AWQ model commit

config.json

@@ -0,0 +1,43 @@
+{
+  "_name_or_path": "/workspace/process/lightonai_alfred-40b-1023/source",
+  "alibi": false,
+  "apply_residual_connection_post_layernorm": false,
+  "architectures": [
+    "RWForCausalLM"
+  ],
+  "attention_dropout": 0.0,
+  "auto_map": {
+    "AutoConfig": "configuration_RW.RWConfig",
+    "AutoModelForCausalLM": "modeling_RW.RWForCausalLM"
+  },
+  "bias": false,
+  "bos_token_id": 11,
+  "embedding_scaling_factor": 4,
+  "eos_token_id": 11,
+  "hidden_dropout": 0.0,
+  "hidden_size": 8192,
+  "initializer_range": 0.02,
+  "layer_norm_epsilon": 1e-05,
+  "model_type": "RefinedWeb",
+  "multi_query": true,
+  "n_head": 128,
+  "n_head_kv": 8,
+  "n_layer": 60,
+  "ntk_scaling_factor": 5,
+  "pad_token_id": 0,
+  "parallel_attn": true,
+  "pretraining_tp": 1,
+  "quantization_config": {
+    "bits": 4,
+    "group_size": 128,
+    "quant_method": "awq",
+    "version": "gemm",
+    "zero_point": true
+  },
+  "single_ln": false,
+  "torch_dtype": "float16",
+  "transformers_version": "4.35.0",
+  "use_cache": true,
+  "vanilla_scaling_factor": null,
+  "vocab_size": 65024
+}

configuration_RW.py

@@ -0,0 +1,92 @@
+# coding=utf-8
+# Copyright 2022 the Big Science Workshop and HuggingFace Inc. team.  All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+""" Bloom configuration"""
+from transformers.configuration_utils import PretrainedConfig
+from transformers.utils import logging
+
+
+logger = logging.get_logger(__name__)
+
+
+class RWConfig(PretrainedConfig):
+    model_type = "RefinedWeb"
+    keys_to_ignore_at_inference = ["past_key_values"]
+    attribute_map = {
+        "num_hidden_layers": "n_layer",
+        "num_attention_heads": "n_head",
+    }
+
+    def __init__(
+        self,
+        vocab_size=250880,
+        hidden_size=64,
+        n_layer=2,
+        n_head=8,
+        layer_norm_epsilon=1e-5,
+        initializer_range=0.02,
+        use_cache=True,
+        bos_token_id=1,
+        eos_token_id=2,
+        apply_residual_connection_post_layernorm=False,
+        hidden_dropout=0.0,
+        attention_dropout=0.0,
+        multi_query=False,
+        alibi=False,
+        bias=False,
+        parallel_attn=False,
+        single_ln=False,
+        n_head_kv=1,
+        ntk_scaling_factor=None,
+        vanilla_scaling_factor=None,
+        embedding_scaling_factor=None,
+        **kwargs,
+    ):
+        self.vocab_size = vocab_size
+        # Backward compatibility with n_embed kwarg
+        n_embed = kwargs.pop("n_embed", None)
+        self.hidden_size = hidden_size if n_embed is None else n_embed
+        self.n_layer = n_layer
+        self.n_head = n_head
+        self.layer_norm_epsilon = layer_norm_epsilon
+        self.initializer_range = initializer_range
+        self.use_cache = use_cache
+        self.apply_residual_connection_post_layernorm = apply_residual_connection_post_layernorm
+        self.hidden_dropout = hidden_dropout
+        self.attention_dropout = attention_dropout
+
+        self.bos_token_id = bos_token_id
+        self.eos_token_id = eos_token_id
+        self.multi_query = multi_query
+        self.alibi = alibi
+        self.bias = bias
+        self.parallel_attn = parallel_attn
+        self.single_ln = single_ln
+        self.n_head_kv = n_head_kv
+        self.ntk_scaling_factor = ntk_scaling_factor
+        self.vanilla_scaling_factor = vanilla_scaling_factor
+        self.embedding_scaling_factor = embedding_scaling_factor
+
+        assert not alibi, "Function of alibi has not been verified yet"
+        assert self.vanilla_scaling_factor is None or self.ntk_scaling_factor is None, "Both scaling modes cannot be used concurrently" 
+
+        super().__init__(bos_token_id=bos_token_id, eos_token_id=eos_token_id, **kwargs)
+
+    @property
+    def head_dim(self):
+        return self.hidden_size // self.n_head
+
+    @property
+    def rotary(self):
+        return not self.alibi

generation_config.json

@@ -0,0 +1,6 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 11,
+  "eos_token_id": 11,
+  "transformers_version": "4.31.0"
+}

model-00001-of-00003.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a031e11e2a344f8c3c0bac086a2ea2ee3ccccc7e6575e51e44d8028cb26e64e2
+size 11031788568

model-00002-of-00003.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:354b62cc30cb9336ceda42e7595ebe63b51ff75ec057b488fd72c4a4cc047693
+size 9886158176

model-00003-of-00003.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a2406bae21592ebf7cbb290577dbbc8665a1db549a5178a30c18e3ae12cc292
+size 2397417352

