Add files using upload-large-folder tool

config.json

@@ -0,0 +1,24 @@
+{
+  "architectures": [
+    "RiemannForCausalLM"
+  ],
+  "attention_dropout": 0.1,
+  "dtype": "float32",
+  "hidden_dropout": 0.1,
+  "hidden_size": 5120,
+  "intermediate_size": 20480,
+  "layer_norm_eps": 1e-05,
+  "max_position_embeddings": 2048,
+  "model_type": "riemann_openpeer",
+  "num_attention_heads": 40,
+  "num_hidden_layers": 40,
+  "tie_word_embeddings": false,
+  "transformers_version": "4.57.6",
+  "use_cache": false,
+  "use_gradient_checkpointing": true,
+  "vocab_size": 50257,
+  "auto_map": {
+    "AutoConfig": "configuration_riemann.RiemannConfig",
+    "AutoModelForCausalLM": "modeling_riemann.RiemannForCausalLM"
+  }
+}

configuration_riemann.py

@@ -0,0 +1,33 @@
+from __future__ import annotations
+from transformers import PretrainedConfig
+
+
+class RiemannConfig(PretrainedConfig):
+    model_type = "riemann_openpeer"
+
+    def __init__(
+        self,
+        vocab_size: int = 50257,
+        hidden_size: int = 768,
+        num_hidden_layers: int = 12,
+        num_attention_heads: int = 12,
+        intermediate_size: int = 3072,
+        max_position_embeddings: int = 1024,
+        layer_norm_eps: float = 1e-05,
+        hidden_dropout: float = 0.1,
+        attention_dropout: float = 0.1,
+        use_gradient_checkpointing: bool = False,
+        **kwargs,
+    ):
+        self.vocab_size = vocab_size
+        self.hidden_size = hidden_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.intermediate_size = intermediate_size
+        self.max_position_embeddings = max_position_embeddings
+        self.layer_norm_eps = layer_norm_eps
+        self.hidden_dropout = hidden_dropout
+        self.attention_dropout = attention_dropout
+        self.use_gradient_checkpointing = use_gradient_checkpointing
+        kwargs.setdefault("tie_word_embeddings", False)
+        super().__init__(**kwargs)

generation_config.json

@@ -0,0 +1,4 @@
+{
+  "_from_model_config": true,
+  "transformers_version": "4.57.6"
+}

model-00001-of-00011.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a078a2cffe24d9d7dab152885b4cedb05523f7d3a0819a7bff0d9b36ece6bafd
+size 4951762312

model-00002-of-00011.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c356bc43243a782f0cc20197932b90d873697a60b75dc3f544b07e3f2482c2cd
+size 4929358272

model-00003-of-00011.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9d4032694bb98c259a199bbb35e091f4cf481cc35ca997f45dd818b687c1270c
+size 4614703120

model-00004-of-00011.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:10bf59a488a94dc94879c121a71df4d8d0659270ec3f4bcf39e0117c983bbbe6
+size 4614724056

model-00005-of-00011.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:182b33840bbb79cb111a865483da98c71219f5eb2bb8c3eb308ea3f8cf57afd6
+size 4929358320

model-00006-of-00011.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:696d49d9f9f929b0d01be23f41a4e296cbf24cad5e20b2f4dd5ca57432d34935
+size 4929358328

model-00007-of-00011.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:527de9ef032ffe2bec824cf3d5ce6e224c634939ac516b116acbbfeeea3fe2c2
+size 4929358336

model-00008-of-00011.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9213ab94567493996178184f1a5b9bd60666a43329ef51b73d0b48d4d0b16043
+size 4929358328

model-00009-of-00011.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cac369c6f044ebc8ea4f5081a62f77b6269b03c70e787ddbb974179fadaeff48
+size 4614703168

model-00010-of-00011.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b494452ee08fcebed2859f9e812ba2ecc99defc472ffb5c06a8d14146294cd4e
+size 4614724056

model-00011-of-00011.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b9e831859ea0a4320d362006a82adfa9db751f3cd5adf8d1f1121cb3ce3fc827
+size 4385468872

