cuda_device)
+        # ln_f to last
+        self.ln_f = self.ln_f.to(self.last_device)
+
+    
+    def deparallelize(self):
+        warnings.warn(
+            "Like `parallelize`, `deparallelize` is deprecated and will be removed in v5 of Transformers.",
+            FutureWarning,
+        )
+        self.model_parallel = False
+        self.device_map = None
+        self.first_device = "cpu"
+        self.last_device = "cpu"
+        self.wte = self.wte.to("cpu")
+        self.wpe = self.wpe.to("cpu")
+        for index in range(len(self.h)):
+            self.h[index] = self.h[index].to("cpu")
+        self.ln_f = self.ln_f.to("cpu")
+        torch.cuda.empty_cache()
+
+    def get_input_embeddings(self):
+        return self.wte
+
+    def set_input_embeddings(self, new_embeddings):
+        self.wte = new_embeddings
+
+    def _prune_heads(self, heads_to_prune):
+        """
+        Prunes heads of the model. heads_to_prune: dict of {layer_num: list of heads to prune in this layer}
+        """
+        for layer, heads in heads_to_prune.items():
+            self.h[layer].attn.prune_heads(heads)
+
+    def forward(
+        self,
+        input_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[Tuple[Tuple[torch.Tensor]]] = None,
+        attention_mask: Optional[torch.FloatTensor] = None,
+        token_type_ids: Optional[torch.LongTensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        head_mask: Optional[torch.FloatTensor] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        encoder_hidden_states: Optional[torch.Tensor] = None,
+        encoder_attention_mask: Optional[torch.FloatTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[Tuple, BaseModelOutputWithPastAndCrossAttentions]:
+        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
+        )
+        use_cache = use_cache if use_cache is not None else self.config.use_cache
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        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:
+            # self.warn_if_padding_and_no_attention_mask(input_ids, attention_mask)
+            input_shape = input_ids.size()
+            input_ids = input_ids.view(-1, input_shape[-1])
+            batch_size = input_ids.shape[0]
+        elif inputs_embeds is not None:
+            input_shape = inputs_embeds.size()[:-1]
+            batch_size = inputs_embeds.shape[0]
+        else:
+            raise ValueError("You have to specify either input_ids or inputs_embeds")
+
+        device = input_ids.device if input_ids is not None else inputs_embeds.device
+
+        if token_type_ids is not None:
+            token_type_ids = token_type_ids.view(-1, input_shape[-1])
+
+        if past_key_values is None:
+            past_length = 0
+            past_key_values = tuple([None] * len(self.h))
+        else:
+            past_length = past_key_values[0][0].size(-2)
+        if position_ids is None:
+            position_ids = torch.arange(past_length, input_shape[-1] + past_length, dtype=torch.long, device=device)
+            position_ids = position_ids.unsqueeze(0)
+
+        # GPT2Attention mask.
+        if attention_mask is not None:
+            if batch_size <= 0:
+                raise ValueError("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
+
+        # If a 2D or 3D attention mask is provided for the cross-attention
+        # we need to make broadcastable to [batch_size, num_heads, seq_length, seq_length]
+        if self.config.add_cross_attention and encoder_hidden_states is not None:
+            encoder_batch_size, encoder_sequence_length, _ = encoder_hidden_states.size()
+            encoder_hidden_shape = (encoder_batch_size, encoder_sequence_length)
+            if encoder_attention_mask is None:
+                encoder_attention_mask = torch.ones(encoder_hidden_shape, device=device)
+            encoder_attention_mask = self.invert_attention_mask(encoder_attention_mask)
+        else:
+            encoder_attention_mask = None
+
+        # 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
+        # head_mask has shape n_layer x batch x n_heads x N x N
+        head_mask = self.get_head_mask(head_mask, self.config.n_layer)
+
+        if inputs_embeds is None:
+            inputs_embeds = self.wte(input_ids)
+        position_embeds = self.wpe(position_ids)
+        hidden_states = inputs_embeds + position_embeds
+
+        if token_type_ids is not None:
+            token_type_embeds = self.wte(token_type_ids)
+            hidden_states = hidden_states + token_type_embeds
+
+        hidden_states = self.drop(hidden_states)
+
+        output_shape = (-1,) + input_shape[1:] + (hidden_states.size(-1),)
+
+        if self.gradient_checkpointing and self.training:
+            if use_cache:
+                logger.warning_once(
+                    "`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`..."
