Push SFT GK model (MMLU 3k + NaturalReasoning 3k, LoRA r=64)

config.json

@@ -4,7 +4,7 @@
   ],
   "attention_bias": false,
   "attention_dropout": 0.0,
-  "bos_token_id": null,
+  "bos_token_id": 151643,
   "dtype": "bfloat16",
   "eos_token_id": 151645,
   "head_dim": 128,
@@ -48,7 +48,7 @@
   "num_attention_heads": 16,
   "num_hidden_layers": 28,
   "num_key_value_heads": 8,
-  "pad_token_id": 151643,
+  "pad_token_id": null,
   "rms_norm_eps": 1e-06,
   "rope_parameters": {
     "rope_theta": 1000000,
@@ -57,7 +57,7 @@
   "sliding_window": null,
   "tie_word_embeddings": true,
   "transformers_version": "5.7.0",
-  "use_cache": false,
+  "use_cache": true,
   "use_sliding_window": false,
   "vocab_size": 151936
 }

generation_config.json

@@ -1,12 +1,13 @@
 {
   "bos_token_id": 151643,
-  "do_sample": false,
+  "do_sample": true,
   "eos_token_id": [
     151645,
     151643
   ],
-  "max_new_tokens": 4096,
   "pad_token_id": 151643,
-  "temperature": 1.0,
+  "temperature": 0.6,
+  "top_k": 20,
+  "top_p": 0.95,
   "transformers_version": "5.7.0"
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a3c236dc027072dd89cc3f6efbe0fad372a064c28fe8b028d6603f9823b39acb
+oid sha256:169ad53ec313c3a34b06c0809216e4fc072cce444a5d4ff2b59690d064130ed5
 size 3441185608