model.safetensors.index.json

@@ -0,0 +1,653 @@
+{
+  "metadata": {
+    "total_parameters": 1638837760,
+    "total_size": 52442808320
+  },
+  "weight_map": {
+    "final_layernorm.bias": "model-00011-of-00011.safetensors",
+    "final_layernorm.weight": "model-00011-of-00011.safetensors",
+    "layers.0.attention.key.bias": "model-00001-of-00011.safetensors",
+    "layers.0.attention.key.weight": "model-00001-of-00011.safetensors",
+    "layers.0.attention.out.bias": "model-00001-of-00011.safetensors",
+    "layers.0.attention.out.weight": "model-00001-of-00011.safetensors",
+    "layers.0.attention.query.bias": "model-00001-of-00011.safetensors",
+    "layers.0.attention.query.weight": "model-00001-of-00011.safetensors",
+    "layers.0.attention.value.bias": "model-00001-of-00011.safetensors",
+    "layers.0.attention.value.weight": "model-00001-of-00011.safetensors",
+    "layers.0.input_layernorm.bias": "model-00001-of-00011.safetensors",
+    "layers.0.input_layernorm.weight": "model-00001-of-00011.safetensors",
+    "layers.0.mlp.dense_4h_to_h.bias": "model-00001-of-00011.safetensors",
+    "layers.0.mlp.dense_4h_to_h.weight": "model-00001-of-00011.safetensors",
+    "layers.0.mlp.dense_h_to_4h.bias": "model-00001-of-00011.safetensors",
+    "layers.0.mlp.dense_h_to_4h.weight": "model-00001-of-00011.safetensors",
+    "layers.0.post_attention_layernorm.bias": "model-00001-of-00011.safetensors",
+    "layers.0.post_attention_layernorm.weight": "model-00001-of-00011.safetensors",
+    "layers.1.attention.key.bias": "model-00001-of-00011.safetensors",
+    "layers.1.attention.key.weight": "model-00001-of-00011.safetensors",
+    "layers.1.attention.out.bias": "model-00001-of-00011.safetensors",
+    "layers.1.attention.out.weight": "model-00001-of-00011.safetensors",
+    "layers.1.attention.query.bias": "model-00001-of-00011.safetensors",
+    "layers.1.attention.query.weight": "model-00001-of-00011.safetensors",
+    "layers.1.attention.value.bias": "model-00001-of-00011.safetensors",
+    "layers.1.attention.value.weight": "model-00001-of-00011.safetensors",
+    "layers.1.input_layernorm.bias": "model-00001-of-00011.safetensors",
+    "layers.1.input_layernorm.weight": "model-00001-of-00011.safetensors",
+    "layers.1.mlp.dense_4h_to_h.bias": "model-00001-of-00011.safetensors",
+    "layers.1.mlp.dense_4h_to_h.weight": "model-00001-of-00011.safetensors",
+    "layers.1.mlp.dense_h_to_4h.bias": "model-00001-of-00011.safetensors",
+    "layers.1.mlp.dense_h_to_4h.weight": "model-00001-of-00011.safetensors",
+    "layers.1.post_attention_layernorm.bias": "model-00001-of-00011.safetensors",
+    "layers.1.post_attention_layernorm.weight": "model-00001-of-00011.safetensors",
+    "layers.10.attention.key.bias": "model-00003-of-00011.safetensors",
+    "layers.10.attention.key.weight": "model-00003-of-00011.safetensors",
+    "layers.10.attention.out.bias": "model-00003-of-00011.safetensors",
+    "layers.10.attention.out.weight": "model-00003-of-00011.safetensors",
+    "layers.10.attention.query.bias": "model-00003-of-00011.safetensors",
+    "layers.10.attention.query.weight": "model-00003-of-00011.safetensors",
+    "layers.10.attention.value.bias": "model-00003-of-00011.safetensors",
+    "layers.10.attention.value.weight": "model-00003-of-00011.safetensors",
+    "layers.10.input_layernorm.bias": "model-00004-of-00011.safetensors",
+    "layers.10.input_layernorm.weight": "model-00004-of-00011.safetensors",
+    "layers.10.mlp.dense_4h_to_h.bias": "model-00004-of-00011.safetensors",
+    "layers.10.mlp.dense_4h_to_h.weight": "model-00004-of-00011.safetensors",
+    "layers.10.mlp.dense_h_to_4h.bias": "model-00003-of-00011.safetensors",
+    "layers.10.mlp.dense_h_to_4h.weight": "model-00003-of-00011.safetensors",
+    "layers.10.post_attention_layernorm.bias": "model-00004-of-00011.safetensors",
+    "layers.10.post_attention_layernorm.weight": "model-00004-of-00011.safetensors",
+    "layers.11.attention.key.bias": "model-00004-of-00011.safetensors",
+    "layers.11.attention.key.weight": "model-00004-of-00011.safetensors",
+    "layers.11.attention.out.bias": "model-00004-of-00011.safetensors",
+    "layers.11.attention.out.weight": "model-00004-of-00011.safetensors",
+    "layers.11.attention.query.bias": "model-00004-of-00011.safetensors",
+    "layers.11.attention.query.weight": "model-00004-of-00011.safetensors",
+    "layers.11.attention.value.bias": "model-00004-of-00011.safetensors",
+    "layers.11.attention.value.weight": "model-00004-of-00011.safetensors",
+    "layers.11.input_layernorm.bias": "model-00004-of-00011.safetensors",
+    "layers.11.input_layernorm.weight": "model-00004-of-00011.safetensors",
+    "layers.11.mlp.dense_4h_to_h.bias": "model-00004-of-00011.safetensors",
+    "layers.11.mlp.dense_4h_to_h.weight": "model-00004-of-00011.safetensors",
+    "layers.11.mlp.dense_h_to_4h.bias": "model-00004-of-00011.safetensors",
+    "layers.11.mlp.dense_h_to_4h.weight": "model-00004-of-00011.safetensors",
+    "layers.11.post_attention_layernorm.bias": "model-00004-of-00011.safetensors",
+    "layers.11.post_attention_layernorm.weight": "model-00004-of-00011.safetensors",
+    "layers.12.attention.key.bias": "model-00004-of-00011.safetensors",
+    "layers.12.attention.key.weight": "model-00004-of-00011.safetensors",
+    "layers.12.attention.out.bias": "model-00004-of-00011.safetensors",
+    "layers.12.attention.out.weight": "model-00004-of-00011.safetensors",
+    "layers.12.attention.query.bias": "model-00004-of-00011.safetensors",
+    "layers.12.attention.query.weight": "model-00004-of-00011.safetensors",
+    "layers.12.attention.value.bias": "model-00004-of-00011.safetensors",
+    "layers.12.attention.value.weight": "model-00004-of-00011.safetensors",
+    "layers.12.input_layernorm.bias": "model-00004-of-00011.safetensors",
+    "layers.12.input_layernorm.weight": "model-00004-of-00011.safetensors",
+    "layers.12.mlp.dense_4h_to_h.bias": "model-00004-of-00011.safetensors",
+    "layers.12.mlp.dense_4h_to_h.weight": "model-00004-of-00011.safetensors",
+    "layers.12.mlp.dense_h_to_4h.bias": "model-00004-of-00011.safetensors",
+    "layers.12.mlp.dense_h_to_4h.weight": "model-00004-of-00011.safetensors",
+    "layers.12.post_attention_layernorm.bias": "model-00004-of-00011.safetensors",
+    "layers.12.post_attention_layernorm.weight": "model-00004-of-00011.safetensors",
+    "layers.13.attention.key.bias": "model-00004-of-00011.safetensors",
+    "layers.13.attention.key.weight": "model-00004-of-00011.safetensors",
+    "layers.13.attention.out.bias": "model-00004-of-00011.safetensors",
+    "layers.13.attention.out.weight": "model-00004-of-00011.safetensors",
+    "layers.13.attention.query.bias": "model-00004-of-00011.safetensors",
+    "layers.13.attention.query.weight": "model-00004-of-00011.safetensors",
+    "layers.13.attention.value.bias": "model-00004-of-00011.safetensors",
+    "layers.13.attention.value.weight": "model-00004-of-00011.safetensors",
+    "layers.13.input_layernorm.bias": "model-00004-of-00011.safetensors",
+    "layers.13.input_layernorm.weight": "model-00004-of-00011.safetensors",
+    "layers.13.mlp.dense_4h_to_h.bias": "model-00004-of-00011.safetensors",
+    "layers.13.mlp.dense_4h_to_h.weight": "model-00004-of-00011.safetensors",
+    "layers.13.mlp.dense_h_to_4h.bias": "model-00004-of-00011.safetensors",
+    "layers.13.mlp.dense_h_to_4h.weight": "model-00004-of-00011.safetensors",
+    "layers.13.post_attention_layernorm.bias": "model-00004-of-00011.safetensors",
+    "layers.13.post_attention_layernorm.weight": "model-00004-of-00011.safetensors",
+    "layers.14.attention.key.bias": "model-00004-of-00011.safetensors",
+    "layers.14.attention.key.weight": "model-00004-of-00011.safetensors",
+    "layers.14.attention.out.bias": "model-00004-of-00011.safetensors",
+    "layers.14.attention.out.weight": "model-00004-of-00011.safetensors",
+    "layers.14.attention.query.bias": "model-00004-of-00011.safetensors",
+    "layers.14.attention.query.weight": "model-00004-of-00011.safetensors",
+    "layers.14.attention.value.bias": "model-00004-of-00011.safetensors",
+    "layers.14.attention.value.weight": "model-00004-of-00011.safetensors",