model.safetensors.index.json

@@ -0,0 +1,971 @@
+{
+    "metadata": {
+        "total_size": 22249897984
+    },
+    "weight_map": {
+        "transformer.word_embeddings.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.0.ln_attn.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.0.ln_attn.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.0.self_attention.query_key_value.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.0.self_attention.query_key_value.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.0.self_attention.query_key_value.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.0.self_attention.dense.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.0.self_attention.dense.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.0.self_attention.dense.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.0.ln_mlp.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.0.ln_mlp.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.0.mlp.dense_h_to_4h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.0.mlp.dense_h_to_4h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.0.mlp.dense_h_to_4h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.0.mlp.dense_4h_to_h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.0.mlp.dense_4h_to_h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.0.mlp.dense_4h_to_h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.1.ln_attn.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.1.ln_attn.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.1.self_attention.query_key_value.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.1.self_attention.query_key_value.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.1.self_attention.query_key_value.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.1.self_attention.dense.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.1.self_attention.dense.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.1.self_attention.dense.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.1.ln_mlp.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.1.ln_mlp.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.1.mlp.dense_h_to_4h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.1.mlp.dense_h_to_4h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.1.mlp.dense_h_to_4h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.1.mlp.dense_4h_to_h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.1.mlp.dense_4h_to_h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.1.mlp.dense_4h_to_h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.2.ln_attn.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.2.ln_attn.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.2.self_attention.query_key_value.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.2.self_attention.query_key_value.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.2.self_attention.query_key_value.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.2.self_attention.dense.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.2.self_attention.dense.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.2.self_attention.dense.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.2.ln_mlp.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.2.ln_mlp.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.2.mlp.dense_h_to_4h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.2.mlp.dense_h_to_4h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.2.mlp.dense_h_to_4h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.2.mlp.dense_4h_to_h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.2.mlp.dense_4h_to_h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.2.mlp.dense_4h_to_h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.3.ln_attn.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.3.ln_attn.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.3.self_attention.query_key_value.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.3.self_attention.query_key_value.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.3.self_attention.query_key_value.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.3.self_attention.dense.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.3.self_attention.dense.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.3.self_attention.dense.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.3.ln_mlp.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.3.ln_mlp.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.3.mlp.dense_h_to_4h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.3.mlp.dense_h_to_4h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.3.mlp.dense_h_to_4h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.3.mlp.dense_4h_to_h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.3.mlp.dense_4h_to_h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.3.mlp.dense_4h_to_h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.4.ln_attn.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.4.ln_attn.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.4.self_attention.query_key_value.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.4.self_attention.query_key_value.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.4.self_attention.query_key_value.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.4.self_attention.dense.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.4.self_attention.dense.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.4.self_attention.dense.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.4.ln_mlp.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.4.ln_mlp.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.4.mlp.dense_h_to_4h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.4.mlp.dense_h_to_4h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.4.mlp.dense_h_to_4h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.4.mlp.dense_4h_to_h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.4.mlp.dense_4h_to_h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.4.mlp.dense_4h_to_h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.5.ln_attn.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.5.ln_attn.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.5.self_attention.query_key_value.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.5.self_attention.query_key_value.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.5.self_attention.query_key_value.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.5.self_attention.dense.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.5.self_attention.dense.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.5.self_attention.dense.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.5.ln_mlp.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.5.ln_mlp.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.5.mlp.dense_h_to_4h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.5.mlp.dense_h_to_4h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.5.mlp.dense_h_to_4h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.5.mlp.dense_4h_to_h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.5.mlp.dense_4h_to_h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.5.mlp.dense_4h_to_h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.6.ln_attn.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.6.ln_attn.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.6.self_attention.query_key_value.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.6.self_attention.query_key_value.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.6.self_attention.query_key_value.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.6.self_attention.dense.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.6.self_attention.dense.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.6.self_attention.dense.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.6.ln_mlp.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.6.ln_mlp.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.6.mlp.dense_h_to_4h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.6.mlp.dense_h_to_4h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.6.mlp.dense_h_to_4h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.6.mlp.dense_4h_to_h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.6.mlp.dense_4h_to_h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.6.mlp.dense_4h_to_h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.7.ln_attn.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.7.ln_attn.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.7.self_attention.query_key_value.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.7.self_attention.query_key_value.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.7.self_attention.query_key_value.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.7.self_attention.dense.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.7.self_attention.dense.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.7.self_attention.dense.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.7.ln_mlp.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.7.ln_mlp.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.7.mlp.dense_h_to_4h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.7.mlp.dense_h_to_4h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.7.mlp.dense_h_to_4h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.7.mlp.dense_4h_to_h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.7.mlp.dense_4h_to_h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.7.mlp.dense_4h_to_h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.8.ln_attn.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.8.ln_attn.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.8.self_attention.query_key_value.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.8.self_attention.query_key_value.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.8.self_attention.query_key_value.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.8.self_attention.dense.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.8.self_attention.dense.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.8.self_attention.dense.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.8.ln_mlp.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.8.ln_mlp.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.8.mlp.dense_h_to_4h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.8.mlp.dense_h_to_4h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.8.mlp.dense_h_to_4h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.8.mlp.dense_4h_to_h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.8.mlp.dense_4h_to_h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.8.mlp.dense_4h_to_h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.9.ln_attn.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.9.ln_attn.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.9.self_attention.query_key_value.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.9.self_attention.query_key_value.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.9.self_attention.query_key_value.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.9.self_attention.dense.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.9.self_attention.dense.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.9.self_attention.dense.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.9.ln_mlp.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.9.ln_mlp.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.9.mlp.dense_h_to_4h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.9.mlp.dense_h_to_4h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.9.mlp.dense_h_to_4h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.9.mlp.dense_4h_to_h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.9.mlp.dense_4h_to_h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.9.mlp.dense_4h_to_h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.10.ln_attn.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.10.ln_attn.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.10.self_attention.query_key_value.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.10.self_attention.query_key_value.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.10.self_attention.query_key_value.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.10.self_attention.dense.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.10.self_attention.dense.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.10.self_attention.dense.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.10.ln_mlp.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.10.ln_mlp.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.10.mlp.dense_h_to_4h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.10.mlp.dense_h_to_4h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.10.mlp.dense_h_to_4h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.10.mlp.dense_4h_to_h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.10.mlp.dense_4h_to_h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.10.mlp.dense_4h_to_h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.11.ln_attn.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.11.ln_attn.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.11.self_attention.query_key_value.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.11.self_attention.query_key_value.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.11.self_attention.query_key_value.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.11.self_attention.dense.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.11.self_attention.dense.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.11.self_attention.dense.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.11.ln_mlp.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.11.ln_mlp.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.11.mlp.dense_h_to_4h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.11.mlp.dense_h_to_4h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.11.mlp.dense_h_to_4h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.11.mlp.dense_4h_to_h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.11.mlp.dense_4h_to_h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.11.mlp.dense_4h_to_h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.12.ln_attn.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.12.ln_attn.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.12.self_attention.query_key_value.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.12.self_attention.query_key_value.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.12.self_attention.query_key_value.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.12.self_attention.dense.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.12.self_attention.dense.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.12.self_attention.dense.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.12.ln_mlp.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.12.ln_mlp.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.12.mlp.dense_h_to_4h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.12.mlp.dense_h_to_4h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.12.mlp.dense_h_to_4h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.12.mlp.dense_4h_to_h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.12.mlp.dense_4h_to_h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.12.mlp.dense_4h_to_h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.13.ln_attn.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.13.ln_attn.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.13.self_attention.query_key_value.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.13.self_attention.query_key_value.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.13.self_attention.query_key_value.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.13.self_attention.dense.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.13.self_attention.dense.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.13.self_attention.dense.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.13.ln_mlp.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.13.ln_mlp.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.13.mlp.dense_h_to_4h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.13.mlp.dense_h_to_4h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.13.mlp.dense_h_to_4h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.13.mlp.dense_4h_to_h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.13.mlp.dense_4h_to_h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.13.mlp.dense_4h_to_h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.14.ln_attn.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.14.ln_attn.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.14.self_attention.query_key_value.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.14.self_attention.query_key_value.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.14.self_attention.query_key_value.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.14.self_attention.dense.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.14.self_attention.dense.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.14.self_attention.dense.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.14.ln_mlp.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.14.ln_mlp.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.14.mlp.dense_h_to_4h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.14.mlp.dense_h_to_4h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.14.mlp.dense_h_to_4h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.14.mlp.dense_4h_to_h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.14.mlp.dense_4h_to_h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.14.mlp.dense_4h_to_h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.15.ln_attn.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.15.ln_attn.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.15.self_attention.query_key_value.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.15.self_attention.query_key_value.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.15.self_attention.query_key_value.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.15.self_attention.dense.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.15.self_attention.dense.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.15.self_attention.dense.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.15.ln_mlp.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.15.ln_mlp.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.15.mlp.dense_h_to_4h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.15.mlp.dense_h_to_4h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.15.mlp.dense_h_to_4h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.15.mlp.dense_4h_to_h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.15.mlp.dense_4h_to_h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.15.mlp.dense_4h_to_h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.16.ln_attn.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.16.ln_attn.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.16.self_attention.query_key_value.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.16.self_attention.query_key_value.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.16.self_attention.query_key_value.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.16.self_attention.dense.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.16.self_attention.dense.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.16.self_attention.dense.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.16.ln_mlp.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.16.ln_mlp.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.16.mlp.dense_h_to_4h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.16.mlp.dense_h_to_4h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.16.mlp.dense_h_to_4h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.16.mlp.dense_4h_to_h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.16.mlp.dense_4h_to_h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.16.mlp.dense_4h_to_h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.17.ln_attn.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.17.ln_attn.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.17.self_attention.query_key_value.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.17.self_attention.query_key_value.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.17.self_attention.query_key_value.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.17.self_attention.dense.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.17.self_attention.dense.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.17.self_attention.dense.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.17.ln_mlp.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.17.ln_mlp.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.17.mlp.dense_h_to_4h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.17.mlp.dense_h_to_4h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.17.mlp.dense_h_to_4h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.17.mlp.dense_4h_to_h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.17.mlp.dense_4h_to_h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.17.mlp.dense_4h_to_h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.18.ln_attn.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.18.ln_attn.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.18.self_attention.query_key_value.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.18.self_attention.query_key_value.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.18.self_attention.query_key_value.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.18.self_attention.dense.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.18.self_attention.dense.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.18.self_attention.dense.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.18.ln_mlp.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.18.ln_mlp.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.18.mlp.dense_h_to_4h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.18.mlp.dense_h_to_4h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.18.mlp.dense_h_to_4h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.18.mlp.dense_4h_to_h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.18.mlp.dense_4h_to_h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.18.mlp.dense_4h_to_h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.19.ln_attn.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.19.ln_attn.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.19.self_attention.query_key_value.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.19.self_attention.query_key_value.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.19.self_attention.query_key_value.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.19.self_attention.dense.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.19.self_attention.dense.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.19.self_attention.dense.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.19.ln_mlp.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.19.ln_mlp.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.19.mlp.dense_h_to_4h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.19.mlp.dense_h_to_4h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.19.mlp.dense_h_to_4h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.19.mlp.dense_4h_to_h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.19.mlp.dense_4h_to_h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.19.mlp.dense_4h_to_h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.20.ln_attn.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.20.ln_attn.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.20.self_attention.query_key_value.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.20.self_attention.query_key_value.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.20.self_attention.query_key_value.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.20.self_attention.dense.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.20.self_attention.dense.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.20.self_attention.dense.