Upload folder using huggingface_hub

GRPO-Qwen2.5-7B/added_tokens.json

@@ -0,0 +1,24 @@
+{
+  "</tool_call>": 151658,
+  "<tool_call>": 151657,
+  "<|box_end|>": 151649,
+  "<|box_start|>": 151648,
+  "<|endoftext|>": 151643,
+  "<|file_sep|>": 151664,
+  "<|fim_middle|>": 151660,
+  "<|fim_pad|>": 151662,
+  "<|fim_prefix|>": 151659,
+  "<|fim_suffix|>": 151661,
+  "<|im_end|>": 151645,
+  "<|im_start|>": 151644,
+  "<|image_pad|>": 151655,
+  "<|object_ref_end|>": 151647,
+  "<|object_ref_start|>": 151646,
+  "<|quad_end|>": 151651,
+  "<|quad_start|>": 151650,
+  "<|repo_name|>": 151663,
+  "<|video_pad|>": 151656,
+  "<|vision_end|>": 151653,
+  "<|vision_pad|>": 151654,
+  "<|vision_start|>": 151652
+}

GRPO-Qwen2.5-7B/config.json

@@ -0,0 +1,28 @@
+{
+  "_name_or_path": "Online-RLHF/sft_checkpoint/Qwen2.5-14B-Instruct",
+  "architectures": [
+    "Qwen2ForCausalLM"
+  ],
+  "attention_dropout": 0.0,
+  "bos_token_id": 151643,
+  "eos_token_id": 151645,
+  "hidden_act": "silu",
+  "hidden_size": 3584,
+  "initializer_range": 0.02,
+  "intermediate_size": 18944,
+  "max_position_embeddings": 32768,
+  "max_window_layers": 28,
+  "model_type": "qwen2",
+  "num_attention_heads": 28,
+  "num_hidden_layers": 28,
+  "num_key_value_heads": 4,
+  "rms_norm_eps": 1e-06,
+  "rope_theta": 1000000.0,
+  "sliding_window": null,
+  "tie_word_embeddings": false,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.43.3",
+  "use_cache": false,
+  "use_sliding_window": false,
+  "vocab_size": 152064
+}

GRPO-Qwen2.5-7B/generation_config.json

@@ -0,0 +1,14 @@
+{
+  "bos_token_id": 151643,
+  "do_sample": true,
+  "eos_token_id": [
+    151645,
+    151643
+  ],
+  "pad_token_id": 151643,
+  "repetition_penalty": 1.05,
+  "temperature": 0.7,
+  "top_k": 20,
+  "top_p": 0.8,
+  "transformers_version": "4.43.3"
+}

GRPO-Qwen2.5-7B/latest

@@ -0,0 +1 @@
+global_step300

GRPO-Qwen2.5-7B/model-00001-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c50c4c0096597f3715d5e4cb697d4a7dd50a1bc4b332860f405c8d36b3c0860c
+size 4877660776

GRPO-Qwen2.5-7B/model-00002-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:abc8789df85c661642bd24f453b2c32533a61d8ad36e00a51801e346a9226f86
+size 4932751008

GRPO-Qwen2.5-7B/model-00003-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a63bca174610be06995e8ac681d90e38a655cfaaa28a99feb57f51204430054f
+size 4330865200

GRPO-Qwen2.5-7B/model-00004-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e0adfc866c34292fb3c9a13532fd6dbf9e640417c8606f2d4953d5f13c0763d7
+size 1089994880

GRPO-Qwen2.5-7B/model.safetensors.index.json

@@ -0,0 +1,346 @@
+{
+  "metadata": {
+    "total_size": 15231233024
+  },
+  "weight_map": {
+    "lm_head.weight": "model-00004-of-00004.safetensors",
+    "model.embed_tokens.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.0.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.0.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.0.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.1.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.1.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.1.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.10.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.10.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.10.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.10.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.11.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.11.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.11.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.12.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.12.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.12.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.13.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.13.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.13.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.14.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.14.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.14.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.15.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.15.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.15.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.16.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.16.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.16.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.17.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.17.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.17.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.18.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.18.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.18.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.18.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.18.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.18.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.18.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.18.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.18.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.19.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.19.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.19.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.19.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.2.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.2.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.2.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.2.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.20.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.20.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.20.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.20.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.20.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.20.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.20.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.20.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.20.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.20.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.20.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.20.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.21.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.21.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.21.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.22.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.22.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.22.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.23.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.23.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.23.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.24.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.24.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.24.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.25.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.25.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.25.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.26.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.26.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.26.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.26.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.26.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.26.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.26.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.26.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.26.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.26.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.26.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.26.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.27.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.27.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.27.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.3.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.3.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.3.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.3.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.4.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.4.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.4.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.5.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.5.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.5.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.6.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.6.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.6.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.7.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.7.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.7.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.8.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.8.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.8.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.8.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.8.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.8.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.8.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.9.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.9.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.9.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.9.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.norm.weight": "model-00003-of-00004.safetensors"
+  }
+}