+    "layers.14.input_layernorm.bias": "model-00005-of-00011.safetensors",
+    "layers.14.input_layernorm.weight": "model-00005-of-00011.safetensors",
+    "layers.14.mlp.dense_4h_to_h.bias": "model-00005-of-00011.safetensors",
+    "layers.14.mlp.dense_4h_to_h.weight": "model-00005-of-00011.safetensors",
+    "layers.14.mlp.dense_h_to_4h.bias": "model-00005-of-00011.safetensors",
+    "layers.14.mlp.dense_h_to_4h.weight": "model-00005-of-00011.safetensors",
+    "layers.14.post_attention_layernorm.bias": "model-00005-of-00011.safetensors",
+    "layers.14.post_attention_layernorm.weight": "model-00005-of-00011.safetensors",
+    "layers.15.attention.key.bias": "model-00005-of-00011.safetensors",
+    "layers.15.attention.key.weight": "model-00005-of-00011.safetensors",
+    "layers.15.attention.out.bias": "model-00005-of-00011.safetensors",
+    "layers.15.attention.out.weight": "model-00005-of-00011.safetensors",
+    "layers.15.attention.query.bias": "model-00005-of-00011.safetensors",
+    "layers.15.attention.query.weight": "model-00005-of-00011.safetensors",
+    "layers.15.attention.value.bias": "model-00005-of-00011.safetensors",
+    "layers.15.attention.value.weight": "model-00005-of-00011.safetensors",
+    "layers.15.input_layernorm.bias": "model-00005-of-00011.safetensors",
+    "layers.15.input_layernorm.weight": "model-00005-of-00011.safetensors",
+    "layers.15.mlp.dense_4h_to_h.bias": "model-00005-of-00011.safetensors",
+    "layers.15.mlp.dense_4h_to_h.weight": "model-00005-of-00011.safetensors",
+    "layers.15.mlp.dense_h_to_4h.bias": "model-00005-of-00011.safetensors",
+    "layers.15.mlp.dense_h_to_4h.weight": "model-00005-of-00011.safetensors",
+    "layers.15.post_attention_layernorm.bias": "model-00005-of-00011.safetensors",
+    "layers.15.post_attention_layernorm.weight": "model-00005-of-00011.safetensors",
+    "layers.16.attention.key.bias": "model-00005-of-00011.safetensors",
+    "layers.16.attention.key.weight": "model-00005-of-00011.safetensors",
+    "layers.16.attention.out.bias": "model-00005-of-00011.safetensors",
+    "layers.16.attention.out.weight": "model-00005-of-00011.safetensors",
+    "layers.16.attention.query.bias": "model-00005-of-00011.safetensors",
+    "layers.16.attention.query.weight": "model-00005-of-00011.safetensors",
+    "layers.16.attention.value.bias": "model-00005-of-00011.safetensors",
+    "layers.16.attention.value.weight": "model-00005-of-00011.safetensors",
+    "layers.16.input_layernorm.bias": "model-00005-of-00011.safetensors",
+    "layers.16.input_layernorm.weight": "model-00005-of-00011.safetensors",
+    "layers.16.mlp.dense_4h_to_h.bias": "model-00005-of-00011.safetensors",
+    "layers.16.mlp.dense_4h_to_h.weight": "model-00005-of-00011.safetensors",
+    "layers.16.mlp.dense_h_to_4h.bias": "model-00005-of-00011.safetensors",
+    "layers.16.mlp.dense_h_to_4h.weight": "model-00005-of-00011.safetensors",
+    "layers.16.post_attention_layernorm.bias": "model-00005-of-00011.safetensors",
+    "layers.16.post_attention_layernorm.weight": "model-00005-of-00011.safetensors",
+    "layers.17.attention.key.bias": "model-00005-of-00011.safetensors",
+    "layers.17.attention.key.weight": "model-00005-of-00011.safetensors",
+    "layers.17.attention.out.bias": "model-00005-of-00011.safetensors",
+    "layers.17.attention.out.weight": "model-00005-of-00011.safetensors",
+    "layers.17.attention.query.bias": "model-00005-of-00011.safetensors",
+    "layers.17.attention.query.weight": "model-00005-of-00011.safetensors",
+    "layers.17.attention.value.bias": "model-00005-of-00011.safetensors",
+    "layers.17.attention.value.weight": "model-00005-of-00011.safetensors",
+    "layers.17.input_layernorm.bias": "model-00005-of-00011.safetensors",
+    "layers.17.input_layernorm.weight": "model-00005-of-00011.safetensors",
+    "layers.17.mlp.dense_4h_to_h.bias": "model-00005-of-00011.safetensors",
+    "layers.17.mlp.dense_4h_to_h.weight": "model-00005-of-00011.safetensors",
+    "layers.17.mlp.dense_h_to_4h.bias": "model-00005-of-00011.safetensors",
+    "layers.17.mlp.dense_h_to_4h.weight": "model-00005-of-00011.safetensors",
+    "layers.17.post_attention_layernorm.bias": "model-00005-of-00011.safetensors",
+    "layers.17.post_attention_layernorm.weight": "model-00005-of-00011.safetensors",
+    "layers.18.attention.key.bias": "model-00005-of-00011.safetensors",
+    "layers.18.attention.key.weight": "model-00005-of-00011.safetensors",
+    "layers.18.attention.out.bias": "model-00006-of-00011.safetensors",
+    "layers.18.attention.out.weight": "model-00006-of-00011.safetensors",
+    "layers.18.attention.query.bias": "model-00005-of-00011.safetensors",
+    "layers.18.attention.query.weight": "model-00005-of-00011.safetensors",
+    "layers.18.attention.value.bias": "model-00005-of-00011.safetensors",
+    "layers.18.attention.value.weight": "model-00005-of-00011.safetensors",
+    "layers.18.input_layernorm.bias": "model-00006-of-00011.safetensors",
+    "layers.18.input_layernorm.weight": "model-00006-of-00011.safetensors",
+    "layers.18.mlp.dense_4h_to_h.bias": "model-00006-of-00011.safetensors",
+    "layers.18.mlp.dense_4h_to_h.weight": "model-00006-of-00011.safetensors",
+    "layers.18.mlp.dense_h_to_4h.bias": "model-00006-of-00011.safetensors",
+    "layers.18.mlp.dense_h_to_4h.weight": "model-00006-of-00011.safetensors",
+    "layers.18.post_attention_layernorm.bias": "model-00006-of-00011.safetensors",
+    "layers.18.post_attention_layernorm.weight": "model-00006-of-00011.safetensors",
+    "layers.19.attention.key.bias": "model-00006-of-00011.safetensors",
+    "layers.19.attention.key.weight": "model-00006-of-00011.safetensors",
+    "layers.19.attention.out.bias": "model-00006-of-00011.safetensors",
+    "layers.19.attention.out.weight": "model-00006-of-00011.safetensors",
+    "layers.19.attention.query.bias": "model-00006-of-00011.safetensors",
+    "layers.19.attention.query.weight": "model-00006-of-00011.safetensors",
+    "layers.19.attention.value.bias": "model-00006-of-00011.safetensors",
+    "layers.19.attention.value.weight": "model-00006-of-00011.safetensors",
+    "layers.19.input_layernorm.bias": "model-00006-of-00011.safetensors",
+    "layers.19.input_layernorm.weight": "model-00006-of-00011.safetensors",
+    "layers.19.mlp.dense_4h_to_h.bias": "model-00006-of-00011.safetensors",
+    "layers.19.mlp.dense_4h_to_h.weight": "model-00006-of-00011.safetensors",
+    "layers.19.mlp.dense_h_to_4h.bias": "model-00006-of-00011.safetensors",
+    "layers.19.mlp.dense_h_to_4h.weight": "model-00006-of-00011.safetensors",
+    "layers.19.post_attention_layernorm.bias": "model-00006-of-00011.safetensors",
+    "layers.19.post_attention_layernorm.weight": "model-00006-of-00011.safetensors",
+    "layers.2.attention.key.bias": "model-00001-of-00011.safetensors",
+    "layers.2.attention.key.weight": "model-00001-of-00011.safetensors",
+    "layers.2.attention.out.bias": "model-00001-of-00011.safetensors",
+    "layers.2.attention.out.weight": "model-00001-of-00011.safetensors",
+    "layers.2.attention.query.bias": "model-00001-of-00011.safetensors",
+    "layers.2.attention.query.weight": "model-00001-of-00011.safetensors",
+    "layers.2.attention.value.bias": "model-00001-of-00011.safetensors",
+    "layers.2.attention.value.weight": "model-00001-of-00011.safetensors",
+    "layers.2.input_layernorm.bias": "model-00001-of-00011.safetensors",
+    "layers.2.input_layernorm.weight": "model-00001-of-00011.safetensors",
+    "layers.2.mlp.dense_4h_to_h.bias": "model-00001-of-00011.safetensors",
+    "layers.2.mlp.dense_4h_to_h.weight": "model-00001-of-00011.safetensors",
+    "layers.2.mlp.dense_h_to_4h.bias": "model-00001-of-00011.safetensors",
+    "layers.2.mlp.dense_h_to_4h.weight": "model-00001-of-00011.safetensors",
+    "layers.2.post_attention_layernorm.bias": "model-00001-of-00011.safetensors",
+    "layers.2.post_attention_layernorm.weight": "model-00001-of-00011.safetensors",
+    "layers.20.attention.key.bias": "model-00006-of-00011.safetensors",
+    "layers.20.attention.key.weight": "model-00006-of-00011.safetensors",
+    "layers.20.attention.out.bias": "model-00006-of-00011.safetensors",
+    "layers.20.attention.out.weight": "model-00006-of-00011.safetensors",