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.20.ln_mlp.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.20.ln_mlp.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.20.mlp.dense_h_to_4h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.20.mlp.dense_h_to_4h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.20.mlp.dense_h_to_4h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.20.mlp.dense_4h_to_h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.20.mlp.dense_4h_to_h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.20.mlp.dense_4h_to_h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.21.ln_attn.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.21.ln_attn.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.21.self_attention.query_key_value.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.21.self_attention.query_key_value.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.21.self_attention.query_key_value.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.21.self_attention.dense.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.21.self_attention.dense.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.21.self_attention.dense.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.21.ln_mlp.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.21.ln_mlp.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.21.mlp.dense_h_to_4h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.21.mlp.dense_h_to_4h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.21.mlp.dense_h_to_4h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.21.mlp.dense_4h_to_h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.21.mlp.dense_4h_to_h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.21.mlp.dense_4h_to_h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.22.ln_attn.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.22.ln_attn.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.22.self_attention.query_key_value.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.22.self_attention.query_key_value.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.22.self_attention.query_key_value.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.22.self_attention.dense.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.22.self_attention.dense.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.22.self_attention.dense.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.22.ln_mlp.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.22.ln_mlp.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.22.mlp.dense_h_to_4h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.22.mlp.dense_h_to_4h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.22.mlp.dense_h_to_4h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.22.mlp.dense_4h_to_h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.22.mlp.dense_4h_to_h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.22.mlp.dense_4h_to_h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.23.ln_attn.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.23.ln_attn.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.23.self_attention.query_key_value.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.23.self_attention.query_key_value.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.23.self_attention.query_key_value.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.23.self_attention.dense.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.23.self_attention.dense.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.23.self_attention.dense.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.23.ln_mlp.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.23.ln_mlp.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.23.mlp.dense_h_to_4h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.23.mlp.dense_h_to_4h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.23.mlp.dense_h_to_4h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.23.mlp.dense_4h_to_h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.23.mlp.dense_4h_to_h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.23.mlp.dense_4h_to_h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.24.ln_attn.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.24.ln_attn.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.24.self_attention.query_key_value.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.24.self_attention.query_key_value.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.24.self_attention.query_key_value.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.24.self_attention.dense.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.24.self_attention.dense.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.24.self_attention.dense.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.24.ln_mlp.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.24.ln_mlp.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.24.mlp.dense_h_to_4h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.24.mlp.dense_h_to_4h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.24.mlp.dense_h_to_4h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.24.mlp.dense_4h_to_h.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.24.mlp.dense_4h_to_h.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.24.mlp.dense_4h_to_h.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.25.ln_attn.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.25.ln_attn.bias": "model-00001-of-00003.safetensors",
+        "transformer.h.25.self_attention.query_key_value.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.25.self_attention.query_key_value.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.25.self_attention.query_key_value.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.25.self_attention.dense.qweight": "model-00001-of-00003.safetensors",
+        "transformer.h.25.self_attention.dense.qzeros": "model-00001-of-00003.safetensors",
+        "transformer.h.25.self_attention.dense.scales": "model-00001-of-00003.safetensors",
+        "transformer.h.25.ln_mlp.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.25.ln_mlp.bias": "model-00001-of-00003.safetensors",
+        "lm_head.weight": "model-00001-of-00003.safetensors",
+        "transformer.h.25.mlp.dense_h_to_4h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.25.mlp.dense_h_to_4h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.25.mlp.dense_h_to_4h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.25.mlp.dense_4h_to_h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.25.mlp.dense_4h_to_h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.25.mlp.dense_4h_to_h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.26.ln_attn.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.26.ln_attn.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.26.self_attention.query_key_value.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.26.self_attention.query_key_value.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.26.self_attention.query_key_value.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.26.self_attention.dense.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.26.self_attention.dense.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.26.self_attention.dense.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.26.ln_mlp.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.26.ln_mlp.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.26.mlp.dense_h_to_4h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.26.mlp.dense_h_to_4h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.26.mlp.dense_h_to_4h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.26.mlp.dense_4h_to_h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.26.mlp.dense_4h_to_h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.26.mlp.dense_4h_to_h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.27.ln_attn.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.27.ln_attn.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.27.self_attention.query_key_value.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.27.self_attention.query_key_value.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.27.self_attention.query_key_value.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.27.self_attention.dense.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.27.self_attention.dense.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.27.self_attention.dense.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.27.ln_mlp.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.27.ln_mlp.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.27.mlp.dense_h_to_4h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.27.mlp.dense_h_to_4h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.27.mlp.dense_h_to_4h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.27.mlp.dense_4h_to_h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.27.mlp.dense_4h_to_h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.27.mlp.dense_4h_to_h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.28.ln_attn.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.28.ln_attn.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.28.self_attention.query_key_value.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.28.self_attention.query_key_value.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.28.self_attention.query_key_value.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.28.self_attention.dense.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.28.self_attention.dense.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.28.self_attention.dense.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.28.ln_mlp.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.28.ln_mlp.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.28.mlp.dense_h_to_4h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.28.mlp.dense_h_to_4h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.28.mlp.dense_h_to_4h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.28.mlp.dense_4h_to_h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.28.mlp.dense_4h_to_h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.28.mlp.dense_4h_to_h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.29.ln_attn.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.29.ln_attn.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.29.self_attention.query_key_value.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.29.self_attention.query_key_value.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.29.self_attention.query_key_value.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.29.self_attention.dense.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.29.self_attention.dense.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.29.self_attention.dense.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.29.ln_mlp.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.29.ln_mlp.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.29.mlp.dense_h_to_4h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.29.mlp.dense_h_to_4h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.29.mlp.dense_h_to_4h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.29.mlp.dense_4h_to_h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.29.mlp.dense_4h_to_h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.29.mlp.dense_4h_to_h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.30.ln_attn.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.30.ln_attn.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.30.self_attention.query_key_value.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.30.self_attention.query_key_value.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.30.self_attention.query_key_value.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.30.self_attention.dense.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.30.self_attention.dense.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.30.self_attention.dense.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.30.ln_mlp.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.30.ln_mlp.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.30.mlp.dense_h_to_4h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.30.mlp.dense_h_to_4h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.30.mlp.dense_h_to_4h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.30.mlp.dense_4h_to_h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.30.mlp.dense_4h_to_h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.30.mlp.dense_4h_to_h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.31.ln_attn.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.31.ln_attn.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.31.self_attention.query_key_value.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.31.self_attention.query_key_value.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.31.self_attention.query_key_value.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.31.self_attention.dense.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.31.self_attention.dense.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.31.self_attention.dense.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.31.ln_mlp.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.31.ln_mlp.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.31.mlp.dense_h_to_4h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.31.mlp.dense_h_to_4h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.31.mlp.dense_h_to_4h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.31.mlp.dense_4h_to_h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.31.mlp.dense_4h_to_h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.31.mlp.dense_4h_to_h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.32.ln_attn.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.32.ln_attn.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.32.self_attention.query_key_value.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.32.self_attention.query_key_value.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.32.self_attention.query_key_value.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.32.self_attention.dense.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.32.self_attention.dense.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.32.self_attention.dense.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.32.ln_mlp.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.32.ln_mlp.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.32.mlp.dense_h_to_4h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.32.mlp.dense_h_to_4h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.32.mlp.dense_h_to_4h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.32.mlp.dense_4h_to_h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.32.mlp.dense_4h_to_h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.32.mlp.dense_4h_to_h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.33.ln_attn.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.33.ln_attn.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.33.self_attention.query_key_value.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.33.self_attention.query_key_value.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.33.self_attention.query_key_value.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.33.self_attention.dense.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.33.self_attention.dense.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.33.self_attention.dense.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.33.ln_mlp.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.33.ln_mlp.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.33.mlp.dense_h_to_4h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.33.mlp.dense_h_to_4h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.33.mlp.dense_h_to_4h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.33.mlp.dense_4h_to_h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.33.mlp.dense_4h_to_h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.33.mlp.dense_4h_to_h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.34.ln_attn.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.34.ln_attn.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.34.self_attention.query_key_value.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.34.self_attention.query_key_value.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.34.self_attention.query_key_value.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.34.self_attention.dense.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.34.self_attention.dense.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.34.self_attention.dense.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.34.ln_mlp.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.34.ln_mlp.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.34.mlp.dense_h_to_4h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.34.mlp.dense_h_to_4h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.34.mlp.dense_h_to_4h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.34.mlp.dense_4h_to_h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.34.mlp.dense_4h_to_h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.34.mlp.dense_4h_to_h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.35.ln_attn.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.35.ln_attn.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.35.self_attention.query_key_value.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.35.self_attention.query_key_value.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.35.self_attention.query_key_value.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.35.self_attention.dense.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.35.self_attention.dense.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.35.self_attention.dense.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.35.ln_mlp.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.35.ln_mlp.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.35.mlp.dense_h_to_4h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.35.mlp.dense_h_to_4h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.35.mlp.dense_h_to_4h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.35.mlp.dense_4h_to_h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.35.mlp.dense_4h_to_h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.35.mlp.dense_4h_to_h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.36.ln_attn.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.36.ln_attn.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.36.self_attention.query_key_value.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.36.self_attention.query_key_value.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.36.self_attention.query_key_value.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.36.self_attention.dense.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.36.self_attention.dense.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.36.self_attention.dense.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.36.ln_mlp.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.36.ln_mlp.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.36.mlp.dense_h_to_4h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.36.mlp.dense_h_to_4h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.36.mlp.dense_h_to_4h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.36.mlp.dense_4h_to_h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.36.mlp.dense_4h_to_h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.36.mlp.dense_4h_to_h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.37.ln_attn.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.37.ln_attn.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.37.self_attention.query_key_value.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.37.self_attention.query_key_value.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.37.self_attention.query_key_value.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.37.self_attention.dense.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.37.self_attention.dense.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.37.self_attention.dense.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.37.ln_mlp.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.37.ln_mlp.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.37.mlp.dense_h_to_4h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.37.mlp.dense_h_to_4h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.37.mlp.dense_h_to_4h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.37.mlp.dense_4h_to_h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.37.mlp.dense_4h_to_h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.37.mlp.dense_4h_to_h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.38.ln_attn.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.38.ln_attn.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.38.self_attention.query_key_value.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.38.self_attention.query_key_value.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.38.self_attention.query_key_value.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.38.self_attention.dense.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.38.self_attention.dense.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.38.self_attention.dense.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.38.ln_mlp.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.38.ln_mlp.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.38.mlp.dense_h_to_4h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.38.mlp.dense_h_to_4h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.38.mlp.dense_h_to_4h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.38.mlp.dense_4h_to_h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.38.mlp.dense_4h_to_h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.38.mlp.dense_4h_to_h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.39.ln_attn.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.39.ln_attn.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.39.self_attention.query_key_value.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.39.self_attention.query_key_value.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.39.self_attention.query_key_value.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.39.self_attention.dense.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.39.self_attention.dense.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.39.self_attention.dense.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.39.ln_mlp.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.39.ln_mlp.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.39.mlp.dense_h_to_4h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.39.mlp.dense_h_to_4h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.39.mlp.dense_h_to_4h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.39.mlp.dense_4h_to_h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.39.mlp.dense_4h_to_h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.39.mlp.dense_4h_to_h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.40.ln_attn.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.40.ln_attn.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.40.self_attention.query_key_value.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.40.self_attention.query_key_value.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.40.self_attention.query_key_value.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.40.self_attention.dense.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.40.self_attention.dense.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.40.self_attention.dense.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.40.ln_mlp.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.40.ln_mlp.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.40.mlp.dense_h_to_4h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.40.mlp.dense_h_to_4h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.40.mlp.dense_h_to_4h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.40.mlp.dense_4h_to_h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.40.mlp.dense_4h_to_h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.40.mlp.dense_4h_to_h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.41.ln_attn.