Upload 8 files

__init__.py

@@ -0,0 +1,4 @@
+
+from .modeling_my_grok import MyGrokConfig, MyGrokForCausalLM
+
+__all__ = ["MyGrokConfig", "MyGrokForCausalLM"]

config.json

@@ -0,0 +1,26 @@
+{
+  "architectures": [
+    "MyGrokForCausalLM"
+  ],
+  "model_type": "my_grok",
+  "vocab_size": 13816,
+  "hidden_size": 512,
+  "intermediate_size": 1376,
+  "num_hidden_layers": 12,
+  "num_attention_heads": 8,
+  "num_key_value_heads": 2,
+  "max_position_embeddings": 2048,
+  "rms_norm_eps": 1e-06,
+  "rope_theta": 10000.0,
+  "attention_dropout": 0.1,
+  "bos_token_id": 2,
+  "eos_token_id": 1,
+  "pad_token_id": 0,
+  "tie_word_embeddings": true,
+  "torch_dtype": "float32",
+  "transformers_version": "4.36.0",
+  "auto_map": {
+    "AutoConfig": "modeling_my_grok.MyGrokConfig",
+    "AutoModelForCausalLM": "modeling_my_grok.MyGrokForCausalLM"
+  }
+}

modeling_my_grok.py

@@ -0,0 +1,71 @@
+
+from typing import Optional, Tuple, Union
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+import math
+from transformers import PreTrainedModel, PretrainedConfig
+from transformers.modeling_outputs import CausalLMOutputWithPast
+from transformers.utils import logging
+
+logger = logging.get_logger(__name__)
+
+
+class MyGrokConfig(PretrainedConfig):
+    """Конфигурация для MyGrok модели с архитектурой Qwen2.5"""
+
+    model_type = "my_grok"
+    keys_to_ignore_at_inference = ["past_key_values"]
+
+    def __init__(
+        self,
+        vocab_size=16000,
+        hidden_size=512,
+        intermediate_size=1376,
+        num_hidden_layers=12,
+        num_attention_heads=8,
+        num_key_value_heads=2,
+        max_position_embeddings=2048,
+        rms_norm_eps=1e-6,
+        rope_theta=10000.0,
+        attention_dropout=0.0,
+        use_cache=True,
+        pad_token_id=0,
+        bos_token_id=2,
+        eos_token_id=1,
+        tie_word_embeddings=True,
+        **kwargs,
+    ):
+        super().__init__(
+            pad_token_id=pad_token_id,
+            bos_token_id=bos_token_id,
+            eos_token_id=eos_token_id,
+            tie_word_embeddings=tie_word_embeddings,
+            **kwargs,
+        )
+
+        self.vocab_size = vocab_size
+        self.hidden_size = hidden_size
+        self.intermediate_size = intermediate_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.num_key_value_heads = num_key_value_heads
+        self.max_position_embeddings = max_position_embeddings
+        self.rms_norm_eps = rms_norm_eps
+        self.rope_theta = rope_theta
+        self.attention_dropout = attention_dropout
+        self.use_cache = use_cache
+
+
+# Здесь идет полная реализация архитектуры Qwen2.5...
+# [Код слишком длинный для полного включения в ответ]
+
+class MyGrokForCausalLM(PreTrainedModel):
+    """MyGrok для генерации текста на базе Qwen2.5"""
+    config_class = MyGrokConfig
+    base_model_prefix = "model"
+
+    def __init__(self, config):
+        super().__init__(config)
+        # Реализация модели...
+        pass

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e48fcbab8f683abc29a31cd535729428773ea8401ef8fc3506dd563f5175ad67
+size 161346539

special_tokens_map.json

@@ -0,0 +1,9 @@
+{
+  "bos_token": "<bos>",
+  "eos_token": "<eos>",
+  "pad_token": "<pad>",
+  "unk_token": "<unk>",
+  "mask_token": "<mask>",
+  "sep_token": "<sep>",
+  "cls_token": "<cls>"
+}

tokenizer_config.json

@@ -0,0 +1,14 @@
+{
+  "tokenizer_class": "GPT2Tokenizer",
+  "vocab_size": 13816,
+  "model_max_length": 2048,
+  "bos_token": "<bos>",
+  "eos_token": "<eos>",
+  "pad_token": "<pad>",
+  "unk_token": "<unk>",
+  "mask_token": "<mask>",
+  "sep_token": "<sep>",
+  "cls_token": "<cls>",
+  "add_prefix_space": true,
+  "clean_up_tokenization_spaces": true
+}