+    "layers.20.attention.query.bias": "model-00006-of-00011.safetensors",
+    "layers.20.attention.query.weight": "model-00006-of-00011.safetensors",
+    "layers.20.attention.value.bias": "model-00006-of-00011.safetensors",
+    "layers.20.attention.value.weight": "model-00006-of-00011.safetensors",
+    "layers.20.input_layernorm.bias": "model-00006-of-00011.safetensors",
+    "layers.20.input_layernorm.weight": "model-00006-of-00011.safetensors",
+    "layers.20.mlp.dense_4h_to_h.bias": "model-00006-of-00011.safetensors",
+    "layers.20.mlp.dense_4h_to_h.weight": "model-00006-of-00011.safetensors",
+    "layers.20.mlp.dense_h_to_4h.bias": "model-00006-of-00011.safetensors",
+    "layers.20.mlp.dense_h_to_4h.weight": "model-00006-of-00011.safetensors",
+    "layers.20.post_attention_layernorm.bias": "model-00006-of-00011.safetensors",
+    "layers.20.post_attention_layernorm.weight": "model-00006-of-00011.safetensors",
+    "layers.21.attention.key.bias": "model-00006-of-00011.safetensors",
+    "layers.21.attention.key.weight": "model-00006-of-00011.safetensors",
+    "layers.21.attention.out.bias": "model-00006-of-00011.safetensors",
+    "layers.21.attention.out.weight": "model-00006-of-00011.safetensors",
+    "layers.21.attention.query.bias": "model-00006-of-00011.safetensors",
+    "layers.21.attention.query.weight": "model-00006-of-00011.safetensors",
+    "layers.21.attention.value.bias": "model-00006-of-00011.safetensors",
+    "layers.21.attention.value.weight": "model-00006-of-00011.safetensors",
+    "layers.21.input_layernorm.bias": "model-00006-of-00011.safetensors",
+    "layers.21.input_layernorm.weight": "model-00006-of-00011.safetensors",
+    "layers.21.mlp.dense_4h_to_h.bias": "model-00006-of-00011.safetensors",
+    "layers.21.mlp.dense_4h_to_h.weight": "model-00006-of-00011.safetensors",
+    "layers.21.mlp.dense_h_to_4h.bias": "model-00006-of-00011.safetensors",
+    "layers.21.mlp.dense_h_to_4h.weight": "model-00006-of-00011.safetensors",
+    "layers.21.post_attention_layernorm.bias": "model-00006-of-00011.safetensors",
+    "layers.21.post_attention_layernorm.weight": "model-00006-of-00011.safetensors",
+    "layers.22.attention.key.bias": "model-00006-of-00011.safetensors",
+    "layers.22.attention.key.weight": "model-00006-of-00011.safetensors",
+    "layers.22.attention.out.bias": "model-00007-of-00011.safetensors",
+    "layers.22.attention.out.weight": "model-00007-of-00011.safetensors",
+    "layers.22.attention.query.bias": "model-00006-of-00011.safetensors",
+    "layers.22.attention.query.weight": "model-00006-of-00011.safetensors",
+    "layers.22.attention.value.bias": "model-00007-of-00011.safetensors",
+    "layers.22.attention.value.weight": "model-00007-of-00011.safetensors",
+    "layers.22.input_layernorm.bias": "model-00007-of-00011.safetensors",
+    "layers.22.input_layernorm.weight": "model-00007-of-00011.safetensors",
+    "layers.22.mlp.dense_4h_to_h.bias": "model-00007-of-00011.safetensors",
+    "layers.22.mlp.dense_4h_to_h.weight": "model-00007-of-00011.safetensors",
+    "layers.22.mlp.dense_h_to_4h.bias": "model-00007-of-00011.safetensors",
+    "layers.22.mlp.dense_h_to_4h.weight": "model-00007-of-00011.safetensors",
+    "layers.22.post_attention_layernorm.bias": "model-00007-of-00011.safetensors",
+    "layers.22.post_attention_layernorm.weight": "model-00007-of-00011.safetensors",
+    "layers.23.attention.key.bias": "model-00007-of-00011.safetensors",
+    "layers.23.attention.key.weight": "model-00007-of-00011.safetensors",
+    "layers.23.attention.out.bias": "model-00007-of-00011.safetensors",
+    "layers.23.attention.out.weight": "model-00007-of-00011.safetensors",
+    "layers.23.attention.query.bias": "model-00007-of-00011.safetensors",
+    "layers.23.attention.query.weight": "model-00007-of-00011.safetensors",
+    "layers.23.attention.value.bias": "model-00007-of-00011.safetensors",
+    "layers.23.attention.value.weight": "model-00007-of-00011.safetensors",
+    "layers.23.input_layernorm.bias": "model-00007-of-00011.safetensors",
+    "layers.23.input_layernorm.weight": "model-00007-of-00011.safetensors",
+    "layers.23.mlp.dense_4h_to_h.bias": "model-00007-of-00011.safetensors",
+    "layers.23.mlp.dense_4h_to_h.weight": "model-00007-of-00011.safetensors",
+    "layers.23.mlp.dense_h_to_4h.bias": "model-00007-of-00011.safetensors",
+    "layers.23.mlp.dense_h_to_4h.weight": "model-00007-of-00011.safetensors",
+    "layers.23.post_attention_layernorm.bias": "model-00007-of-00011.safetensors",
+    "layers.23.post_attention_layernorm.weight": "model-00007-of-00011.safetensors",
+    "layers.24.attention.key.bias": "model-00007-of-00011.safetensors",
+    "layers.24.attention.key.weight": "model-00007-of-00011.safetensors",
+    "layers.24.attention.out.bias": "model-00007-of-00011.safetensors",
+    "layers.24.attention.out.weight": "model-00007-of-00011.safetensors",
+    "layers.24.attention.query.bias": "model-00007-of-00011.safetensors",
+    "layers.24.attention.query.weight": "model-00007-of-00011.safetensors",
+    "layers.24.attention.value.bias": "model-00007-of-00011.safetensors",
+    "layers.24.attention.value.weight": "model-00007-of-00011.safetensors",
+    "layers.24.input_layernorm.bias": "model-00007-of-00011.safetensors",
+    "layers.24.input_layernorm.weight": "model-00007-of-00011.safetensors",
+    "layers.24.mlp.dense_4h_to_h.bias": "model-00007-of-00011.safetensors",
+    "layers.24.mlp.dense_4h_to_h.weight": "model-00007-of-00011.safetensors",
+    "layers.24.mlp.dense_h_to_4h.bias": "model-00007-of-00011.safetensors",
+    "layers.24.mlp.dense_h_to_4h.weight": "model-00007-of-00011.safetensors",
+    "layers.24.post_attention_layernorm.bias": "model-00007-of-00011.safetensors",
+    "layers.24.post_attention_layernorm.weight": "model-00007-of-00011.safetensors",
+    "layers.25.attention.key.bias": "model-00007-of-00011.safetensors",
+    "layers.25.attention.key.weight": "model-00007-of-00011.safetensors",
+    "layers.25.attention.out.bias": "model-00007-of-00011.safetensors",
+    "layers.25.attention.out.weight": "model-00007-of-00011.safetensors",
+    "layers.25.attention.query.bias": "model-00007-of-00011.safetensors",
+    "layers.25.attention.query.weight": "model-00007-of-00011.safetensors",
+    "layers.25.attention.value.bias": "model-00007-of-00011.safetensors",
+    "layers.25.attention.value.weight": "model-00007-of-00011.safetensors",
+    "layers.25.input_layernorm.bias": "model-00007-of-00011.safetensors",
+    "layers.25.input_layernorm.weight": "model-00007-of-00011.safetensors",
+    "layers.25.mlp.dense_4h_to_h.bias": "model-00007-of-00011.safetensors",
+    "layers.25.mlp.dense_4h_to_h.weight": "model-00007-of-00011.safetensors",
+    "layers.25.mlp.dense_h_to_4h.bias": "model-00007-of-00011.safetensors",
+    "layers.25.mlp.dense_h_to_4h.weight": "model-00007-of-00011.safetensors",
+    "layers.25.post_attention_layernorm.bias": "model-00007-of-00011.safetensors",
+    "layers.25.post_attention_layernorm.weight": "model-00007-of-00011.safetensors",
+    "layers.26.attention.key.bias": "model-00008-of-00011.safetensors",
+    "layers.26.attention.key.weight": "model-00008-of-00011.safetensors",
+    "layers.26.attention.out.bias": "model-00008-of-00011.safetensors",
+    "layers.26.attention.out.weight": "model-00008-of-00011.safetensors",
+    "layers.26.attention.query.bias": "model-00007-of-00011.safetensors",
+    "layers.26.attention.query.weight": "model-00007-of-00011.safetensors",
+    "layers.26.attention.value.bias": "model-00008-of-00011.safetensors",
+    "layers.26.attention.value.weight": "model-00008-of-00011.safetensors",
+    "layers.26.input_layernorm.bias": "model-00008-of-00011.safetensors",
+    "layers.26.input_layernorm.weight": "model-00008-of-00011.safetensors",
+    "layers.26.mlp.dense_4h_to_h.bias": "model-00008-of-00011.safetensors",
+    "layers.26.mlp.dense_4h_to_h.weight": "model-00008-of-00011.safetensors",
+    "layers.26.mlp.dense_h_to_4h.bias": "model-00008-of-00011.safetensors",
+    "layers.26.mlp.dense_h_to_4h.weight": "model-00008-of-00011.safetensors",
+    "layers.26.post_attention_layernorm.bias": "model-00008-of-00011.safetensors",