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.41.ln_attn.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.41.self_attention.query_key_value.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.41.self_attention.query_key_value.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.41.self_attention.query_key_value.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.41.self_attention.dense.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.41.self_attention.dense.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.41.self_attention.dense.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.41.ln_mlp.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.41.ln_mlp.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.41.mlp.dense_h_to_4h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.41.mlp.dense_h_to_4h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.41.mlp.dense_h_to_4h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.41.mlp.dense_4h_to_h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.41.mlp.dense_4h_to_h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.41.mlp.dense_4h_to_h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.42.ln_attn.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.42.ln_attn.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.42.self_attention.query_key_value.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.42.self_attention.query_key_value.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.42.self_attention.query_key_value.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.42.self_attention.dense.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.42.self_attention.dense.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.42.self_attention.dense.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.42.ln_mlp.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.42.ln_mlp.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.42.mlp.dense_h_to_4h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.42.mlp.dense_h_to_4h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.42.mlp.dense_h_to_4h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.42.mlp.dense_4h_to_h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.42.mlp.dense_4h_to_h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.42.mlp.dense_4h_to_h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.43.ln_attn.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.43.ln_attn.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.43.self_attention.query_key_value.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.43.self_attention.query_key_value.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.43.self_attention.query_key_value.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.43.self_attention.dense.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.43.self_attention.dense.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.43.self_attention.dense.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.43.ln_mlp.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.43.ln_mlp.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.43.mlp.dense_h_to_4h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.43.mlp.dense_h_to_4h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.43.mlp.dense_h_to_4h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.43.mlp.dense_4h_to_h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.43.mlp.dense_4h_to_h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.43.mlp.dense_4h_to_h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.44.ln_attn.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.44.ln_attn.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.44.self_attention.query_key_value.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.44.self_attention.query_key_value.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.44.self_attention.query_key_value.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.44.self_attention.dense.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.44.self_attention.dense.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.44.self_attention.dense.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.44.ln_mlp.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.44.ln_mlp.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.44.mlp.dense_h_to_4h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.44.mlp.dense_h_to_4h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.44.mlp.dense_h_to_4h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.44.mlp.dense_4h_to_h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.44.mlp.dense_4h_to_h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.44.mlp.dense_4h_to_h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.45.ln_attn.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.45.ln_attn.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.45.self_attention.query_key_value.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.45.self_attention.query_key_value.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.45.self_attention.query_key_value.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.45.self_attention.dense.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.45.self_attention.dense.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.45.self_attention.dense.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.45.ln_mlp.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.45.ln_mlp.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.45.mlp.dense_h_to_4h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.45.mlp.dense_h_to_4h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.45.mlp.dense_h_to_4h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.45.mlp.dense_4h_to_h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.45.mlp.dense_4h_to_h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.45.mlp.dense_4h_to_h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.46.ln_attn.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.46.ln_attn.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.46.self_attention.query_key_value.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.46.self_attention.query_key_value.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.46.self_attention.query_key_value.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.46.self_attention.dense.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.46.self_attention.dense.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.46.self_attention.dense.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.46.ln_mlp.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.46.ln_mlp.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.46.mlp.dense_h_to_4h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.46.mlp.dense_h_to_4h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.46.mlp.dense_h_to_4h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.46.mlp.dense_4h_to_h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.46.mlp.dense_4h_to_h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.46.mlp.dense_4h_to_h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.47.ln_attn.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.47.ln_attn.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.47.self_attention.query_key_value.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.47.self_attention.query_key_value.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.47.self_attention.query_key_value.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.47.self_attention.dense.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.47.self_attention.dense.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.47.self_attention.dense.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.47.ln_mlp.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.47.ln_mlp.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.47.mlp.dense_h_to_4h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.47.mlp.dense_h_to_4h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.47.mlp.dense_h_to_4h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.47.mlp.dense_4h_to_h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.47.mlp.dense_4h_to_h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.47.mlp.dense_4h_to_h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.48.ln_attn.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.48.ln_attn.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.48.self_attention.query_key_value.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.48.self_attention.query_key_value.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.48.self_attention.query_key_value.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.48.self_attention.dense.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.48.self_attention.dense.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.48.self_attention.dense.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.48.ln_mlp.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.48.ln_mlp.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.48.mlp.dense_h_to_4h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.48.mlp.dense_h_to_4h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.48.mlp.dense_h_to_4h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.48.mlp.dense_4h_to_h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.48.mlp.dense_4h_to_h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.48.mlp.dense_4h_to_h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.49.ln_attn.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.49.ln_attn.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.49.self_attention.query_key_value.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.49.self_attention.query_key_value.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.49.self_attention.query_key_value.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.49.self_attention.dense.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.49.self_attention.dense.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.49.self_attention.dense.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.49.ln_mlp.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.49.ln_mlp.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.49.mlp.dense_h_to_4h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.49.mlp.dense_h_to_4h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.49.mlp.dense_h_to_4h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.49.mlp.dense_4h_to_h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.49.mlp.dense_4h_to_h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.49.mlp.dense_4h_to_h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.50.ln_attn.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.50.ln_attn.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.50.self_attention.query_key_value.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.50.self_attention.query_key_value.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.50.self_attention.query_key_value.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.50.self_attention.dense.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.50.self_attention.dense.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.50.self_attention.dense.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.50.ln_mlp.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.50.ln_mlp.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.50.mlp.dense_h_to_4h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.50.mlp.dense_h_to_4h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.50.mlp.dense_h_to_4h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.50.mlp.dense_4h_to_h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.50.mlp.dense_4h_to_h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.50.mlp.dense_4h_to_h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.51.ln_attn.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.51.ln_attn.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.51.self_attention.query_key_value.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.51.self_attention.query_key_value.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.51.self_attention.query_key_value.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.51.self_attention.dense.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.51.self_attention.dense.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.51.self_attention.dense.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.51.ln_mlp.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.51.ln_mlp.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.51.mlp.dense_h_to_4h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.51.mlp.dense_h_to_4h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.51.mlp.dense_h_to_4h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.51.mlp.dense_4h_to_h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.51.mlp.dense_4h_to_h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.51.mlp.dense_4h_to_h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.52.ln_attn.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.52.ln_attn.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.52.self_attention.query_key_value.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.52.self_attention.query_key_value.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.52.self_attention.query_key_value.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.52.self_attention.dense.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.52.self_attention.dense.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.52.self_attention.dense.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.52.ln_mlp.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.52.ln_mlp.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.52.mlp.dense_h_to_4h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.52.mlp.dense_h_to_4h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.52.mlp.dense_h_to_4h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.52.mlp.dense_4h_to_h.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.52.mlp.dense_4h_to_h.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.52.mlp.dense_4h_to_h.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.53.ln_attn.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.53.ln_attn.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.53.self_attention.query_key_value.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.53.self_attention.query_key_value.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.53.self_attention.query_key_value.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.53.self_attention.dense.qweight": "model-00002-of-00003.safetensors",
+        "transformer.h.53.self_attention.dense.qzeros": "model-00002-of-00003.safetensors",
+        "transformer.h.53.self_attention.dense.scales": "model-00002-of-00003.safetensors",
+        "transformer.h.53.ln_mlp.weight": "model-00002-of-00003.safetensors",
+        "transformer.h.53.ln_mlp.bias": "model-00002-of-00003.safetensors",
+        "transformer.h.53.mlp.dense_h_to_4h.qweight": "model-00003-of-00003.safetensors",
+        "transformer.h.53.mlp.dense_h_to_4h.qzeros": "model-00003-of-00003.safetensors",
+        "transformer.h.53.mlp.dense_h_to_4h.scales": "model-00003-of-00003.safetensors",
+        "transformer.h.53.mlp.dense_4h_to_h.qweight": "model-00003-of-00003.safetensors",
+        "transformer.h.53.mlp.dense_4h_to_h.qzeros": "model-00003-of-00003.safetensors",
+        "transformer.h.53.mlp.dense_4h_to_h.scales": "model-00003-of-00003.safetensors",
+        "transformer.h.54.ln_attn.weight": "model-00003-of-00003.safetensors",
+        "transformer.h.54.ln_attn.bias": "model-00003-of-00003.safetensors",
+        "transformer.h.54.self_attention.query_key_value.qweight": "model-00003-of-00003.safetensors",
+        "transformer.h.54.self_attention.query_key_value.qzeros": "model-00003-of-00003.safetensors",
+        "transformer.h.54.self_attention.query_key_value.scales": "model-00003-of-00003.safetensors",
+        "transformer.h.54.self_attention.dense.qweight": "model-00003-of-00003.safetensors",
+        "transformer.h.54.self_attention.dense.qzeros": "model-00003-of-00003.safetensors",
+        "transformer.h.54.self_attention.dense.scales": "model-00003-of-00003.safetensors",
+        "transformer.h.54.ln_mlp.weight": "model-00003-of-00003.safetensors",
+        "transformer.h.54.ln_mlp.bias": "model-00003-of-00003.safetensors",
+        "transformer.h.54.mlp.dense_h_to_4h.qweight": "model-00003-of-00003.safetensors",
+        "transformer.h.54.mlp.dense_h_to_4h.qzeros": "model-00003-of-00003.safetensors",
+        "transformer.h.54.mlp.dense_h_to_4h.scales": "model-00003-of-00003.safetensors",
+        "transformer.h.54.mlp.dense_4h_to_h.qweight": "model-00003-of-00003.safetensors",
+        "transformer.h.54.mlp.dense_4h_to_h.qzeros": "model-00003-of-00003.safetensors",
+        "transformer.h.54.mlp.dense_4h_to_h.scales": "model-00003-of-00003.safetensors",
+        "transformer.h.55.ln_attn.weight": "model-00003-of-00003.safetensors",
+        "transformer.h.55.ln_attn.bias": "model-00003-of-00003.safetensors",
+        "transformer.h.55.self_attention.query_key_value.qweight": "model-00003-of-00003.safetensors",
+        "transformer.h.55.self_attention.query_key_value.qzeros": "model-00003-of-00003.safetensors",
+        "transformer.h.55.self_attention.query_key_value.scales": "model-00003-of-00003.safetensors",
+        "transformer.h.55.self_attention.dense.qweight": "model-00003-of-00003.safetensors",
+        "transformer.h.55.self_attention.dense.qzeros": "model-00003-of-00003.safetensors",
+        "transformer.h.55.self_attention.dense.scales": "model-00003-of-00003.safetensors",
+        "transformer.h.55.ln_mlp.weight": "model-00003-of-00003.safetensors",
+        "transformer.h.55.ln_mlp.bias": "model-00003-of-00003.safetensors",
+        "transformer.h.55.mlp.dense_h_to_4h.qweight": "model-00003-of-00003.safetensors",
+        "transformer.h.55.mlp.dense_h_to_4h.qzeros": "model-00003-of-00003.safetensors",
+        "transformer.h.55.mlp.dense_h_to_4h.scales": "model-00003-of-00003.safetensors",
+        "transformer.h.55.mlp.dense_4h_to_h.qweight": "model-00003-of-00003.safetensors",
+        "transformer.h.55.mlp.dense_4h_to_h.qzeros": "model-00003-of-00003.safetensors",
+        "transformer.h.55.mlp.dense_4h_to_h.scales": "model-00003-of-00003.safetensors",
+        "transformer.h.56.ln_attn.weight": "model-00003-of-00003.safetensors",
+        "transformer.h.56.ln_attn.bias": "model-00003-of-00003.safetensors",
+        "transformer.h.56.self_attention.query_key_value.qweight": "model-00003-of-00003.safetensors",
+        "transformer.h.56.self_attention.query_key_value.qzeros": "model-00003-of-00003.safetensors",
+        "transformer.h.56.self_attention.query_key_value.scales": "model-00003-of-00003.safetensors",
+        "transformer.h.56.self_attention.dense.qweight": "model-00003-of-00003.safetensors",
+        "transformer.h.56.self_attention.dense.qzeros": "model-00003-of-00003.safetensors",
+        "transformer.h.56.self_attention.dense.scales": "model-00003-of-00003.safetensors",
+        "transformer.h.56.ln_mlp.weight": "model-00003-of-00003.safetensors",
+        "transformer.h.56.ln_mlp.bias": "model-00003-of-00003.safetensors",
+        "transformer.h.56.mlp.dense_h_to_4h.qweight": "model-00003-of-00003.safetensors",
+        "transformer.h.56.mlp.dense_h_to_4h.qzeros": "model-00003-of-00003.safetensors",
+        "transformer.h.56.mlp.dense_h_to_4h.scales": "model-00003-of-00003.safetensors",
+        "transformer.h.56.mlp.dense_4h_to_h.qweight": "model-00003-of-00003.safetensors",
+        "transformer.h.56.mlp.dense_4h_to_h.qzeros": "model-00003-of-00003.safetensors",
+        "transformer.h.56.mlp.dense_4h_to_h.scales": "model-00003-of-00003.safetensors",
+        "transformer.h.57.ln_attn.weight": "model-00003-of-00003.safetensors",
+        "transformer.h.57.ln_attn.bias": "model-00003-of-00003.safetensors",
+        "transformer.h.57.self_attention.query_key_value.qweight": "model-00003-of-00003.safetensors",
+        "transformer.h.57.self_attention.query_key_value.qzeros": "model-00003-of-00003.safetensors",
+        "transformer.h.57.self_attention.query_key_value.scales": "model-00003-of-00003.safetensors",
+        "transformer.h.57.self_attention.dense.qweight": "model-00003-of-00003.safetensors",
+        "transformer.h.57.self_attention.dense.qzeros": "model-00003-of-00003.safetensors",
+        "transformer.h.57.self_attention.dense.scales": "model-00003-of-00003.safetensors",
+        "transformer.h.57.ln_mlp.weight": "model-00003-of-00003.safetensors",
+        "transformer.h.57.ln_mlp.bias": "model-00003-of-00003.safetensors",
+        "transformer.h.57.mlp.dense_h_to_4h.qweight": "model-00003-of-00003.safetensors",
+        "transformer.h.57.mlp.dense_h_to_4h.qzeros": "model-00003-of-00003.safetensors",
+        "transformer.h.57.mlp.dense_h_to_4h.scales": "model-00003-of-00003.safetensors",
+        "transformer.h.57.mlp.dense_4h_to_h.qweight": "model-00003-of-00003.safetensors",
+        "transformer.h.57.mlp.dense_4h_to_h.qzeros": "model-00003-of-00003.safetensors",
+        "transformer.h.57.mlp.dense_4h_to_h.scales": "model-00003-of-00003.safetensors",
+        "transformer.h.58.ln_attn.weight": "model-00003-of-00003.safetensors",
+        "transformer.h.58.ln_attn.bias": "model-00003-of-00003.safetensors",
+        "transformer.h.58.self_attention.query_key_value.qweight": "model-00003-of-00003.safetensors",
+        "transformer.h.58.self_attention.query_key_value.qzeros": "model-00003-of-00003.safetensors",
+        "transformer.h.58.self_attention.query_key_value.scales": "model-00003-of-00003.safetensors",
+        "transformer.h.58.self_attention.dense.qweight": "model-00003-of-00003.safetensors",
+        "transformer.h.58.self_attention.dense.qzeros": "model-00003-of-00003.safetensors",
+        "transformer.h.58.self_attention.dense.scales": "model-00003-of-00003.safetensors",
+        "transformer.h.58.ln_mlp.weight": "model-00003-of-00003.safetensors",
+        "transformer.h.58.ln_mlp.bias": "model-00003-of-00003.safetensors",
+        "transformer.h.58.mlp.dense_h_to_4h.qweight": "model-00003-of-00003.safetensors",
+        "transformer.h.58.mlp.dense_h_to_4h.qzeros": "model-00003-of-00003.safetensors",
+        "transformer.h.58.mlp.dense_h_to_4h.scales": "model-00003-of-00003.safetensors",
+        "transformer.h.58.mlp.dense_4h_to_h.qweight": "model-00003-of-00003.safetensors",
+        "transformer.h.58.mlp.dense_4h_to_h.qzeros": "model-00003-of-00003.safetensors",
+        "transformer.h.58.mlp.dense_4h_to_h.scales": "model-00003-of-00003.safetensors",
+        "transformer.h.59.ln_attn.weight": "model-00003-of-00003.safetensors",
+        "transformer.h.59.ln_attn.bias": "model-00003-of-00003.safetensors",
+        "transformer.h.59.self_attention.query_key_value.qweight": "model-00003-of-00003.safetensors",
+        "transformer.h.59.self_attention.query_key_value.qzeros": "model-00003-of-00003.safetensors",
+        "transformer.h.59.self_attention.query_key_value.scales": "model-00003-of-00003.safetensors",
+        "transformer.h.59.self_attention.dense.qweight": "model-00003-of-00003.safetensors",
+        "transformer.h.59.self_attention.dense.qzeros": "model-00003-of-00003.safetensors",
+        "transformer.h.59.self_attention.dense.scales": "model-00003-of-00003.safetensors",
+        "transformer.h.59.ln_mlp.weight": "model-00003-of-00003.safetensors",
+        "transformer.h.59.ln_mlp.bias": "model-00003-of-00003.safetensors",
+        "transformer.h.59.mlp.dense_h_to_4h.qweight": "model-00003-of-00003.safetensors",
+        "transformer.h.59.mlp.dense_h_to_4h.qzeros": "model-00003-of-00003.safetensors",
+        "transformer.h.59.mlp.dense_h_to_4h.scales": "model-00003-of-00003.safetensors",
+        "transformer.h.59.mlp.dense_4h_to_h.qweight": "model-00003-of-00003.safetensors",
+        "transformer.h.59.mlp.dense_4h_to_h.qzeros": "model-00003-of-00003.safetensors",
+        "transformer.h.59.mlp.dense_4h_to_h.scales": "model-00003-of-00003.safetensors",
+        "transformer.ln_f.weight": "model-00003-of-00003.safetensors",
+        "transformer.ln_f.bias": "model-00003-of-00003.safetensors"
+    }
+}