+                )
+                use_cache = False
+
+        presents = () if use_cache else None
+        all_self_attentions = () if output_attentions else None
+        all_cross_attentions = () if output_attentions and self.config.add_cross_attention else None
+        all_hidden_states = () if output_hidden_states else None
+        for i, (block, layer_past) in enumerate(zip(self.h, past_key_values)):
+            # Model parallel
+            if self.model_parallel:
+                torch.cuda.set_device(hidden_states.device)
+                # Ensure layer_past is on same device as hidden_states (might not be correct)
+                if layer_past is not None:
+                    layer_past = tuple(past_state.to(hidden_states.device) for past_state in layer_past)
+                # Ensure that attention_mask is always on the same device as hidden_states
+                if attention_mask is not None:
+                    attention_mask = attention_mask.to(hidden_states.device)
+                if isinstance(head_mask, torch.Tensor):
+                    head_mask = head_mask.to(hidden_states.device)
+            if output_hidden_states:
+                all_hidden_states = all_hidden_states + (hidden_states,)
+
+            if self.gradient_checkpointing and self.training:
+                outputs = self._gradient_checkpointing_func(
+                    block.__call__,
+                    hidden_states,
+                    None,
+                    attention_mask,
+                    head_mask[i],
+                    encoder_hidden_states,
+                    encoder_attention_mask,
+                    use_cache,
+                    output_attentions,
+                )
+            else:
+                outputs = block(
+                    hidden_states,
+                    layer_past=layer_past,
+                    attention_mask=attention_mask,
+                    head_mask=head_mask[i],
+                    encoder_hidden_states=encoder_hidden_states,
+                    encoder_attention_mask=encoder_attention_mask,
+                    use_cache=use_cache,
+                    output_attentions=output_attentions,
+                )
+
+            hidden_states = outputs[0]
+            if use_cache is True:
+                presents = presents + (outputs[1],)
+
+            if output_attentions:
+                all_self_attentions = all_self_attentions + (outputs[2 if use_cache else 1],)
+                if self.config.add_cross_attention:
+                    all_cross_attentions = all_cross_attentions + (outputs[3 if use_cache else 2],)
+
+            # Model Parallel: If it's the last layer for that device, put things on the next device
+            if self.model_parallel:
+                for k, v in self.device_map.items():
+                    if i == v[-1] and "cuda:" + str(k) != self.last_device:
+                        hidden_states = hidden_states.to("cuda:" + str(k + 1))
+
+        hidden_states = self.ln_f(hidden_states)
+
+        hidden_states = hidden_states.view(output_shape)
+        # 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_self_attentions, all_cross_attentions]
+                if v is not None
+            )
+
+        return BaseModelOutputWithPastAndCrossAttentions(
+            last_hidden_state=hidden_states,
+            past_key_values=presents,
+            hidden_states=all_hidden_states,
+            attentions=all_self_attentions,
+            cross_attentions=all_cross_attentions,
+        )
+
+class LOLABlock(nn.Module):
+    def __init__(self, config, layer_idx=None):
+        super().__init__()
+        hidden_size = config.hidden_size
+        inner_dim = config.n_inner if config.n_inner is not None else 4 * hidden_size
+
+        self.ln_1 = nn.LayerNorm(hidden_size, eps=config.layer_norm_epsilon)
+        self.attn = GPT2Attention(config, layer_idx=layer_idx)
+        self.ln_2 = nn.LayerNorm(hidden_size, eps=config.layer_norm_epsilon)
+
+        self.moe = LOLAMOE(
+            hidden_size,
+            inner_dim,
+            config,
+            config.num_experts,
+            k=config.topk,
+            # capacity_factor=1.0,
+            # min_capacity=4,
+            # drop_tokens=False,
+            # use_tutel=False,
+            # enable_expert_tensor_parallelism=False,
+        )
+
+    def forward(
+        self,
+        hidden_states: Optional[Tuple[torch.FloatTensor]],
+        layer_past: Optional[Tuple[torch.Tensor]] = None,
+        attention_mask: Optional[torch.FloatTensor] = None,
+        head_mask: Optional[torch.FloatTensor] = None,
+        encoder_hidden_states: Optional[torch.Tensor] = None,
+        encoder_attention_mask: Optional[torch.FloatTensor] = None,
+        use_cache: Optional[bool] = False,
+        output_attentions: Optional[bool] = False,
+    ) -> Union[Tuple[torch.Tensor], Optional[Tuple[torch.Tensor, Tuple[torch.FloatTensor, ...]]]]:
