Rollback to pre-MCQ full fine-tune

config.json

@@ -3,7 +3,8 @@
     "Qwen2ForCausalLM"
   ],
   "attention_dropout": 0.0,
-  "dtype": "float32",
+  "bos_token_id": 151643,
+  "dtype": "bfloat16",
   "eos_token_id": 151645,
   "hidden_act": "silu",
   "hidden_size": 1536,
@@ -45,7 +46,6 @@
   "num_attention_heads": 12,
   "num_hidden_layers": 28,
   "num_key_value_heads": 2,
-  "pad_token_id": 151643,
   "rms_norm_eps": 1e-06,
   "rope_parameters": {
     "rope_theta": 1000000.0,

generation_config.json

@@ -1,4 +1,5 @@
 {
+  "bos_token_id": 151643,
   "do_sample": true,
   "eos_token_id": [
     151645,

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:299d3434b7055f0939d1a23c209d6b3fc889f13486f30c9c9bd6aa463731ef74
-size 3554389592
+oid sha256:9dcbe6f21b10abd806761f576293109e17a3af10db5d2a4c14cf95d0448caab2
+size 3087467144