modeling_RW.py

@@ -0,0 +1,1134 @@
+# port of models described in RW
+# We use the bloom model as a starting point for these model.
+# Please refer to the bloom models for usage instructions.
+
+import math
+import warnings
+from typing import Optional, Tuple, Union
+
+import torch
+import torch.utils.checkpoint
+from torch import nn
+from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, LayerNorm, MSELoss
+from torch.nn import functional as F
+
+from transformers.modeling_outputs import (
+    BaseModelOutputWithPastAndCrossAttentions,
+    CausalLMOutputWithCrossAttentions,
+    QuestionAnsweringModelOutput,
+    SequenceClassifierOutputWithPast,
+    TokenClassifierOutput,
+)
+from transformers.modeling_utils import PreTrainedModel
+from transformers.utils import logging
+from .configuration_RW import RWConfig
+
+logger = logging.get_logger(__name__)
+
+# NOTE: Unfortunately we did not fuse matmul and bias during training, this means that there's one additional quantization to bfloat16 between the operations.
+# In order not to degrade the quality of our HF-port, we keep these characteristics in the final model.
+class Linear(nn.Linear):
+    def forward(self, input: torch.Tensor) -> torch.Tensor:
+        ret = input @ self.weight.T
+        if self.bias is None:
+            return ret
+        else:
+            return ret + self.bias
+
+
+from einops import rearrange
+
+# rotary pos emb helpers (torch.jit.script does not seem to support staticmethod...)
+def rotate_half(x):
+    x1, x2 = x[..., : x.shape[-1] // 2], x[..., x.shape[-1] // 2 :]
+    return torch.cat((-x2, x1), dim=x1.ndim - 1)  # dim=-1 triggers a bug in torch < 1.8.0
+
+
+class RotaryEmbedding(torch.nn.Module):
+    """Implementation of RotaryEmbedding from GPT-NeoX.
+    This implementation is design to operate on queries and keys that are compatible with
+    [batch_size, n_heads_per_partition, seq_len, head_dim] (e.g. MinGPTAttention format).
+    """
+
+    def __init__(
+        self,
+        head_dim: int,
+        base=10000,
+        position_interpolation_factor=1,
+        embedding_factor=None
+    ):
+        super().__init__()
+        inv_freq = 1.0 / (base ** (torch.arange(0, head_dim, 2).float() / head_dim))
+        self.position_interpolation_factor = position_interpolation_factor 
+        self.embedding_factor = embedding_factor
+        self.register_buffer("inv_freq", inv_freq, persistent=False)
+        self.head_dim = head_dim
+        self.seq_len_cached = None
+        self.batch_size_cached = None
+        self.cos_cached: torch.Tensor | None = None
+        self.sin_cached: torch.Tensor | None = None
+
+    def cos_sin(
+        self,
+        seq_len: int,
+        device="cuda",
+        dtype=torch.bfloat16,
+    ) -> torch.Tensor:
+        if seq_len != self.seq_len_cached:
+            self.seq_len_cached = seq_len
+            t = torch.arange(seq_len, device=device).type_as(self.inv_freq)
+            freqs = torch.einsum("i,j->ij", t, self.inv_freq) / self.position_interpolation_factor
+            emb = torch.cat((freqs, freqs), dim=-1).to(device)
+
+            if dtype in [torch.float16, torch.bfloat16]:
+                emb = emb.float()
+
+            self.cos_cached = emb.cos()[None, :, :]
+            self.sin_cached = emb.sin()[None, :, :]
+
+            self.cos_cached = self.cos_cached.type(dtype)
+            self.sin_cached = self.sin_cached.type(dtype)
+
+        return self.cos_cached, self.sin_cached
+
+    def forward(self, q, k):
+        batch, seq_len, head_dim = q.shape
+        cos, sin = self.cos_sin(seq_len, q.device, q.dtype) 
+        if self.embedding_factor is not None:
+            cos = cos * self.embedding_factor
+            sin = sin * self.embedding_factor
+        return (q * cos) + (rotate_half(q) * sin), (k * cos) + (rotate_half(k) * sin)
+
+
+def _make_causal_mask(
+    input_ids_shape: torch.Size, device: torch.device, past_key_values_length: int
+) -> torch.BoolTensor:
+    batch_size, target_length = input_ids_shape
+    mask = torch.empty((target_length, target_length + past_key_values_length), dtype=torch.bool, device=device)
+    # ONNX doesn't support `torch.Tensor.triu` properly, thus we use this workaround
+    seq_ids = torch.arange(target_length, device=device)
+    mask[:, past_key_values_length:] = seq_ids[:, None] < seq_ids[None, :]
+
+    if past_key_values_length > 0:
+        mask[:, :past_key_values_length] = False
+
+    expanded_mask = mask[None, None, :, :].expand(batch_size, 1, target_length, target_length + past_key_values_length)
+    return expanded_mask
+
+
+def _expand_mask(mask: torch.Tensor, tgt_length: int) -> torch.BoolTensor:
+    batch_size, src_length = mask.shape
+    tgt_length = tgt_length if tgt_length is not None else src_length
+
+    expanded_mask = ~(mask[:, None, None, :].to(torch.bool))
+    return expanded_mask.expand(batch_size, 1, tgt_length, src_length)
+
+
+def build_alibi_tensor(attention_mask: torch.Tensor, num_heads: int, dtype: torch.dtype) -> torch.Tensor:
+    batch_size, seq_length = attention_mask.shape
+    closest_power_of_2 = 2 ** math.floor(math.log2(num_heads))
+    base = torch.tensor(
+        2 ** (-(2 ** -(math.log2(closest_power_of_2) - 3))), device=attention_mask.device, dtype=torch.float32
+    )
+    powers = torch.arange(1, 1 + closest_power_of_2, device=attention_mask.device, dtype=torch.int32)
+    slopes = torch.pow(base, powers)
+
+    if closest_power_of_2 != num_heads:
+        extra_base = torch.tensor(
+            2 ** (-(2 ** -(math.log2(2 * closest_power_of_2) - 3))), device=attention_mask.device, dtype=torch.float32
+        )
+        num_remaining_heads = min(closest_power_of_2, num_heads - closest_power_of_2)
+        extra_powers = torch.arange(1, 1 + 2 * num_remaining_heads, 2, device=attention_mask.device, dtype=torch.int32)
+        slopes = torch.cat([slopes, torch.pow(extra_base, extra_powers)], dim=0)
+
+    # Note: alibi will added to the attention bias that will be applied to the query, key product of attention
+    # => therefore alibi will have to be of shape (batch_size, num_heads, query_length, key_length)
+    # => here we set (batch_size=1, num_heads=num_heads, query_length=1, key_length=max_length)
+    # => the query_length dimension will then be broadcasted correctly
+    # This is more or less identical to T5's relative position bias:
+    # https://github.com/huggingface/transformers/blob/f681437203baa7671de3174b0fa583c349d9d5e1/src/transformers/models/t5/modeling_t5.py#L527
+    arange_tensor = ((attention_mask.cumsum(dim=-1) - 1) * attention_mask)[:, None, :]
+    alibi = slopes[..., None].bfloat16() * arange_tensor
+    return alibi.reshape(batch_size * num_heads, 1, seq_length).to(dtype)
+
+
+def dropout_add(x: torch.Tensor, residual: torch.Tensor, prob: float, training: bool) -> torch.Tensor:
+    out = F.dropout(x, p=prob, training=training)
+    out = residual + out
+    return out
+
+
+class Attention(nn.Module):
+    def __init__(self, config: RWConfig):
+        super().__init__()
+
+        self.hidden_size = config.hidden_size
+        self.num_heads = config.n_head
+        self.head_dim = self.hidden_size // self.num_heads
+        self.split_size = self.hidden_size
+        self.hidden_dropout = config.hidden_dropout
+
+        if self.head_dim * self.num_heads != self.hidden_size:
+            raise ValueError(
+                f"`hidden_size` must be divisible by num_heads (got `hidden_size`: {self.hidden_size} and `num_heads`:"
+                f" {self.num_heads})."
+            )
+
+        # create rotary embedding and apply context scaling if any       
+        base = 10000
+        scale = 1 if config.vanilla_scaling_factor is None else config.vanilla_scaling_factor
+        if config.ntk_scaling_factor is not None:
+            base = base * (config.ntk_scaling_factor ** (self.head_dim / (self.head_dim - 2)))
+        embedding_scale = None if config.embedding_scaling_factor is None else 0.1 * math.log(config.embedding_scaling_factor) + 1
+        self.maybe_rotary = RotaryEmbedding(config.head_dim, base, scale, embedding_scale) if config.rotary else lambda q, k: (q, k)
+
+        # Layer-wise attention scaling
+        self.inv_norm_factor = 1.0 / math.sqrt(self.head_dim)
+        self.beta = self.inv_norm_factor
+        self.num_kv = config.n_head if not config.multi_query else config.n_head_kv
+        self.query_key_value = Linear(
+            self.hidden_size,
+            (self.num_kv * 2 + config.n_head) * self.head_dim,
+            bias=config.bias,
+        )
+        self.multi_query = config.multi_query
+        self.dense = Linear(self.hidden_size, self.hidden_size, bias=config.bias)
+        self.attention_dropout = nn.Dropout(config.attention_dropout)
+
+
+    def _split_heads(self, fused_qkv: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor]:
+        """
+        Split the last dimension into (num_heads, head_dim) without making any copies, results share same memory
+        storage as `fused_qkv`
+
+        Args:
+            fused_qkv (`torch.tensor`, *required*): [batch_size, seq_length, (num_heads + 2 * num_kv) * head_dim]
+
+        Returns:
+            query: [batch_size, seq_length, num_heads, head_dim] 
+            key: [batch_size, seq_length, num_heads, head_dim]
+            value: [batch_size, seq_length, num_heads, head_dim]
+        """
+        if not self.multi_query:
+            batch_size, seq_length, _ = fused_qkv.shape
+            fused_qkv = fused_qkv.view(batch_size, seq_length, self.num_heads, 3, self.head_dim)
+            return fused_qkv[..., 0, :], fused_qkv[..., 1, :], fused_qkv[..., 2, :]
+        else:
+            batch, seq_len, _ = fused_qkv.shape
+            qkv = fused_qkv.view(batch, seq_len, -1, self.num_heads // self.num_kv + 2, 64)
+            q = qkv[:, :, :, :-2]
+            k = qkv[:, :, :, [-2]]
+            v = qkv[:, :, :, [-1]]
+            k = torch.broadcast_to(k, q.shape)
+            v = torch.broadcast_to(v, q.shape)
+
+
+            q, k, v = [
+                rearrange(
+                    x,
+                    "batch seq_len group num_heads head_dim ->\
+                    batch seq_len (group num_heads) head_dim",
+                    head_dim=self.head_dim,
+                )
+                for x in [q, k, v]
+            ]
+            return q, k, v
+
+    def _merge_heads(self, x: torch.Tensor) -> torch.Tensor:
+        """
+        Merge heads together over the last dimenstion
+
+        Args:
+            x: (`torch.tensor`, *required*): [batch_size * num_heads, seq_length, head_dim]
+
+        Returns:
+            torch.tensor: [batch_size, seq_length, num_heads * head_dim]
+        """
+        # What we want to achieve is:
+        # batch_size * num_heads, seq_length, head_dim -> batch_size, seq_length, num_heads * head_dim
+        batch_size_and_num_heads, seq_length, _ = x.shape
+        batch_size = batch_size_and_num_heads // self.num_heads
+
+        # First view to decompose the batch size