+        residual = hidden_states
+        hidden_states = self.ln_1(hidden_states)
+        attn_outputs = self.attn(
+            hidden_states,
+            layer_past=layer_past,
+            attention_mask=attention_mask,
+            head_mask=head_mask,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+        )
+        attn_output = attn_outputs[0]  # output_attn: a, present, (attentions)
+        outputs = attn_outputs[1:]
+        # residual connection
+        hidden_states = attn_output + residual
+
+        if encoder_hidden_states is not None:
+            # add one self-attention block for cross-attention
+            if not hasattr(self, "crossattention"):
+                raise ValueError(
+                    f"If `encoder_hidden_states` are passed, {self} has to be instantiated with "
+                    "cross-attention layers by setting `config.add_cross_attention=True`"
+                )
+            residual = hidden_states
+            hidden_states = self.ln_cross_attn(hidden_states)
+            cross_attn_outputs = self.crossattention(
+                hidden_states,
+                attention_mask=attention_mask,
+                head_mask=head_mask,
+                encoder_hidden_states=encoder_hidden_states,
+                encoder_attention_mask=encoder_attention_mask,
+                output_attentions=output_attentions,
+            )
+            attn_output = cross_attn_outputs[0]
+            # residual connection
+            hidden_states = residual + attn_output
+            outputs = outputs + cross_attn_outputs[2:]  # add cross attentions if we output attention weights
+
+        residual = hidden_states
+        hidden_states = self.ln_2(hidden_states)
+        feed_forward_hidden_states, _ = self.moe(hidden_states)
+        # residual connection
+        hidden_states = residual + feed_forward_hidden_states
+
+        if use_cache:
+            outputs = (hidden_states,) + outputs
+        else:
+            outputs = (hidden_states,) + outputs[1:]
+
+        return outputs  # hidden_states, present, (attentions, cross_attentions)
+
+class LOLAMOE(nn.Module):
+    def __init__(self,
+                 hidden_size,
+                 inner_dim,
+                 config,
+                 num_experts,
+                 k
+                 ):
+        super().__init__()
+        self.hidden_dim = hidden_size
+        self.num_experts = num_experts
+        self.top_k = k
+
+        self.gate = nn.Linear(self.hidden_dim, self.num_experts, bias=False)
+        self.experts = nn.ModuleList([GPT2MLP(inner_dim, config) for _ in range(self.num_experts)])
+
+    def forward(self, hidden_states):
+        # https://github.com/huggingface/transformers/blob/main/src/transformers/models/mixtral/modeling_mixtral.py#L816
+        # FIXME do it as in top1gating
+        # https://github.com/microsoft/DeepSpeed/blob/master/deepspeed/moe/sharded_moe.py
+
+        batch_size, sequence_length, hidden_dim = hidden_states.shape
+        hidden_states = hidden_states.view(-1, hidden_dim)
+
+        router_logits = self.gate(hidden_states)
+        # router_logits = router_logits.squeeze(dim=0)
+
+        # TODO: fix the weights logic to be the same as Megatron
+        routing_weights = F.softmax(router_logits, dim=1)
+        routing_weights, selected_experts = torch.topk(routing_weights, self.top_k, dim=-1)
+        # routing_weights /= routing_weights.sum(dim=-1, keepdim=True)
+        # commenting the statement above for LOLA and removing the "/" operator to avoid getting weights as 1
+        routing_weights = routing_weights.sum(dim=-1, keepdim=True)
+        routing_weights = routing_weights.to(hidden_states.dtype)
+
+        final_hidden_states = torch.zeros(
+            (batch_size * sequence_length, hidden_dim), dtype=hidden_states.dtype, device=hidden_states.device
+        )
+        expert_mask = torch.nn.functional.one_hot(selected_experts, num_classes=self.num_experts).permute(2, 1, 0)
+        for expert_idx in range(self.num_experts):
+            expert_layer = self.experts[expert_idx]
+            idx, top_x = torch.where(expert_mask[expert_idx])
+
+            if top_x.shape[0] == 0:
+                continue
+
+            # in torch it is faster to index using lists than torch tensors
+            top_x_list = top_x.tolist()
+            idx_list = idx.tolist()
+
+            # Index the correct hidden states and compute the expert hidden state for
+            # the current expert. We need to make sure to multiply the output hidden
+            # states by `routing_weights` on the corresponding tokens (top-1 and top-2)
+            current_state = hidden_states[None, top_x_list].reshape(-1, hidden_dim)
+            current_hidden_states = expert_layer(current_state) * routing_weights[top_x_list, idx_list, None]
+
+            # However `index_add_` only support torch tensors for indexing so we'll use
+            # the `top_x` tensor here.