+    "layers.26.post_attention_layernorm.weight": "model-00008-of-00011.safetensors",
+    "layers.27.attention.key.bias": "model-00008-of-00011.safetensors",
+    "layers.27.attention.key.weight": "model-00008-of-00011.safetensors",
+    "layers.27.attention.out.bias": "model-00008-of-00011.safetensors",
+    "layers.27.attention.out.weight": "model-00008-of-00011.safetensors",
+    "layers.27.attention.query.bias": "model-00008-of-00011.safetensors",
+    "layers.27.attention.query.weight": "model-00008-of-00011.safetensors",
+    "layers.27.attention.value.bias": "model-00008-of-00011.safetensors",
+    "layers.27.attention.value.weight": "model-00008-of-00011.safetensors",
+    "layers.27.input_layernorm.bias": "model-00008-of-00011.safetensors",
+    "layers.27.input_layernorm.weight": "model-00008-of-00011.safetensors",
+    "layers.27.mlp.dense_4h_to_h.bias": "model-00008-of-00011.safetensors",
+    "layers.27.mlp.dense_4h_to_h.weight": "model-00008-of-00011.safetensors",
+    "layers.27.mlp.dense_h_to_4h.bias": "model-00008-of-00011.safetensors",
+    "layers.27.mlp.dense_h_to_4h.weight": "model-00008-of-00011.safetensors",
+    "layers.27.post_attention_layernorm.bias": "model-00008-of-00011.safetensors",
+    "layers.27.post_attention_layernorm.weight": "model-00008-of-00011.safetensors",
+    "layers.28.attention.key.bias": "model-00008-of-00011.safetensors",
+    "layers.28.attention.key.weight": "model-00008-of-00011.safetensors",
+    "layers.28.attention.out.bias": "model-00008-of-00011.safetensors",
+    "layers.28.attention.out.weight": "model-00008-of-00011.safetensors",
+    "layers.28.attention.query.bias": "model-00008-of-00011.safetensors",
+    "layers.28.attention.query.weight": "model-00008-of-00011.safetensors",
+    "layers.28.attention.value.bias": "model-00008-of-00011.safetensors",
+    "layers.28.attention.value.weight": "model-00008-of-00011.safetensors",
+    "layers.28.input_layernorm.bias": "model-00008-of-00011.safetensors",
+    "layers.28.input_layernorm.weight": "model-00008-of-00011.safetensors",
+    "layers.28.mlp.dense_4h_to_h.bias": "model-00008-of-00011.safetensors",
+    "layers.28.mlp.dense_4h_to_h.weight": "model-00008-of-00011.safetensors",
+    "layers.28.mlp.dense_h_to_4h.bias": "model-00008-of-00011.safetensors",
+    "layers.28.mlp.dense_h_to_4h.weight": "model-00008-of-00011.safetensors",
+    "layers.28.post_attention_layernorm.bias": "model-00008-of-00011.safetensors",
+    "layers.28.post_attention_layernorm.weight": "model-00008-of-00011.safetensors",
+    "layers.29.attention.key.bias": "model-00008-of-00011.safetensors",
+    "layers.29.attention.key.weight": "model-00008-of-00011.safetensors",
+    "layers.29.attention.out.bias": "model-00008-of-00011.safetensors",
+    "layers.29.attention.out.weight": "model-00008-of-00011.safetensors",
+    "layers.29.attention.query.bias": "model-00008-of-00011.safetensors",
+    "layers.29.attention.query.weight": "model-00008-of-00011.safetensors",
+    "layers.29.attention.value.bias": "model-00008-of-00011.safetensors",
+    "layers.29.attention.value.weight": "model-00008-of-00011.safetensors",
+    "layers.29.input_layernorm.bias": "model-00008-of-00011.safetensors",
+    "layers.29.input_layernorm.weight": "model-00008-of-00011.safetensors",
+    "layers.29.mlp.dense_4h_to_h.bias": "model-00008-of-00011.safetensors",
+    "layers.29.mlp.dense_4h_to_h.weight": "model-00008-of-00011.safetensors",
+    "layers.29.mlp.dense_h_to_4h.bias": "model-00008-of-00011.safetensors",
+    "layers.29.mlp.dense_h_to_4h.weight": "model-00008-of-00011.safetensors",
+    "layers.29.post_attention_layernorm.bias": "model-00008-of-00011.safetensors",
+    "layers.29.post_attention_layernorm.weight": "model-00008-of-00011.safetensors",
+    "layers.3.attention.key.bias": "model-00002-of-00011.safetensors",
+    "layers.3.attention.key.weight": "model-00002-of-00011.safetensors",
+    "layers.3.attention.out.bias": "model-00002-of-00011.safetensors",
+    "layers.3.attention.out.weight": "model-00002-of-00011.safetensors",
+    "layers.3.attention.query.bias": "model-00001-of-00011.safetensors",
+    "layers.3.attention.query.weight": "model-00001-of-00011.safetensors",
+    "layers.3.attention.value.bias": "model-00002-of-00011.safetensors",
+    "layers.3.attention.value.weight": "model-00002-of-00011.safetensors",
+    "layers.3.input_layernorm.bias": "model-00002-of-00011.safetensors",
+    "layers.3.input_layernorm.weight": "model-00002-of-00011.safetensors",
+    "layers.3.mlp.dense_4h_to_h.bias": "model-00002-of-00011.safetensors",
+    "layers.3.mlp.dense_4h_to_h.weight": "model-00002-of-00011.safetensors",
+    "layers.3.mlp.dense_h_to_4h.bias": "model-00002-of-00011.safetensors",
+    "layers.3.mlp.dense_h_to_4h.weight": "model-00002-of-00011.safetensors",
+    "layers.3.post_attention_layernorm.bias": "model-00002-of-00011.safetensors",
+    "layers.3.post_attention_layernorm.weight": "model-00002-of-00011.safetensors",
+    "layers.30.attention.key.bias": "model-00009-of-00011.safetensors",
+    "layers.30.attention.key.weight": "model-00009-of-00011.safetensors",
+    "layers.30.attention.out.bias": "model-00009-of-00011.safetensors",
+    "layers.30.attention.out.weight": "model-00009-of-00011.safetensors",
+    "layers.30.attention.query.bias": "model-00009-of-00011.safetensors",
+    "layers.30.attention.query.weight": "model-00009-of-00011.safetensors",
+    "layers.30.attention.value.bias": "model-00009-of-00011.safetensors",
+    "layers.30.attention.value.weight": "model-00009-of-00011.safetensors",
+    "layers.30.input_layernorm.bias": "model-00009-of-00011.safetensors",
+    "layers.30.input_layernorm.weight": "model-00009-of-00011.safetensors",
+    "layers.30.mlp.dense_4h_to_h.bias": "model-00009-of-00011.safetensors",
+    "layers.30.mlp.dense_4h_to_h.weight": "model-00009-of-00011.safetensors",
+    "layers.30.mlp.dense_h_to_4h.bias": "model-00009-of-00011.safetensors",
+    "layers.30.mlp.dense_h_to_4h.weight": "model-00009-of-00011.safetensors",
+    "layers.30.post_attention_layernorm.bias": "model-00009-of-00011.safetensors",
+    "layers.30.post_attention_layernorm.weight": "model-00009-of-00011.safetensors",
+    "layers.31.attention.key.bias": "model-00009-of-00011.safetensors",
+    "layers.31.attention.key.weight": "model-00009-of-00011.safetensors",
+    "layers.31.attention.out.bias": "model-00009-of-00011.safetensors",
+    "layers.31.attention.out.weight": "model-00009-of-00011.safetensors",
+    "layers.31.attention.query.bias": "model-00009-of-00011.safetensors",
+    "layers.31.attention.query.weight": "model-00009-of-00011.safetensors",
+    "layers.31.attention.value.bias": "model-00009-of-00011.safetensors",
+    "layers.31.attention.value.weight": "model-00009-of-00011.safetensors",
+    "layers.31.input_layernorm.bias": "model-00009-of-00011.safetensors",
+    "layers.31.input_layernorm.weight": "model-00009-of-00011.safetensors",
+    "layers.31.mlp.dense_4h_to_h.bias": "model-00009-of-00011.safetensors",
+    "layers.31.mlp.dense_4h_to_h.weight": "model-00009-of-00011.safetensors",
+    "layers.31.mlp.dense_h_to_4h.bias": "model-00009-of-00011.safetensors",
+    "layers.31.mlp.dense_h_to_4h.weight": "model-00009-of-00011.safetensors",
+    "layers.31.post_attention_layernorm.bias": "model-00009-of-00011.safetensors",
+    "layers.31.post_attention_layernorm.weight": "model-00009-of-00011.safetensors",
+    "layers.32.attention.key.bias": "model-00009-of-00011.safetensors",
+    "layers.32.attention.key.weight": "model-00009-of-00011.safetensors",
+    "layers.32.attention.out.bias": "model-00009-of-00011.safetensors",
+    "layers.32.attention.out.weight": "model-00009-of-00011.safetensors",
+    "layers.32.attention.query.bias": "model-00009-of-00011.safetensors",
+    "layers.32.attention.query.weight": "model-00009-of-00011.safetensors",
+    "layers.32.attention.value.bias": "model-00009-of-00011.safetensors",
+    "layers.32.attention.value.weight": "model-00009-of-00011.safetensors",
+    "layers.32.input_layernorm.bias": "model-00009-of-00011.safetensors",
+    "layers.32.input_layernorm.weight": "model-00009-of-00011.safetensors",
+    "layers.32.mlp.dense_4h_to_h.bias": "model-00009-of-00011.safetensors",