+        # batch_size * num_heads, seq_length, head_dim -> batch_size, num_heads, seq_length, head_dim
+        x = x.view(batch_size, self.num_heads, seq_length, self.head_dim)
+
+        # batch_size, num_heads, seq_length, head_dim -> batch_size, seq_length, num_heads, head_dim
+        x = x.permute(0, 2, 1, 3)
+
+        # batch_size, seq_length, num_heads, head_dim -> batch_size, seq_length, num_heads * head_dim
+        return x.reshape(batch_size, seq_length, self.num_heads * self.head_dim)
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        alibi: torch.Tensor,
+        attention_mask: torch.Tensor,
+        layer_past: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,
+        head_mask: Optional[torch.Tensor] = None,
+        use_cache: bool = False,
+        output_attentions: bool = False,
+    ):
+        fused_qkv = self.query_key_value(hidden_states)  # [batch_size, seq_length, 3 x hidden_size]
+
+        # 3 x [batch_size, seq_length, num_heads, head_dim]
+        (query_layer, key_layer, value_layer) = self._split_heads(fused_qkv)
+
+        batch_size, q_length, _, _ = query_layer.shape
+
+        query_layer = query_layer.transpose(1, 2).reshape(batch_size * self.num_heads, q_length, self.head_dim)
+        key_layer = key_layer.transpose(1, 2).reshape(
+            batch_size * self.num_heads,
+            q_length,
+            self.head_dim,
+        )
+        value_layer = value_layer.transpose(1, 2).reshape(batch_size * self.num_heads, q_length, self.head_dim)
+
+        query_layer, key_layer = self.maybe_rotary(query_layer, key_layer)
+
+        if layer_past is not None:
+            past_key, past_value = layer_past
+            # concatenate along seq_length dimension:
+            #  - key: [batch_size * self.num_heads, head_dim, kv_length]
+            #  - value: [batch_size * self.num_heads, kv_length, head_dim]
+            key_layer = torch.cat((past_key, key_layer), dim=1)
+            value_layer = torch.cat((past_value, value_layer), dim=1)
+
+        _, kv_length, _ = key_layer.shape
+
+        if use_cache is True:
+            present = (key_layer, value_layer)
+        else:
+            present = None
+
+        if alibi is None:
+            query_layer_ = query_layer.reshape(batch_size, self.num_heads, -1, self.head_dim)
+            key_layer_ = key_layer.reshape(batch_size, self.num_heads, -1, self.head_dim)
+            value_layer_ = value_layer.reshape(batch_size, self.num_heads, -1, self.head_dim)
+
+            attn_output = F.scaled_dot_product_attention(
+                query_layer_, key_layer_, value_layer_, None, 0.0, is_causal=True
+            )
+
+            x = attn_output.view(batch_size, self.num_heads, q_length, self.head_dim)
+            x = x.permute(0, 2, 1, 3)
+            attn_output = x.reshape(batch_size, q_length, self.num_heads * self.head_dim)
+
+            output_tensor = self.dense(attn_output)
+
+            outputs = (output_tensor, present)
+            assert not output_attentions  # not supported.
+            return outputs
+        else:
+            attention_mask_float = (attention_mask * 1.0).masked_fill(attention_mask, -1e9).to(torch.bfloat16)
+            matmul_result = query_layer @ key_layer.transpose(-1, -2)
+
+            # change view to [batch_size, num_heads, q_length, kv_length]
+            attention_scores = matmul_result.view(batch_size, self.num_heads, q_length, kv_length)
+
+            # cast attention scores to fp32, compute scaled softmax and cast back to initial dtype - [batch_size, num_heads, q_length, kv_length]
+            input_dtype = attention_scores.dtype
+            # `float16` has a minimum value of -65504.0, whereas `bfloat16` and `float32` have a minimum value of `-3.4e+38`
+            if input_dtype == torch.float16 or input_dtype == torch.bfloat16:
+                attention_scores = attention_scores.to(torch.float32)
+            # attn_weights = torch.masked_fill(attention_scores, attention_mask, torch.finfo(attention_scores.dtype).min)
+            attention_probs = F.softmax(
+                (attention_scores + alibi) * self.inv_norm_factor + attention_mask_float,
+                dim=-1,
+                dtype=hidden_states.dtype,
+            )
+            # [batch_size, num_heads, q_length, kv_length]
+            attention_probs = self.attention_dropout(attention_probs)
+
+            if head_mask is not None:
+                attention_probs = attention_probs * head_mask
+
+            # change view [batch_size x num_heads, q_length, kv_length]
+            attention_probs_reshaped = attention_probs.view(batch_size * self.num_heads, q_length, kv_length)
+
+            # matmul: [batch_size * num_heads, q_length, head_dim]
+            context_layer = attention_probs_reshaped @ value_layer
+
+            # change view [batch_size, num_heads, q_length, head_dim]
+            context_layer = self._merge_heads(context_layer)
+
+            output_tensor = self.dense(context_layer)
+
+            outputs = (output_tensor, present)
+            if output_attentions:
+                outputs += (attention_probs,)
+
+            return outputs
+
+
+class MLP(nn.Module):
+    def __init__(self, config: RWConfig):
+        super().__init__()
+        hidden_size = config.hidden_size
+
+        self.dense_h_to_4h = Linear(hidden_size, 4 * hidden_size, bias=config.bias)
+        self.act = nn.GELU()
+        self.dense_4h_to_h = Linear(4 * hidden_size, hidden_size, bias=config.bias)
+        self.hidden_dropout = config.hidden_dropout
+
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        x = self.act(self.dense_h_to_4h(x))
+        x = self.dense_4h_to_h(x)
+        return x
+
+
+class DecoderLayer(nn.Module):
+    def __init__(self, config: RWConfig):
+        super().__init__()
+        hidden_size = config.hidden_size
+
+        self.ln_attn = LayerNorm(hidden_size, eps=config.layer_norm_epsilon)
+        self.num_heads = config.n_head
+        self.self_attention = Attention(config)
+
+        self.ln_mlp = LayerNorm(hidden_size, eps=config.layer_norm_epsilon)
+
+        self.mlp = MLP(config)
+
+        self.apply_residual_connection_post_layernorm = config.apply_residual_connection_post_layernorm
+        self.hidden_dropout = config.hidden_dropout
+
+        self.config = config
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        alibi: torch.Tensor,
+        attention_mask: torch.Tensor,
+        layer_past: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,
+        head_mask: Optional[torch.Tensor] = None,
+        use_cache: bool = False,
+        output_attentions: bool = False,
+    ):
+        # hidden_states: [batch_size, seq_length, hidden_size]
+
+        # Layer norm at the beginning of the transformer layer.
+        layernorm_output = self.ln_attn(hidden_states)
+
+        # Layer norm post the self attention.
+        residual = hidden_states
+
+        # Self attention.
+        attn_outputs = self.self_attention(
+            layernorm_output,
+            layer_past=layer_past,
+            attention_mask=attention_mask,
+            alibi=alibi,
+            head_mask=head_mask,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+        )
+
+        attention_output = attn_outputs[0]
+
+        if not self.config.parallel_attn:
+            residual = dropout_add(attention_output, residual, self.config.attention_dropout, training=self.training)
+            layernorm_output = self.ln_mlp(residual)
+        elif not self.config.single_ln:
+            layernorm_output = self.ln_mlp(residual)
+
+        outputs = attn_outputs[1:]
+
+        # MLP.
+        mlp_output = self.mlp(layernorm_output)
+
+        if self.config.parallel_attn:
+            mlp_output += attention_output
+
+        output = dropout_add(mlp_output, residual, self.config.hidden_dropout, training=self.training)
+
+        if use_cache:
+            outputs = (output,) + outputs
+        else:
+            outputs = (output,) + outputs[1:]
+
+        return outputs  # hidden_states, present, attentions
+
+
+class RWPreTrainedModel(PreTrainedModel):
+    _keys_to_ignore_on_load_missing = [r"h.*.self_attention.scale_mask_softmax.causal_mask", r"lm_head.weight"]
+    """
+    An abstract class to handle weights initialization and a simple interface for downloading and loading pretrained
+    models.
+    """
+
+    config_class = RWConfig
+    base_model_prefix = "transformer"
+    supports_gradient_checkpointing = True
+    _no_split_modules = ["DecoderLayer"]
+
+    def __init__(self, *inputs, **kwargs):
+        super().__init__(*inputs, **kwargs)
+
+    def _init_weights(self, module: nn.Module):
+        """Initialize the weights."""
+        if isinstance(module, nn.Linear) or isinstance(module, Linear):
+            # Slightly different from the TF version which uses truncated_normal for initialization
+            # cf https://github.com/pytorch/pytorch/pull/5617
+            module.weight.data.normal_(mean=0.0, std=self.config.initializer_range)
+            if module.bias is not None:
+                module.bias.data.zero_()
+        elif isinstance(module, nn.Embedding):
+            module.weight.data.normal_(mean=0.0, std=self.config.initializer_range)
+            if module.padding_idx is not None:
+                module.weight.data[module.padding_idx].zero_()
+        elif isinstance(module, LayerNorm):
+            module.bias.data.zero_()
+            module.weight.data.fill_(1.0)
+
+    def _set_gradient_checkpointing(self, module: nn.Module, value: bool = False):
+        if isinstance(module, RWModel):
+            module.gradient_checkpointing = value
+
+    @staticmethod
+    def _convert_to_standard_cache(
+        past_key_value: Tuple[Tuple[torch.Tensor, torch.Tensor]], batch_size: int
+    ) -> Tuple[Tuple[torch.Tensor, torch.Tensor]]:
+        """
+        Standardizes the format of the cache so as to match most implementations, i.e. to tuple(tuple([batch_size,
+        num_heads, ...]))
+        """
+        batch_size_times_num_heads, head_dim, seq_length = past_key_value[0][0].shape
+        num_heads = batch_size_times_num_heads // batch_size
+        # key: [batch_size * num_heads, head_dim, seq_length] -> [batch_size, num_heads, head_dim, seq_length]
+        # value: [batch_size * num_heads, seq_length, head_dim] -> [batch_size, num_heads, seq_length, head_dim]
+        return tuple(
+            (
+                layer_past[0].view(batch_size, num_heads, head_dim, seq_length),
+                layer_past[1].view(batch_size, num_heads, seq_length, head_dim),
+            )
+            for layer_past in past_key_value
+        )
+
+    @staticmethod
+    def _convert_to_rw_cache(
+        past_key_value: Tuple[Tuple[torch.Tensor, torch.Tensor]]
+    ) -> Tuple[Tuple[torch.Tensor, torch.Tensor]]:
+        batch_size, num_heads, head_dim, seq_length = past_key_value[0][0].shape
+        batch_size_times_num_heads = batch_size * num_heads
+        # key:  [batch_size, num_heads, head_dim, seq_length] -> [batch_size * num_heads, head_dim, seq_length]