GRPO-Qwen2.5-7B/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:74c8d337722463e03302f4febf685dad8b5696faf483eba44fc08dad412ada63
+size 16389

GRPO-Qwen2.5-7B/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:53c4237c1614c4c633a43025a0090a990233816338927c9c955846324d5b2a50
+size 16389

GRPO-Qwen2.5-7B/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1ef41f6e29cc1a22a05075b00f43810e0c57a9da1e5acbaeb5dbc75ba17687c8
+size 16389

GRPO-Qwen2.5-7B/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:230b943449a21bb79084e2e41f16597020eec2c876a22f1a50b7f9e90aeb9a03
+size 16389

GRPO-Qwen2.5-7B/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b9fb5842a00fbaaf416fd46e2a48f3cc0ebb987e5f71c6539a053019b2579d2e
+size 16389

GRPO-Qwen2.5-7B/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:beba699c5fc429b85822cd284c107538cf35942d1dfee2eed2811b645857b2c2
+size 16389

GRPO-Qwen2.5-7B/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6c0409921d93b5a657666afceacaf68ace4bec5b421ced13ace851fc935778c6
+size 16389

GRPO-Qwen2.5-7B/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1db8beb7baf2946f5820ec75d95bd275110ea83cd4d2c10904804078dddf05a2
+size 16389

GRPO-Qwen2.5-7B/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:39e3b445bdee7f608178c2d9dddec8bda7d4cb7646dc90d110a0ac2d9619d790
+size 1465

GRPO-Qwen2.5-7B/special_tokens_map.json

@@ -0,0 +1,25 @@
+{
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "eos_token": {
+    "content": "<|im_end|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": "<|im_end|>"
+}

GRPO-Qwen2.5-7B/tokenizer_config.json

@@ -0,0 +1,207 @@
+{
+  "add_bos_token": false,
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "151643": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151644": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151645": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151646": {
+      "content": "<|object_ref_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151647": {
+      "content": "<|object_ref_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151648": {
+      "content": "<|box_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151649": {
+      "content": "<|box_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151650": {
+      "content": "<|quad_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151651": {
+      "content": "<|quad_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151652": {
+      "content": "<|vision_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151653": {
+      "content": "<|vision_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151654": {
+      "content": "<|vision_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151655": {
+      "content": "<|image_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151656": {
+      "content": "<|video_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151657": {
+      "content": "<tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151658": {
+      "content": "</tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151659": {
+      "content": "<|fim_prefix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151660": {
+      "content": "<|fim_middle|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151661": {
+      "content": "<|fim_suffix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151662": {
+      "content": "<|fim_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151663": {
+      "content": "<|repo_name|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151664": {
+      "content": "<|file_sep|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    }
+  },
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "bos_token": null,
+  "chat_template": "{%- if tools %}\n    {{- '<|im_start|>system\\n' }}\n    {%- if messages[0]['role'] == 'system' %}\n        {{- messages[0]['content'] }}\n    {%- else %}\n        {{- 'You are Qwen, created by Alibaba Cloud. You are a helpful assistant.' }}\n    {%- endif %}\n    {{- \"\\n\\n# Tools\\n\\nYou may call one or more functions to assist with the user query.\\n\\nYou are provided with function signatures within <tools></tools> XML tags:\\n<tools>\" }}\n    {%- for tool in tools %}\n        {{- \"\\n\" }}\n        {{- tool | tojson }}\n    {%- endfor %}\n    {{- \"\\n</tools>\\n\\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\\n<tool_call>\\n{\\\"name\\\": <function-name>, \\\"arguments\\\": <args-json-object>}\\n</tool_call><|im_end|>\\n\" }}\n{%- else %}\n    {%- if messages[0]['role'] == 'system' %}\n        {{- '<|im_start|>system\\n' + messages[0]['content'] + '<|im_end|>\\n' }}\n    {%- else %}\n        {{- '<|im_start|>system\\nYou are Qwen, created by Alibaba Cloud. You are a helpful assistant.<|im_end|>\\n' }}\n    {%- endif %}\n{%- endif %}\n{%- for message in messages %}\n    {%- if (message.role == \"user\") or (message.role == \"system\" and not loop.first) or (message.role == \"assistant\" and not message.tool_calls) %}\n        {{- '<|im_start|>' + message.role + '\\n' + message.content + '<|im_end|>' + '\\n' }}\n    {%- elif message.role == \"assistant\" %}\n        {{- '<|im_start|>' + message.role }}\n        {%- if message.content %}\n            {{- '\\n' + message.content }}\n        {%- endif %}\n        {%- for tool_call in message.tool_calls %}\n            {%- if tool_call.function is defined %}\n                {%- set tool_call = tool_call.function %}\n            {%- endif %}\n            {{- '\\n<tool_call>\\n{\"name\": \"' }}\n            {{- tool_call.name }}\n            {{- '\", \"arguments\": ' }}\n            {{- tool_call.arguments | tojson }}\n            {{- '}\\n</tool_call>' }}\n        {%- endfor %}\n        {{- '<|im_end|>\\n' }}\n    {%- elif message.role == \"tool\" %}\n        {%- if (loop.index0 == 0) or (messages[loop.index0 - 1].role != \"tool\") %}\n            {{- '<|im_start|>user' }}\n        {%- endif %}\n        {{- '\\n<tool_response>\\n' }}\n        {{- message.content }}\n        {{- '\\n</tool_response>' }}\n        {%- if loop.last or (messages[loop.index0 + 1].role != \"tool\") %}\n            {{- '<|im_end|>\\n' }}\n        {%- endif %}\n    {%- endif %}\n{%- endfor %}\n{%- if add_generation_prompt %}\n    {{- '<|im_start|>assistant\\n' }}\n{%- endif %}\n",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "model_max_length": 131072,
+  "pad_token": "<|im_end|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}

