update(weights): train rand init model for 5k steps on cosmopedia

config.json

@@ -1,30 +1,33 @@
 {
+  "_name_or_path": "/weka2/home-guac/stable-lm/stable-neox/ckpts/stablelm_2_tiny/global_step6000",
   "architectures": [
     "StableLmForCausalLM"
   ],
   "attention_dropout": 0.0,
-  "bos_token_id": 0,
-  "eos_token_id": 0,
+  "bos_token_id": 100257,
+  "eos_token_id": 100257,
   "hidden_act": "silu",
   "hidden_dropout": 0.0,
   "hidden_size": 512,
-  "initializer_range": 0.02,
-  "intermediate_size": 64,
+  "initializer_range": 0.01,
+  "intermediate_size": 1536,
   "layer_norm_eps": 1e-05,
   "max_position_embeddings": 4096,
   "model_type": "stablelm",
   "num_attention_heads": 16,
-  "num_hidden_layers": 2,
-  "num_key_value_heads": 2,
+  "num_hidden_layers": 8,
+  "num_key_value_heads": 4,
   "partial_rotary_factor": 0.25,
   "qk_layernorm": true,
   "rope_scaling": null,
   "rope_theta": 10000,
+  "rotary_scaling_factor": 1.0,
   "tie_word_embeddings": false,
-  "torch_dtype": "float32",
+  "torch_dtype": "bfloat16",
   "transformers_version": "4.39.0.dev0",
   "use_cache": true,
+  "use_norm_bias": false,
   "use_parallel_residual": true,
   "use_qkv_bias": false,
-  "vocab_size": 50304
+  "vocab_size": 100352
 }

generation_config.json

@@ -1,6 +1,6 @@
 {
   "_from_model_config": true,
-  "bos_token_id": 0,
-  "eos_token_id": 0,
+  "bos_token_id": 100257,
+  "eos_token_id": 100257,
   "transformers_version": "4.39.0.dev0"
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:43482736fab1c471b5112f63e63b92a96244153a058bdb710d3617188ff90672
-size 211573792
+oid sha256:5d04033878a589c26fea21e661a7f251d493790ac6bd595ad16a5834e6b17b5f
+size 253811856