+        # value: [batch_size, num_heads, seq_length, head_dim] -> [batch_size * num_heads, seq_length, head_dim]
+        return tuple(
+            (
+                layer_past[0].view(batch_size_times_num_heads, head_dim, seq_length),
+                layer_past[1].view(batch_size_times_num_heads, seq_length, head_dim),
+            )
+            for layer_past in past_key_value
+        )
+
+
+class RWModel(RWPreTrainedModel):
+    def __init__(self, config: RWConfig):
+        super().__init__(config)
+
+        self.embed_dim = config.hidden_size
+        self.num_heads = config.n_head
+        self.alibi = config.alibi
+
+        # Embedding + LN Embedding
+        self.word_embeddings = nn.Embedding(config.vocab_size, self.embed_dim)
+
+        # Transformer blocks
+        self.h = nn.ModuleList([DecoderLayer(config) for _ in range(config.num_hidden_layers)])
+
+        # Final Layer Norm
+        self.ln_f = LayerNorm(self.embed_dim, eps=config.layer_norm_epsilon)
+
+        self.gradient_checkpointing = False
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def get_input_embeddings(self):
+        return self.word_embeddings
+
+    def _prepare_attn_mask(
+        self, attention_mask: torch.Tensor, input_shape: Tuple[int, int], past_key_values_length: int
+    ) -> torch.BoolTensor:
+        # create causal mask
+        # [batch_size, seq_length] -> [batch_size, 1, tgt_length, src_length]
+        combined_attention_mask = None
+        device = attention_mask.device
+        _, src_length = input_shape
+
+        if src_length > 1:
+            combined_attention_mask = _make_causal_mask(
+                input_shape, device=device, past_key_values_length=past_key_values_length
+            )
+
+        # [batch_size, seq_length] -> [batch_size, 1, tgt_length, src_length]
+        expanded_attn_mask = _expand_mask(attention_mask, tgt_length=src_length)
+        combined_attention_mask = (
+            expanded_attn_mask if combined_attention_mask is None else expanded_attn_mask | combined_attention_mask
+        )
+
+        return combined_attention_mask
+
+    def set_input_embeddings(self, new_embeddings: torch.Tensor):
+        self.word_embeddings = new_embeddings
+
+    def forward(
+        self,
+        input_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[Tuple[Tuple[torch.Tensor, torch.Tensor], ...]] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        head_mask: Optional[torch.LongTensor] = None,
+        inputs_embeds: Optional[torch.LongTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+        **deprecated_arguments,
+    ) -> Union[Tuple[torch.Tensor, ...], BaseModelOutputWithPastAndCrossAttentions]:
+        if deprecated_arguments.pop("position_ids", False) is not False:
+            # `position_ids` could have been `torch.Tensor` or `None` so defaulting pop to `False` allows to detect if users were passing explicitly `None`
+            warnings.warn(
+                "`position_ids` have no functionality in BLOOM and will be removed in v5.0.0. You can safely ignore"
+                " passing `position_ids`.",
+                FutureWarning,
+            )
+        if len(deprecated_arguments) > 0:
+            raise ValueError(f"Got unexpected arguments: {deprecated_arguments}")
+
+        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:
+            batch_size, seq_length = input_ids.shape
+        elif inputs_embeds is not None:
+            batch_size, seq_length, _ = inputs_embeds.shape
+        else:
+            raise ValueError("You have to specify either input_ids or inputs_embeds")
+
+        if past_key_values is None:
+            past_key_values = tuple([None] * len(self.h))
+
+        # Prepare head mask if needed
+        # 1.0 in head_mask indicate we keep the head
+        # attention_probs has shape batch_size x num_heads x N x N
+        # head_mask has shape n_layer x batch x num_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.word_embeddings(input_ids)
+
+        hidden_states = inputs_embeds
+
+        presents = () if use_cache else None
+        all_self_attentions = () if output_attentions else None
+        all_hidden_states = () if output_hidden_states else None
+
+        # Compute alibi tensor: check build_alibi_tensor documentation
+        seq_length_with_past = seq_length
+        past_key_values_length = 0
+        if past_key_values[0] is not None:
+            past_key_values_length = past_key_values[0][0].shape[2]
+            seq_length_with_past = seq_length_with_past + past_key_values_length
+        if attention_mask is None:
+            attention_mask = torch.ones((batch_size, seq_length_with_past), device=hidden_states.device)
+        else:
+            attention_mask = attention_mask.to(hidden_states.device)
+
+        if self.alibi:
+            alibi = build_alibi_tensor(attention_mask, self.num_heads, dtype=hidden_states.dtype)
+        else:
+            alibi = None
+
+        causal_mask = self._prepare_attn_mask(
+            attention_mask,
+            input_shape=(batch_size, seq_length),
+            past_key_values_length=past_key_values_length,
+        )
+
+        for i, (block, layer_past) in enumerate(zip(self.h, past_key_values)):
+
+            if output_hidden_states:
+                all_hidden_states = all_hidden_states + (hidden_states,)
+
+            if self.gradient_checkpointing and self.training:
+
+                if use_cache:
+                    logger.warning(
+                        "`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`..."
+                    )
+                    use_cache = False
+
+                def create_custom_forward(module):
+                    def custom_forward(*inputs):
+                        # None for past_key_value
+                        return module(*inputs, use_cache=use_cache, output_attentions=output_attentions)
+
+                    return custom_forward
+
+                outputs = torch.utils.checkpoint.checkpoint(
+                    create_custom_forward(block),
+                    hidden_states,
+                    alibi,
+                    causal_mask,
+                    head_mask[i],
+                )
+            else:
+                outputs = block(
+                    hidden_states,
+                    layer_past=layer_past,
+                    attention_mask=causal_mask,
+                    head_mask=head_mask[i],
+                    use_cache=use_cache,
+                    output_attentions=output_attentions,
+                    alibi=alibi,
+                )
+
+            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],)
+
+        # Add last hidden state
+        hidden_states = self.ln_f(hidden_states)
+
+        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] 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,
+        )
+
+
+class RWForCausalLM(RWPreTrainedModel):
+    _keys_to_ignore_on_load_missing = [r"h.*.self_attention.scale_mask_softmax.causal_mask", r"lm_head.weight"]
+
+    def __init__(self, config: RWConfig):
+        super().__init__(config)
+        self.transformer = RWModel(config)
+        self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, bias=False)
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def get_output_embeddings(self):
+        return self.lm_head
+
+    def set_output_embeddings(self, new_embeddings: torch.Tensor):
+        self.lm_head = new_embeddings
+
+    def prepare_inputs_for_generation(
+        self,
+        input_ids: torch.LongTensor,
+        past: Optional[torch.Tensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        **kwargs,
+    ) -> dict:
+        # only last token for input_ids if past is not None
+        if past:
+            input_ids = input_ids[:, -1].unsqueeze(-1)
+
+            # the cache may be in the stardard format (e.g. in contrastive search), convert to our's format if needed
+            if past[0][0].shape[0] == input_ids.shape[0]:
+                past = self._convert_to_rw_cache(past)
+
+        return {
+            "input_ids": input_ids,
+            "past_key_values": past,
+            "use_cache": kwargs.get("use_cache"),
+            "attention_mask": attention_mask,
+        }
+
+    def forward(
+        self,
+        input_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[Tuple[Tuple[torch.Tensor, torch.Tensor], ...]] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        head_mask: Optional[torch.Tensor] = None,
+        inputs_embeds: Optional[torch.Tensor] = None,
+        labels: Optional[torch.Tensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+        **deprecated_arguments,
+    ) -> Union[Tuple[torch.Tensor], CausalLMOutputWithCrossAttentions]:
+        r"""
+        labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
+            Labels for language modeling. Note that the labels **are shifted** inside the model, i.e. you can set
+            `labels = input_ids` Indices are selected in `[-100, 0, ..., config.vocab_size]` All labels set to `-100`
+            are ignored (masked), the loss is only computed for labels in `[0, ..., config.vocab_size]`
+        """
+        if deprecated_arguments.pop("position_ids", False) is not False:
+            # `position_ids` could have been `torch.Tensor` or `None` so defaulting pop to `False` allows to detect if users were passing explicitly `None`
+            warnings.warn(
+                "`position_ids` have no functionality in BLOOM and will be removed in v5.0.0. You can safely ignore"
+                " passing `position_ids`.",
+                FutureWarning,
+            )
+        if len(deprecated_arguments) > 0:
+            raise ValueError(f"Got unexpected arguments: {deprecated_arguments}")
+
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        transformer_outputs = self.transformer(
+            input_ids,
+            past_key_values=past_key_values,
+            attention_mask=attention_mask,
+            head_mask=head_mask,
+            inputs_embeds=inputs_embeds,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+        hidden_states = transformer_outputs[0]
+
+        lm_logits = self.lm_head(hidden_states)
+
+        loss = None
+        if labels is not None:
+            # Shift so that tokens < n predict n
+            shift_logits = lm_logits[..., :-1, :].contiguous()
+            shift_labels = labels[..., 1:].contiguous()
+            batch_size, seq_length, vocab_size = shift_logits.shape
+            # Flatten the tokens
+            loss_fct = CrossEntropyLoss()
+            loss = loss_fct(
+                shift_logits.view(batch_size * seq_length, vocab_size), shift_labels.view(batch_size * seq_length)
+            )
+
+        if not return_dict:
+            output = (lm_logits,) + transformer_outputs[1:]
+            return ((loss,) + output) if loss is not None else output
+
+        return CausalLMOutputWithCrossAttentions(
+            loss=loss,
+            logits=lm_logits,
+            past_key_values=transformer_outputs.past_key_values,
