all configs and tokenizer files from expts

README.md

@@ -14,6 +14,6 @@ We first fine-tuned gpt-2 for 1 epoch on the English corpus (after cleaning*) of
 
 Since the above dataset contains news regarding Indian subcontinent. We thought that with this fine-tuning, the model will get familiary with India specific terms.
 
-Then we further fine-tuned this model on sentences written by Gandhi.
+Then we further fine-tuned this model on sentences written by Gandhi (for 3 epochs).
 
 *Before cleaning #sents = 54M, after cleaning 42M. We simply took those English sentences which ends with a full-stop.

added_tokens.json

@@ -0,0 +1,4 @@
+{
+  "<|pad|>": 50258,
+  "<|startoftext|>": 50257
+}

config.json

@@ -1,4 +1,5 @@
 {
+  "_name_or_path": "gpt2",
   "activation_function": "gelu_new",
   "architectures": [
     "GPT2LMHeadModel"
@@ -7,15 +8,18 @@
   "bos_token_id": 50256,
   "embd_pdrop": 0.1,
   "eos_token_id": 50256,
+  "gradient_checkpointing": false,
   "initializer_range": 0.02,
   "layer_norm_epsilon": 1e-05,
   "model_type": "gpt2",
   "n_ctx": 1024,
   "n_embd": 768,
   "n_head": 12,
+  "n_inner": null,
   "n_layer": 12,
   "n_positions": 1024,
   "resid_pdrop": 0.1,
+  "scale_attn_weights": true,
   "summary_activation": null,
   "summary_first_dropout": 0.1,
   "summary_proj_to_labels": true,
@@ -27,5 +31,7 @@
       "max_length": 50
     }
   },
-  "vocab_size": 50257
+  "transformers_version": "4.8.2",
+  "use_cache": true,
+  "vocab_size": 50259
 }

special_tokens_map.json

@@ -0,0 +1,24 @@
+{
+  "bos_token": {
+    "content": "<|startoftext|>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": "<|pad|>",
+  "unk_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  }
+}