+    "layers.32.mlp.dense_4h_to_h.weight": "model-00009-of-00011.safetensors",
+    "layers.32.mlp.dense_h_to_4h.bias": "model-00009-of-00011.safetensors",
+    "layers.32.mlp.dense_h_to_4h.weight": "model-00009-of-00011.safetensors",
+    "layers.32.post_attention_layernorm.bias": "model-00009-of-00011.safetensors",
+    "layers.32.post_attention_layernorm.weight": "model-00009-of-00011.safetensors",
+    "layers.33.attention.key.bias": "model-00009-of-00011.safetensors",
+    "layers.33.attention.key.weight": "model-00009-of-00011.safetensors",
+    "layers.33.attention.out.bias": "model-00009-of-00011.safetensors",
+    "layers.33.attention.out.weight": "model-00009-of-00011.safetensors",
+    "layers.33.attention.query.bias": "model-00009-of-00011.safetensors",
+    "layers.33.attention.query.weight": "model-00009-of-00011.safetensors",
+    "layers.33.attention.value.bias": "model-00009-of-00011.safetensors",
+    "layers.33.attention.value.weight": "model-00009-of-00011.safetensors",
+    "layers.33.input_layernorm.bias": "model-00010-of-00011.safetensors",
+    "layers.33.input_layernorm.weight": "model-00010-of-00011.safetensors",
+    "layers.33.mlp.dense_4h_to_h.bias": "model-00010-of-00011.safetensors",
+    "layers.33.mlp.dense_4h_to_h.weight": "model-00010-of-00011.safetensors",
+    "layers.33.mlp.dense_h_to_4h.bias": "model-00009-of-00011.safetensors",
+    "layers.33.mlp.dense_h_to_4h.weight": "model-00009-of-00011.safetensors",
+    "layers.33.post_attention_layernorm.bias": "model-00010-of-00011.safetensors",
+    "layers.33.post_attention_layernorm.weight": "model-00010-of-00011.safetensors",
+    "layers.34.attention.key.bias": "model-00010-of-00011.safetensors",
+    "layers.34.attention.key.weight": "model-00010-of-00011.safetensors",
+    "layers.34.attention.out.bias": "model-00010-of-00011.safetensors",
+    "layers.34.attention.out.weight": "model-00010-of-00011.safetensors",
+    "layers.34.attention.query.bias": "model-00010-of-00011.safetensors",
+    "layers.34.attention.query.weight": "model-00010-of-00011.safetensors",
+    "layers.34.attention.value.bias": "model-00010-of-00011.safetensors",
+    "layers.34.attention.value.weight": "model-00010-of-00011.safetensors",
+    "layers.34.input_layernorm.bias": "model-00010-of-00011.safetensors",
+    "layers.34.input_layernorm.weight": "model-00010-of-00011.safetensors",
+    "layers.34.mlp.dense_4h_to_h.bias": "model-00010-of-00011.safetensors",
+    "layers.34.mlp.dense_4h_to_h.weight": "model-00010-of-00011.safetensors",
+    "layers.34.mlp.dense_h_to_4h.bias": "model-00010-of-00011.safetensors",
+    "layers.34.mlp.dense_h_to_4h.weight": "model-00010-of-00011.safetensors",
+    "layers.34.post_attention_layernorm.bias": "model-00010-of-00011.safetensors",
+    "layers.34.post_attention_layernorm.weight": "model-00010-of-00011.safetensors",
+    "layers.35.attention.key.bias": "model-00010-of-00011.safetensors",
+    "layers.35.attention.key.weight": "model-00010-of-00011.safetensors",
+    "layers.35.attention.out.bias": "model-00010-of-00011.safetensors",
+    "layers.35.attention.out.weight": "model-00010-of-00011.safetensors",
+    "layers.35.attention.query.bias": "model-00010-of-00011.safetensors",
+    "layers.35.attention.query.weight": "model-00010-of-00011.safetensors",
+    "layers.35.attention.value.bias": "model-00010-of-00011.safetensors",
+    "layers.35.attention.value.weight": "model-00010-of-00011.safetensors",
+    "layers.35.input_layernorm.bias": "model-00010-of-00011.safetensors",
+    "layers.35.input_layernorm.weight": "model-00010-of-00011.safetensors",
+    "layers.35.mlp.dense_4h_to_h.bias": "model-00010-of-00011.safetensors",
+    "layers.35.mlp.dense_4h_to_h.weight": "model-00010-of-00011.safetensors",
+    "layers.35.mlp.dense_h_to_4h.bias": "model-00010-of-00011.safetensors",
+    "layers.35.mlp.dense_h_to_4h.weight": "model-00010-of-00011.safetensors",
+    "layers.35.post_attention_layernorm.bias": "model-00010-of-00011.safetensors",
+    "layers.35.post_attention_layernorm.weight": "model-00010-of-00011.safetensors",
+    "layers.36.attention.key.bias": "model-00010-of-00011.safetensors",
+    "layers.36.attention.key.weight": "model-00010-of-00011.safetensors",
+    "layers.36.attention.out.bias": "model-00010-of-00011.safetensors",
+    "layers.36.attention.out.weight": "model-00010-of-00011.safetensors",
+    "layers.36.attention.query.bias": "model-00010-of-00011.safetensors",
+    "layers.36.attention.query.weight": "model-00010-of-00011.safetensors",
+    "layers.36.attention.value.bias": "model-00010-of-00011.safetensors",
+    "layers.36.attention.value.weight": "model-00010-of-00011.safetensors",
+    "layers.36.input_layernorm.bias": "model-00010-of-00011.safetensors",
+    "layers.36.input_layernorm.weight": "model-00010-of-00011.safetensors",
+    "layers.36.mlp.dense_4h_to_h.bias": "model-00010-of-00011.safetensors",
+    "layers.36.mlp.dense_4h_to_h.weight": "model-00010-of-00011.safetensors",
+    "layers.36.mlp.dense_h_to_4h.bias": "model-00010-of-00011.safetensors",
+    "layers.36.mlp.dense_h_to_4h.weight": "model-00010-of-00011.safetensors",
+    "layers.36.post_attention_layernorm.bias": "model-00010-of-00011.safetensors",
+    "layers.36.post_attention_layernorm.weight": "model-00010-of-00011.safetensors",
+    "layers.37.attention.key.bias": "model-00010-of-00011.safetensors",
+    "layers.37.attention.key.weight": "model-00010-of-00011.safetensors",
+    "layers.37.attention.out.bias": "model-00010-of-00011.safetensors",
+    "layers.37.attention.out.weight": "model-00010-of-00011.safetensors",
+    "layers.37.attention.query.bias": "model-00010-of-00011.safetensors",
+    "layers.37.attention.query.weight": "model-00010-of-00011.safetensors",
+    "layers.37.attention.value.bias": "model-00010-of-00011.safetensors",
+    "layers.37.attention.value.weight": "model-00010-of-00011.safetensors",
+    "layers.37.input_layernorm.bias": "model-00011-of-00011.safetensors",
+    "layers.37.input_layernorm.weight": "model-00011-of-00011.safetensors",
+    "layers.37.mlp.dense_4h_to_h.bias": "model-00011-of-00011.safetensors",
+    "layers.37.mlp.dense_4h_to_h.weight": "model-00011-of-00011.safetensors",
+    "layers.37.mlp.dense_h_to_4h.bias": "model-00011-of-00011.safetensors",
+    "layers.37.mlp.dense_h_to_4h.weight": "model-00011-of-00011.safetensors",
+    "layers.37.post_attention_layernorm.bias": "model-00011-of-00011.safetensors",
+    "layers.37.post_attention_layernorm.weight": "model-00011-of-00011.safetensors",
+    "layers.38.attention.key.bias": "model-00011-of-00011.safetensors",
+    "layers.38.attention.key.weight": "model-00011-of-00011.safetensors",
+    "layers.38.attention.out.bias": "model-00011-of-00011.safetensors",
+    "layers.38.attention.out.weight": "model-00011-of-00011.safetensors",
+    "layers.38.attention.query.bias": "model-00011-of-00011.safetensors",
+    "layers.38.attention.query.weight": "model-00011-of-00011.safetensors",
+    "layers.38.attention.value.bias": "model-00011-of-00011.safetensors",
+    "layers.38.attention.value.weight": "model-00011-of-00011.safetensors",
+    "layers.38.input_layernorm.bias": "model-00011-of-00011.safetensors",
+    "layers.38.input_layernorm.weight": "model-00011-of-00011.safetensors",
+    "layers.38.mlp.dense_4h_to_h.bias": "model-00011-of-00011.safetensors",
+    "layers.38.mlp.dense_4h_to_h.weight": "model-00011-of-00011.safetensors",
+    "layers.38.mlp.dense_h_to_4h.bias": "model-00011-of-00011.safetensors",
+    "layers.38.mlp.dense_h_to_4h.weight": "model-00011-of-00011.safetensors",
+    "layers.38.post_attention_layernorm.bias": "model-00011-of-00011.safetensors",
+    "layers.38.post_attention_layernorm.weight": "model-00011-of-00011.safetensors",
+    "layers.39.attention.key.bias": "model-00011-of-00011.safetensors",
+    "layers.39.attention.key.weight": "model-00011-of-00011.safetensors",
+    "layers.39.attention.out.bias": "model-00011-of-00011.safetensors",
+    "layers.39.attention.out.weight": "model-00011-of-00011.safetensors",
+    "layers.39.attention.query.bias": "model-00011-of-00011.safetensors",
+    "layers.39.attention.query.weight": "model-00011-of-00011.safetensors",