+            final_hidden_states.index_add_(0, top_x, current_hidden_states.to(hidden_states.dtype))
+        final_hidden_states = final_hidden_states.reshape(batch_size, sequence_length, hidden_dim)
+        return final_hidden_states, router_logits
+
+class LOLAAttention(GPT2Attention):
+    def __init__(self, config, is_cross_attention=False, layer_idx=None):
+        super(GPT2Attention, SequenceClassifierOutputWithPast).__init__()
+
+        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
+            ),
+            #persistent=False,
+        )
+        self.register_buffer("masked_bias", torch.tensor(-1e4), 
+                             #persistent=False
+                             )
+
+        self.embed_dim = config.hidden_size
+        self.num_heads = config.num_attention_heads
+        self.head_dim = self.embed_dim // self.num_heads
+        self.split_size = self.embed_dim
+        if self.head_dim * self.num_heads != self.embed_dim:
+            raise ValueError(
+                f"`embed_dim` must be divisible by num_heads (got `embed_dim`: {self.embed_dim} and `num_heads`:"
+                f" {self.num_heads})."
+            )
+
+        self.scale_attn_weights = config.scale_attn_weights
+        self.is_cross_attention = is_cross_attention
+
+        # Layer-wise attention scaling, reordering, and upcasting
+        self.scale_attn_by_inverse_layer_idx = config.scale_attn_by_inverse_layer_idx
+        self.layer_idx = layer_idx
+        self.reorder_and_upcast_attn = config.reorder_and_upcast_attn
+
+        if self.is_cross_attention:
+            self.c_attn = Conv1D(2 * self.embed_dim, self.embed_dim)
+            self.q_attn = Conv1D(self.embed_dim, self.embed_dim)
+        else:
+            self.c_attn = Conv1D(3 * self.embed_dim, self.embed_dim)
+        self.c_proj = Conv1D(self.embed_dim, self.embed_dim)
+
+        self.attn_dropout = nn.Dropout(config.attn_pdrop)
+        self.resid_dropout = nn.Dropout(config.resid_pdrop)
+
+        self.pruned_heads = set()
+
+
+class LOLALMHeadModel(GPT2LMHeadModel):
+    
+    config_class = LOLAConfig
+
+    def __init__(self, config):
+        # preventing initiation of GPT2LMHeadModel directly
+        super(GPT2LMHeadModel, self).__init__(config)
+        self.transformer = LOLAModel(config)
+        self.lm_head = nn.Linear(config.n_embd, config.vocab_size, bias=False)
+
+        # Model parallel
+        self.model_parallel = False
+        self.device_map = None
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+
+class LOLADoubleHeadsModel(GPT2DoubleHeadsModel):
+    
+    config_class = LOLAConfig
+
+    def __init__(self, config):
+        super(GPT2DoubleHeadsModel, self).__init__(config)
+        config.num_labels = 1
+        self.transformer = LOLAModel(config)
+        self.lm_head = nn.Linear(config.n_embd, config.vocab_size, bias=False)
+        self.multiple_choice_head = SequenceSummary(config)
+
+        # Model parallel
+        self.model_parallel = False
+        self.device_map = None
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+      
+class LOLAForSequenceClassification(GPT2ForSequenceClassification):
+    
+    config_class = LOLAConfig
+    
+    def __init__(self, config):
+        super(GPT2ForSequenceClassification, self).__init__(config)
+        self.num_labels = config.num_labels
+        self.transformer = LOLAModel(config)
+        self.score = nn.Linear(config.n_embd, self.num_labels, bias=False)
+
+        # Model parallel
+        self.model_parallel = False
+        self.device_map = None
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+class LOLAForTokenClassification(GPT2ForTokenClassification):
+    
+    config_class = LOLAConfig
+    
+    def __init__(self, config):
+        super(GPT2ForTokenClassification, self).__init__(config)
+        self.num_labels = config.num_labels
+
+        self.transformer = LOLAModel(config)
+        if hasattr(config, "classifier_dropout") and config.classifier_dropout is not None:
+            classifier_dropout = config.classifier_dropout
+        elif hasattr(config, "hidden_dropout") and config.hidden_dropout is not None:
