First model version_12

config.json

@@ -1,5 +1,4 @@
 {
-  "_name_or_path": "sberbank-ai/rugpt3small_based_on_gpt2",
   "activation_function": "gelu_new",
   "architectures": [
     "GPT2LMHeadModel"
@@ -19,14 +18,10 @@
   "n_layer": 12,
   "n_positions": 2048,
   "resid_pdrop": 0.1,
-  "scale_attn_weights": true,
   "summary_activation": null,
   "summary_first_dropout": 0.1,
   "summary_proj_to_labels": true,
   "summary_type": "cls_index",
   "summary_use_proj": true,
-  "torch_dtype": "float32",
-  "transformers_version": "4.10.3",
-  "use_cache": true,
   "vocab_size": 50264
 }

flax_model.msgpack

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:322182be41c143f2f8ecd8b9e76018d09ae5ca4c752b8858a257811bb4c5f5da
+size 500931352

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:50470554560bd166fde24cd8d3f6ce9a19c521662ef0eb4d427137628fd4d100
-size 551322215
+oid sha256:0d2a4a387b0430654bb1f6813436bd31f8e10a06e18ea0e5c41b33120b78458e
+size 551290714