+    "layers.39.attention.value.bias": "model-00011-of-00011.safetensors",
+    "layers.39.attention.value.weight": "model-00011-of-00011.safetensors",
+    "layers.39.input_layernorm.bias": "model-00011-of-00011.safetensors",
+    "layers.39.input_layernorm.weight": "model-00011-of-00011.safetensors",
+    "layers.39.mlp.dense_4h_to_h.bias": "model-00011-of-00011.safetensors",
+    "layers.39.mlp.dense_4h_to_h.weight": "model-00011-of-00011.safetensors",
+    "layers.39.mlp.dense_h_to_4h.bias": "model-00011-of-00011.safetensors",
+    "layers.39.mlp.dense_h_to_4h.weight": "model-00011-of-00011.safetensors",
+    "layers.39.post_attention_layernorm.bias": "model-00011-of-00011.safetensors",
+    "layers.39.post_attention_layernorm.weight": "model-00011-of-00011.safetensors",
+    "layers.4.attention.key.bias": "model-00002-of-00011.safetensors",
+    "layers.4.attention.key.weight": "model-00002-of-00011.safetensors",
+    "layers.4.attention.out.bias": "model-00002-of-00011.safetensors",
+    "layers.4.attention.out.weight": "model-00002-of-00011.safetensors",
+    "layers.4.attention.query.bias": "model-00002-of-00011.safetensors",
+    "layers.4.attention.query.weight": "model-00002-of-00011.safetensors",
+    "layers.4.attention.value.bias": "model-00002-of-00011.safetensors",
+    "layers.4.attention.value.weight": "model-00002-of-00011.safetensors",
+    "layers.4.input_layernorm.bias": "model-00002-of-00011.safetensors",
+    "layers.4.input_layernorm.weight": "model-00002-of-00011.safetensors",
+    "layers.4.mlp.dense_4h_to_h.bias": "model-00002-of-00011.safetensors",
+    "layers.4.mlp.dense_4h_to_h.weight": "model-00002-of-00011.safetensors",
+    "layers.4.mlp.dense_h_to_4h.bias": "model-00002-of-00011.safetensors",
+    "layers.4.mlp.dense_h_to_4h.weight": "model-00002-of-00011.safetensors",
+    "layers.4.post_attention_layernorm.bias": "model-00002-of-00011.safetensors",
+    "layers.4.post_attention_layernorm.weight": "model-00002-of-00011.safetensors",
+    "layers.5.attention.key.bias": "model-00002-of-00011.safetensors",
+    "layers.5.attention.key.weight": "model-00002-of-00011.safetensors",
+    "layers.5.attention.out.bias": "model-00002-of-00011.safetensors",
+    "layers.5.attention.out.weight": "model-00002-of-00011.safetensors",
+    "layers.5.attention.query.bias": "model-00002-of-00011.safetensors",
+    "layers.5.attention.query.weight": "model-00002-of-00011.safetensors",
+    "layers.5.attention.value.bias": "model-00002-of-00011.safetensors",
+    "layers.5.attention.value.weight": "model-00002-of-00011.safetensors",
+    "layers.5.input_layernorm.bias": "model-00002-of-00011.safetensors",
+    "layers.5.input_layernorm.weight": "model-00002-of-00011.safetensors",
+    "layers.5.mlp.dense_4h_to_h.bias": "model-00002-of-00011.safetensors",
+    "layers.5.mlp.dense_4h_to_h.weight": "model-00002-of-00011.safetensors",
+    "layers.5.mlp.dense_h_to_4h.bias": "model-00002-of-00011.safetensors",
+    "layers.5.mlp.dense_h_to_4h.weight": "model-00002-of-00011.safetensors",
+    "layers.5.post_attention_layernorm.bias": "model-00002-of-00011.safetensors",
+    "layers.5.post_attention_layernorm.weight": "model-00002-of-00011.safetensors",
+    "layers.6.attention.key.bias": "model-00002-of-00011.safetensors",
+    "layers.6.attention.key.weight": "model-00002-of-00011.safetensors",
+    "layers.6.attention.out.bias": "model-00002-of-00011.safetensors",
+    "layers.6.attention.out.weight": "model-00002-of-00011.safetensors",
+    "layers.6.attention.query.bias": "model-00002-of-00011.safetensors",
+    "layers.6.attention.query.weight": "model-00002-of-00011.safetensors",
+    "layers.6.attention.value.bias": "model-00002-of-00011.safetensors",
+    "layers.6.attention.value.weight": "model-00002-of-00011.safetensors",
+    "layers.6.input_layernorm.bias": "model-00002-of-00011.safetensors",
+    "layers.6.input_layernorm.weight": "model-00002-of-00011.safetensors",
+    "layers.6.mlp.dense_4h_to_h.bias": "model-00002-of-00011.safetensors",
+    "layers.6.mlp.dense_4h_to_h.weight": "model-00002-of-00011.safetensors",
+    "layers.6.mlp.dense_h_to_4h.bias": "model-00002-of-00011.safetensors",
+    "layers.6.mlp.dense_h_to_4h.weight": "model-00002-of-00011.safetensors",
+    "layers.6.post_attention_layernorm.bias": "model-00002-of-00011.safetensors",
+    "layers.6.post_attention_layernorm.weight": "model-00002-of-00011.safetensors",
+    "layers.7.attention.key.bias": "model-00003-of-00011.safetensors",
+    "layers.7.attention.key.weight": "model-00003-of-00011.safetensors",
+    "layers.7.attention.out.bias": "model-00003-of-00011.safetensors",
+    "layers.7.attention.out.weight": "model-00003-of-00011.safetensors",
+    "layers.7.attention.query.bias": "model-00003-of-00011.safetensors",
+    "layers.7.attention.query.weight": "model-00003-of-00011.safetensors",
+    "layers.7.attention.value.bias": "model-00003-of-00011.safetensors",
+    "layers.7.attention.value.weight": "model-00003-of-00011.safetensors",
+    "layers.7.input_layernorm.bias": "model-00003-of-00011.safetensors",
+    "layers.7.input_layernorm.weight": "model-00003-of-00011.safetensors",
+    "layers.7.mlp.dense_4h_to_h.bias": "model-00003-of-00011.safetensors",
+    "layers.7.mlp.dense_4h_to_h.weight": "model-00003-of-00011.safetensors",
+    "layers.7.mlp.dense_h_to_4h.bias": "model-00003-of-00011.safetensors",
+    "layers.7.mlp.dense_h_to_4h.weight": "model-00003-of-00011.safetensors",
+    "layers.7.post_attention_layernorm.bias": "model-00003-of-00011.safetensors",
+    "layers.7.post_attention_layernorm.weight": "model-00003-of-00011.safetensors",
+    "layers.8.attention.key.bias": "model-00003-of-00011.safetensors",
+    "layers.8.attention.key.weight": "model-00003-of-00011.safetensors",
+    "layers.8.attention.out.bias": "model-00003-of-00011.safetensors",
+    "layers.8.attention.out.weight": "model-00003-of-00011.safetensors",
+    "layers.8.attention.query.bias": "model-00003-of-00011.safetensors",
+    "layers.8.attention.query.weight": "model-00003-of-00011.safetensors",
+    "layers.8.attention.value.bias": "model-00003-of-00011.safetensors",
+    "layers.8.attention.value.weight": "model-00003-of-00011.safetensors",
+    "layers.8.input_layernorm.bias": "model-00003-of-00011.safetensors",
+    "layers.8.input_layernorm.weight": "model-00003-of-00011.safetensors",
+    "layers.8.mlp.dense_4h_to_h.bias": "model-00003-of-00011.safetensors",
+    "layers.8.mlp.dense_4h_to_h.weight": "model-00003-of-00011.safetensors",
+    "layers.8.mlp.dense_h_to_4h.bias": "model-00003-of-00011.safetensors",
+    "layers.8.mlp.dense_h_to_4h.weight": "model-00003-of-00011.safetensors",
+    "layers.8.post_attention_layernorm.bias": "model-00003-of-00011.safetensors",
+    "layers.8.post_attention_layernorm.weight": "model-00003-of-00011.safetensors",
+    "layers.9.attention.key.bias": "model-00003-of-00011.safetensors",
+    "layers.9.attention.key.weight": "model-00003-of-00011.safetensors",
+    "layers.9.attention.out.bias": "model-00003-of-00011.safetensors",
+    "layers.9.attention.out.weight": "model-00003-of-00011.safetensors",
+    "layers.9.attention.query.bias": "model-00003-of-00011.safetensors",
+    "layers.9.attention.query.weight": "model-00003-of-00011.safetensors",
+    "layers.9.attention.value.bias": "model-00003-of-00011.safetensors",
+    "layers.9.attention.value.weight": "model-00003-of-00011.safetensors",
+    "layers.9.input_layernorm.bias": "model-00003-of-00011.safetensors",
+    "layers.9.input_layernorm.weight": "model-00003-of-00011.safetensors",
+    "layers.9.mlp.dense_4h_to_h.bias": "model-00003-of-00011.safetensors",
+    "layers.9.mlp.dense_4h_to_h.weight": "model-00003-of-00011.safetensors",
+    "layers.9.mlp.dense_h_to_4h.bias": "model-00003-of-00011.safetensors",
+    "layers.9.mlp.dense_h_to_4h.weight": "model-00003-of-00011.safetensors",
+    "layers.9.post_attention_layernorm.bias": "model-00003-of-00011.safetensors",
+    "layers.9.post_attention_layernorm.weight": "model-00003-of-00011.safetensors",
+    "lm_head.weight": "model-00011-of-00011.safetensors",
+    "position_embeddings.weight": "model-00001-of-00011.safetensors",
+    "word_embeddings.weight": "model-00001-of-00011.safetensors"
+  }
+}