+            classifier_dropout = config.hidden_dropout
+        else:
+            classifier_dropout = 0.1
+        self.dropout = nn.Dropout(classifier_dropout)
+        self.classifier = nn.Linear(config.hidden_size, config.num_labels)
+
+        # Model parallel
+        self.model_parallel = False
+        self.device_map = None
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+class LOLAForQuestionAnswering(GPT2PreTrainedModel):
+    
+    config_class = LOLAConfig
+    
+    def __init__(self, config):
+        super().__init__(config)
+        self.num_labels = config.num_labels
+        self.transformer = LOLAModel(config)
+        self.qa_outputs = nn.Linear(config.hidden_size, 2)
+
+        # Model parallel
+        self.model_parallel = False
+        self.device_map = None
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def forward(
+        self,
+        input_ids: Optional[torch.LongTensor] = None,
+        attention_mask: Optional[torch.FloatTensor] = None,
+        token_type_ids: Optional[torch.LongTensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        head_mask: Optional[torch.FloatTensor] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        start_positions: Optional[torch.LongTensor] = None,
+        end_positions: Optional[torch.LongTensor] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[Tuple, QuestionAnsweringModelOutput]:
+        r"""
+        start_positions (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
+            Labels for position (index) of the start of the labelled span for computing the token classification loss.
+            Positions are clamped to the length of the sequence (`sequence_length`). Position outside of the sequence
+            are not taken into account for computing the loss.
+        end_positions (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
+            Labels for position (index) of the end of the labelled span for computing the token classification loss.
+            Positions are clamped to the length of the sequence (`sequence_length`). Position outside of the sequence
+            are not taken into account for computing the loss.
+        """
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        outputs = self.transformer(
+            input_ids,
+            attention_mask=attention_mask,
+            token_type_ids=token_type_ids,
+            position_ids=position_ids,
+            head_mask=head_mask,
+            inputs_embeds=inputs_embeds,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+
+        sequence_output = outputs[0]
+
+        logits = self.qa_outputs(sequence_output)
+        start_logits, end_logits = logits.split(1, dim=-1)
+        start_logits = start_logits.squeeze(-1).contiguous()
+        end_logits = end_logits.squeeze(-1).contiguous()
+
+        total_loss = None
+        if start_positions is not None and end_positions is not None:
+            # If we are on multi-GPU, split add a dimension
+            if len(start_positions.size()) > 1:
+                start_positions = start_positions.squeeze(-1).to(start_logits.device)
+            if len(end_positions.size()) > 1:
+                end_positions = end_positions.squeeze(-1).to(end_logits.device)
+            # sometimes the start/end positions are outside our model inputs, we ignore these terms
+            ignored_index = start_logits.size(1)
+            start_positions = start_positions.clamp(0, ignored_index)
+            end_positions = end_positions.clamp(0, ignored_index)
+
+            loss_fct = CrossEntropyLoss(ignore_index=ignored_index)
+            start_loss = loss_fct(start_logits, start_positions)
+            end_loss = loss_fct(end_logits, end_positions)
+            total_loss = (start_loss + end_loss) / 2
+
+        if not return_dict:
+            output = (start_logits, end_logits) + outputs[2:]
+            return ((total_loss,) + output) if total_loss is not None else output
+
+        return QuestionAnsweringModelOutput(
+            loss=total_loss,
+            start_logits=start_logits,
+            end_logits=end_logits,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )