First model version

.ipynb_checkpoints/README-checkpoint.md

@@ -0,0 +1 @@
+Ushbu model, HuggingFace-da RoBERTa transformatorini amalga oshirishga asoslangan. Bizning RoBERTa dasturimiz 12 ta diqqat boshi va 6 ta qatlamdan foydalanadi, natijada 72 ta aniq e'tibor mexanizmlari paydo bo'ladi. Biz har bir kirish satridagi tokenlarning 15 foizini niqoblaydigan RoBERTa-dan dastlabki tekshirish protsedurasini qabul qildik. Biz maksimal 52K tokenli lug'atdan va maksimal 512 ta ketma-ketlik uzunligidan foydalanganmiz. Biz 1M PubChem to'plamlarida 10 ta davr uchun o'qitdik. Loss funksiya 2.9 dan 0.33 gacha tushdi. Ushbu modelni sizga taqdim qilamiz.

config.json

@@ -0,0 +1,25 @@
+{
+  "architectures": [
+    "RobertaForMaskedLM"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "bos_token_id": 0,
+  "eos_token_id": 2,
+  "gradient_checkpointing": false,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 768,
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "layer_norm_eps": 1e-12,
+  "max_position_embeddings": 512,
+  "model_type": "roberta",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 6,
+  "pad_token_id": 1,
+  "position_embedding_type": "absolute",
+  "transformers_version": "4.7.0",
+  "type_vocab_size": 1,
+  "use_cache": true,
+  "vocab_size": 52000
+}

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3d786c79b6c644a07af010aa530fdcef684df92cff0af755819c9ef0d6941d11
+size 334061362

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e9a70217d4fa0f8d791e8b5b56916f85ace3c3c218676dfeee4ef36e3ac65785
+size 2479