modeling_riemann.py

@@ -0,0 +1,244 @@
+from __future__ import annotations
+import math
+from typing import Optional
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+import torch.utils.checkpoint
+from transformers import GenerationMixin, PreTrainedModel
+from transformers.modeling_outputs import CausalLMOutputWithPast
+from .configuration_riemann import RiemannConfig
+
+
+class MultiHeadAttention(nn.Module):
+
+    def __init__(self, config: RiemannConfig):
+        super().__init__()
+        self.num_heads = config.num_attention_heads
+        self.hidden_size = config.hidden_size
+        self.head_size = self.hidden_size // self.num_heads
+        self.query = nn.Linear(config.hidden_size, config.hidden_size)
+        self.key = nn.Linear(config.hidden_size, config.hidden_size)
+        self.value = nn.Linear(config.hidden_size, config.hidden_size)
+        self.out = nn.Linear(config.hidden_size, config.hidden_size)
+        self.dropout = nn.Dropout(config.attention_dropout)
+
+    def forward(
+        self, hidden_states: torch.Tensor, attention_mask: Optional[torch.Tensor] = None
+    ):
+        batch_size, seq_length = hidden_states.shape[:2]
+        query_states = self.query(hidden_states)
+        key_states = self.key(hidden_states)
+        value_states = self.value(hidden_states)
+        query_states = query_states.view(
+            batch_size, seq_length, self.num_heads, self.head_size
+        ).transpose(1, 2)
+        key_states = key_states.view(
+            batch_size, seq_length, self.num_heads, self.head_size
+        ).transpose(1, 2)
+        value_states = value_states.view(
+            batch_size, seq_length, self.num_heads, self.head_size
+        ).transpose(1, 2)
+        use_sdpa = hidden_states.device.type in ("cuda", "cpu")
+        if use_sdpa:
+            context_layer = F.scaled_dot_product_attention(
+                query_states,
+                key_states,
+                value_states,
+                attn_mask=attention_mask,
+                dropout_p=self.dropout.p if self.training else 0.0,
+                is_causal=attention_mask is None,
+            )
+        else:
+            attention_scores = torch.matmul(query_states, key_states.transpose(-1, -2))
+            attention_scores = attention_scores / math.sqrt(self.head_size)
+            if attention_mask is None:
+                attention_mask = torch.triu(
+                    torch.full(
+                        (seq_length, seq_length),
+                        torch.finfo(attention_scores.dtype).min,
+                        dtype=attention_scores.dtype,
+                        device=attention_scores.device,
+                    ),
+                    diagonal=1,
+                )
+            attention_scores = attention_scores + attention_mask
+            attention_probs = F.softmax(attention_scores, dim=-1)
+            attention_probs = self.dropout(attention_probs)
+            context_layer = torch.matmul(attention_probs, value_states)
+        context_layer = context_layer.transpose(1, 2).contiguous()
+        context_layer = context_layer.view(batch_size, seq_length, self.hidden_size)
+        context_layer = self.out(context_layer)
+        return context_layer
+
+
+class MLP(nn.Module):
+
+    def __init__(self, config: RiemannConfig):
+        super().__init__()
+        self.dense_h_to_4h = nn.Linear(config.hidden_size, config.intermediate_size)
+        self.dense_4h_to_h = nn.Linear(config.intermediate_size, config.hidden_size)
+        self.act = nn.GELU()
+        self.dropout = nn.Dropout(config.hidden_dropout)
+
+    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
+        hidden_states = self.dense_h_to_4h(hidden_states)
+        hidden_states = self.act(hidden_states)
+        hidden_states = self.dense_4h_to_h(hidden_states)
+        hidden_states = self.dropout(hidden_states)
+        return hidden_states
+
+
+class TransformerBlock(nn.Module):
+
+    def __init__(self, config: RiemannConfig):
+        super().__init__()
+        self.attention = MultiHeadAttention(config)
+        self.mlp = MLP(config)
+        self.input_layernorm = nn.LayerNorm(
+            config.hidden_size, eps=config.layer_norm_eps
+        )
+        self.post_attention_layernorm = nn.LayerNorm(
+            config.hidden_size, eps=config.layer_norm_eps
+        )
+        self.dropout = nn.Dropout(config.hidden_dropout)
+
+    def forward(
+        self, hidden_states: torch.Tensor, attention_mask: Optional[torch.Tensor] = None
+    ):
+        attention_layernorm_out = self.input_layernorm(hidden_states)
+        attention_output = self.attention(
+            attention_layernorm_out, attention_mask=attention_mask
+        )
+        attention_output = self.dropout(attention_output)
+        attention_output = attention_output + hidden_states
+        mlp_layernorm_out = self.post_attention_layernorm(attention_output)
+        mlp_output = self.mlp(mlp_layernorm_out)
+        layer_output = mlp_output + attention_output
+        return layer_output
+
+
+class RiemannPreTrainedModel(PreTrainedModel):
+    config_class = RiemannConfig
+    base_model_prefix = "riemann"
+    supports_gradient_checkpointing = True
+
+    def _init_weights(self, module):
+        if isinstance(module, nn.Linear):
+            module.weight.data.normal_(mean=0.0, std=0.02)
+            if module.bias is not None:
+                module.bias.data.zero_()
+        elif isinstance(module, nn.Embedding):
+            module.weight.data.normal_(mean=0.0, std=0.02)
+        elif isinstance(module, nn.LayerNorm):
+            module.bias.data.zero_()
+            module.weight.data.fill_(1.0)
+
+
+class RiemannForCausalLM(RiemannPreTrainedModel, GenerationMixin):
+
+    def __init__(self, config: RiemannConfig):
+        super().__init__(config)
+        self.config = config
+        self.word_embeddings = nn.Embedding(config.vocab_size, config.hidden_size)
+        self.position_embeddings = nn.Embedding(
+            config.max_position_embeddings, config.hidden_size
+        )
+        self.layers = nn.ModuleList(
+            [TransformerBlock(config) for _ in range(config.num_hidden_layers)]
+        )
+        self.final_layernorm = nn.LayerNorm(
+            config.hidden_size, eps=config.layer_norm_eps
+        )
+        self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, bias=False)
+        self.gradient_checkpointing = False
+        self.post_init()
+
+    def get_input_embeddings(self):
+        return self.word_embeddings
+
+    def set_input_embeddings(self, value):
+        self.word_embeddings = value
+
+    def get_output_embeddings(self):
+        return self.lm_head
+
+    def set_output_embeddings(self, value):
+        self.lm_head = value
+
+    def _build_attention_mask(self, input_shape, attention_mask, dtype, device):
+        batch_size, seq_length = input_shape
+        causal = torch.triu(
+            torch.full(
+                (seq_length, seq_length),
+                torch.finfo(dtype).min,
+                dtype=dtype,
+                device=device,
+            ),
+            diagonal=1,
+        )
+        causal = causal.unsqueeze(0).unsqueeze(0)
+        if attention_mask is not None:
+            padding = attention_mask[:, None, None, :].to(dtype=dtype)
+            padding = (1.0 - padding) * torch.finfo(dtype).min
+            return causal + padding
+        return causal
+
+    def forward(
+        self,
+        input_ids: Optional[torch.Tensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        inputs_embeds: Optional[torch.Tensor] = None,
+        labels: Optional[torch.Tensor] = None,
+        use_cache: Optional[bool] = None,
+        **kwargs,
+    ):
+        if input_ids is not None and inputs_embeds is not None:
+            raise ValueError("pass either input_ids or inputs_embeds, not both")
+        if input_ids is not None:
+            inputs_embeds = self.word_embeddings(input_ids)
+            batch_size, seq_length = input_ids.shape
+        elif inputs_embeds is not None:
+            batch_size, seq_length = inputs_embeds.shape[:2]
+        else:
+            raise ValueError("need either input_ids or inputs_embeds")
+        position_ids = torch.arange(
+            seq_length, dtype=torch.long, device=inputs_embeds.device
+        )
+        position_ids = position_ids.unsqueeze(0).expand(batch_size, -1)
+        position_embeds = self.position_embeddings(position_ids)
+        hidden_states = inputs_embeds + position_embeds
+        full_mask = (
+            None
+            if attention_mask is None
+            else self._build_attention_mask(
+                (batch_size, seq_length),
+                attention_mask,
+                hidden_states.dtype,
+                hidden_states.device,
+            )
+        )
+        for layer in self.layers:
+            if self.gradient_checkpointing and self.training:
+                hidden_states = torch.utils.checkpoint.checkpoint(
+                    layer, hidden_states, full_mask, use_reentrant=False
+                )
+            else:
+                hidden_states = layer(hidden_states, full_mask)
+        hidden_states = self.final_layernorm(hidden_states)
+        logits = self.lm_head(hidden_states)
+        loss = None
+        if labels is not None:
+            shift_logits = logits[..., :-1, :].contiguous()
+            shift_labels = labels[..., 1:].contiguous()
+            loss = F.cross_entropy(
+                shift_logits.view(-1, self.config.vocab_size),
+                shift_labels.view(-1),
+                ignore_index=-100,
+            )
+        return CausalLMOutputWithPast(
+            loss=loss, logits=logits, hidden_states=hidden_states
+        )
+
+    def prepare_inputs_for_generation(self, input_ids, **kwargs):
+        return {"input_ids": input_ids}

special_tokens_map.json

@@ -0,0 +1,6 @@
+{
+  "bos_token": "<|endoftext|>",
+  "eos_token": "<|endoftext|>",
+  "pad_token": "<|endoftext|>",
+  "unk_token": "<|endoftext|>"
+}

tokenizer_config.json

@@ -0,0 +1,21 @@
+{
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "50256": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "bos_token": "<|endoftext|>",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "extra_special_tokens": {},
+  "model_max_length": 1024,
+  "pad_token": "<|endoftext|>",
+  "tokenizer_class": "GPT2Tokenizer",
+  "unk_token": "<|endoftext|>"
+}

training_state.json

@@ -0,0 +1,4 @@
+{
+  "stage": "stage6_12.85b",
+  "global_step": 2064
+}