+            hidden_states=transformer_outputs.hidden_states,
+            attentions=transformer_outputs.attentions,
+        )
+
+    def _reorder_cache(
+        self, past: Tuple[Tuple[torch.Tensor, torch.Tensor], ...], beam_idx: torch.LongTensor
+    ) -> Tuple[Tuple[torch.Tensor, torch.Tensor], ...]:
+        """
+        This function is used to re-order the `past_key_values` cache if [`~PreTrainedModel.beam_search`] or
+        [`~PreTrainedModel.beam_sample`] is called. This is required to match `past_key_values` with the correct
+        beam_idx at every generation step.
+
+        Output shares the same memory storage as `past`.
+        """
+        standardized_past = self._convert_to_standard_cache(past, batch_size=len(beam_idx))
+
+        # Get a copy of `beam_idx` on all the devices where we need those indices.
+        device_to_beam_idx = {
+            past_state.device: beam_idx.to(past_state.device) for layer_past in past for past_state in layer_past
+        }
+        reordered_past = tuple(
+            (
+                layer_past[0].index_select(0, device_to_beam_idx[layer_past[0].device]),
+                layer_past[1].index_select(0, device_to_beam_idx[layer_past[0].device]),
+            )
+            for layer_past in standardized_past
+        )
+        return self._convert_to_rw_cache(reordered_past)
+
+
+class RWForSequenceClassification(RWPreTrainedModel):
+    _keys_to_ignore_on_load_missing = [r"h.*.self_attention.scale_mask_softmax.causal_mask", r"lm_head.weight"]
+
+    def __init__(self, config: RWConfig):
+        super().__init__(config)
+        self.num_labels = config.num_labels
+        self.transformer = RWModel(config)
+        self.score = nn.Linear(config.hidden_size, config.num_labels, bias=False)
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def forward(
+        self,
+        input_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[Tuple[Tuple[torch.Tensor, torch.Tensor], ...]] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        head_mask: Optional[torch.Tensor] = None,
+        inputs_embeds: Optional[torch.Tensor] = None,
+        labels: Optional[torch.Tensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+        **deprecated_arguments,
+    ) -> Union[Tuple[torch.Tensor], SequenceClassifierOutputWithPast]:
+        r"""
+        labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
+            Labels for computing the sequence classification/regression loss. Indices should be in `[0, ...,
+            config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If
+            `config.num_labels > 1` a classification loss is computed (Cross-Entropy).
+        """
+        if deprecated_arguments.pop("position_ids", False) is not False:
+            # `position_ids` could have been `torch.Tensor` or `None` so defaulting pop to `False` allows to detect if users were passing explicitly `None`
+            warnings.warn(
+                "`position_ids` have no functionality in BLOOM and will be removed in v5.0.0. You can safely ignore"
+                " passing `position_ids`.",
+                FutureWarning,
+            )
+        if len(deprecated_arguments) > 0:
+            raise ValueError(f"Got unexpected arguments: {deprecated_arguments}")
+
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        transformer_outputs = self.transformer(
+            input_ids,
+            past_key_values=past_key_values,
+            attention_mask=attention_mask,
+            head_mask=head_mask,
+            inputs_embeds=inputs_embeds,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+
+        hidden_states = transformer_outputs[0]
+        logits = self.score(hidden_states)
+
+        if input_ids is not None:
+            batch_size = input_ids.shape[0]
+        else:
+            batch_size = inputs_embeds.shape[0]
+
+        if self.config.pad_token_id is None and batch_size != 1:
+            raise ValueError("Cannot handle batch sizes > 1 if no padding token is defined.")
+        if self.config.pad_token_id is None:
+            sequence_lengths = -1
+        else:
+            if input_ids is not None:
+                sequence_lengths = torch.ne(input_ids, self.config.pad_token_id).sum(dim=-1) - 1
+            else:
+                sequence_lengths = -1
+                logger.warning(
+                    f"{self.__class__.__name__} will not detect padding tokens in `inputs_embeds`. Results may be "
+                    "unexpected if using padding tokens in conjunction with `inputs_embeds.`"
+                )
+
+        pooled_logits = logits[torch.arange(batch_size, device=logits.device), sequence_lengths]
+
+        loss = None
+        if labels is not None:
+            if self.config.problem_type is None:
+                if self.num_labels == 1:
+                    self.config.problem_type = "regression"
+                elif self.num_labels > 1 and (labels.dtype == torch.long or labels.dtype == torch.int):
+                    self.config.problem_type = "single_label_classification"
+                else:
+                    self.config.problem_type = "multi_label_classification"
+
+            if self.config.problem_type == "regression":
+                loss_fct = MSELoss()
+                if self.num_labels == 1:
+                    loss = loss_fct(pooled_logits.squeeze(), labels.squeeze())
+                else:
+                    loss = loss_fct(pooled_logits, labels)
+            elif self.config.problem_type == "single_label_classification":
+                loss_fct = CrossEntropyLoss()
+                loss = loss_fct(pooled_logits, labels)
+            elif self.config.problem_type == "multi_label_classification":
+                loss_fct = BCEWithLogitsLoss()
+                loss = loss_fct(pooled_logits, labels)
+        if not return_dict:
+            output = (pooled_logits,) + transformer_outputs[1:]
+            return ((loss,) + output) if loss is not None else output
+
+        return SequenceClassifierOutputWithPast(
+            loss=loss,
+            logits=pooled_logits,
+            past_key_values=transformer_outputs.past_key_values,
+            hidden_states=transformer_outputs.hidden_states,
+            attentions=transformer_outputs.attentions,
+        )
+
+
+class RWForTokenClassification(RWPreTrainedModel):
+    _keys_to_ignore_on_load_missing = [r"h.*.self_attention.scale_mask_softmax.causal_mask", r"lm_head.weight"]
+
+    def __init__(self, config: RWConfig):
+        super().__init__(config)
+        self.num_labels = config.num_labels
+
+        self.transformer = RWModel(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)
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def forward(
+        self,
+        input_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[Tuple[Tuple[torch.Tensor, torch.Tensor], ...]] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        head_mask: Optional[torch.Tensor] = None,
+        inputs_embeds: Optional[torch.Tensor] = None,
+        labels: Optional[torch.Tensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+        **deprecated_arguments,
+    ) -> Union[Tuple[torch.Tensor], TokenClassifierOutput]:
+        r"""
+        labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
+            Labels for computing the sequence classification/regression loss. Indices should be in `[0, ...,
+            config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If
+            `config.num_labels > 1` a classification loss is computed (Cross-Entropy).
+        """
+        if deprecated_arguments.pop("position_ids", False) is not False:
+            # `position_ids` could have been `torch.Tensor` or `None` so defaulting pop to `False` allows to detect if users were passing explicitly `None`
+            warnings.warn(
+                "`position_ids` have no functionality in BLOOM and will be removed in v5.0.0. You can safely ignore"
+                " passing `position_ids`.",
+                FutureWarning,
+            )
+        if len(deprecated_arguments) > 0:
+            raise ValueError(f"Got unexpected arguments: {deprecated_arguments}")
+
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        transformer_outputs = self.transformer(
+            input_ids,
+            past_key_values=past_key_values,
+            attention_mask=attention_mask,
+            head_mask=head_mask,
+            inputs_embeds=inputs_embeds,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+
+        hidden_states = transformer_outputs[0]
+        hidden_states = self.dropout(hidden_states)
+        logits = self.classifier(hidden_states)
+
+        loss = None
+        if labels is not None:
+            batch_size, seq_length = labels.shape
+            loss_fct = CrossEntropyLoss()
+            loss = loss_fct(logits.view(batch_size * seq_length, self.num_labels), labels.view(batch_size * seq_length))
+
+        if not return_dict:
+            output = (logits,) + transformer_outputs[2:]
+            return ((loss,) + output) if loss is not None else output
+
+        return TokenClassifierOutput(
+            loss=loss,
+            logits=logits,
+            hidden_states=transformer_outputs.hidden_states,
+            attentions=transformer_outputs.attentions,
+        )
+
+
+class RWForQuestionAnswering(RWPreTrainedModel):
+    _keys_to_ignore_on_load_missing = [r"h.*.self_attention.scale_mask_softmax.causal_mask", r"lm_head.weight"]
+
+    def __init__(self, config):
+        super().__init__(config)
+        self.transformer = RWModel(config)
+        self.qa_outputs = nn.Linear(config.hidden_size, 2)
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def forward(
+        self,
+        input_ids: Optional[torch.LongTensor] = None,
+        attention_mask: Optional[torch.FloatTensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        head_mask: Optional[torch.FloatTensor] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        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,
+            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)
+            if len(end_positions.size()) > 1:
+                end_positions = end_positions.squeeze(-1)
+            # 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,
+        )

quant_config.json

@@ -0,0 +1,6 @@
+{
+    "zero_point": true,
+    "q_group_size": 128,
+    "w_bit": 4,
+    "version": "GEMM"
+}

special_tokens_map.json

@@ -0,0 +1,20 @@
+{
+  "additional_special_tokens": [
+    ">>TITLE<<",
+    ">>ABSTRACT<<",
+    ">>INTRODUCTION<<",
+    ">>SUMMARY<<",
+    ">>COMMENT<<",
+    ">>ANSWER<<",
+    ">>QUESTION<<",
+    ">>DOMAIN<<",
+    ">>PREFIX<<",
+    ">>SUFFIX<<",
+    ">>MIDDLE<<",
+    "<start_system>",
+    "<start_user>",
+    "<start_assistant>",
+    "<end_message>"
+  ],
+  "eos_token": "<end_message>"
+}

tokenizer_config.json

@@ -0,0 +1,7 @@
+{
+  "add_prefix_space": false,
+  "eos_token": "<end_message>",
+  "model_max_length": 8192,
+  "special_tokens_map_file": null,
+  "tokenizer_class": "PreTrainedTokenizerFast"
+}