GRPO-Qwen2.5-7B/trainer_state.json

@@ -0,0 +1,2331 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.9803921568627451,
+  "eval_steps": 100,
+  "global_step": 300,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.006535947712418301,
+      "grad_norm": 45.186977516883296,
+      "learning_rate": 1e-08,
+      "logits/chosen": -0.27009034156799316,
+      "logits/rejected": -0.24063682556152344,
+      "logps/chosen": -158.87863159179688,
+      "logps/rejected": -161.36410522460938,
+      "loss": 0.0081,
+      "rewards/accuracies": 0.0,
+      "rewards/chosen": 0.0,
+      "rewards/margins": 0.0,
+      "rewards/rejected": 0.0,
+      "step": 2
+    },
+    {
+      "epoch": 0.013071895424836602,
+      "grad_norm": 73.24549714056225,
+      "learning_rate": 2e-08,
+      "logits/chosen": -0.5123661160469055,
+      "logits/rejected": -0.6410996317863464,
+      "logps/chosen": -138.18080139160156,
+      "logps/rejected": -117.99702453613281,
+      "loss": 0.0384,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": -0.005687572993338108,
+      "rewards/margins": 0.0037205684930086136,
+      "rewards/rejected": -0.009408143348991871,
+      "step": 4
+    },
+    {
+      "epoch": 0.0196078431372549,
+      "grad_norm": 109.83831851122726,
+      "learning_rate": 3e-08,
+      "logits/chosen": -0.4367651343345642,
+      "logits/rejected": -0.45796388387680054,
+      "logps/chosen": -198.53912353515625,
+      "logps/rejected": -173.95309448242188,
+      "loss": 0.0584,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.0017266026698052883,
+      "rewards/margins": 2.5259330868721008e-05,
+      "rewards/rejected": -0.0017518643289804459,
+      "step": 6
+    },
+    {
+      "epoch": 0.026143790849673203,
+      "grad_norm": 110.0579993785592,
+      "learning_rate": 4e-08,
+      "logits/chosen": -0.2823714017868042,
+      "logits/rejected": -0.23136189579963684,
+      "logps/chosen": -287.70977783203125,
+      "logps/rejected": -280.503662109375,
+      "loss": 0.0512,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.017410974949598312,
+      "rewards/margins": 0.042463820427656174,
+      "rewards/rejected": -0.025052841752767563,
+      "step": 8
+    },
+    {
+      "epoch": 0.032679738562091505,
+      "grad_norm": 95.23381976178658,
+      "learning_rate": 5e-08,
+      "logits/chosen": -0.24388760328292847,
+      "logits/rejected": -0.3079368472099304,
+      "logps/chosen": -163.5430908203125,
+      "logps/rejected": -172.42404174804688,
+      "loss": 0.047,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": -0.012023413553833961,
+      "rewards/margins": 0.001053391955792904,
+      "rewards/rejected": -0.01307680644094944,
+      "step": 10
+    },
+    {
+      "epoch": 0.0392156862745098,
+      "grad_norm": 105.90774713530806,
+      "learning_rate": 6e-08,
+      "logits/chosen": -0.5560111403465271,
+      "logits/rejected": -0.5856660008430481,
+      "logps/chosen": -130.87429809570312,
+      "logps/rejected": -126.42973327636719,
+      "loss": 0.0442,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.019864223897457123,
+      "rewards/margins": 0.04785549268126488,
+      "rewards/rejected": -0.027991265058517456,
+      "step": 12
+    },
+    {
+      "epoch": 0.0457516339869281,
+      "grad_norm": 103.85273262722002,
+      "learning_rate": 7e-08,
+      "logits/chosen": -0.5803347229957581,
+      "logits/rejected": -0.5331610441207886,
+      "logps/chosen": -192.33383178710938,
+      "logps/rejected": -208.4179229736328,
+      "loss": 0.05,
+      "rewards/accuracies": 0.4375,
+      "rewards/chosen": -0.009264146909117699,
+      "rewards/margins": -0.002938077785074711,
+      "rewards/rejected": -0.0063260686583817005,
+      "step": 14
+    },
+    {
+      "epoch": 0.05228758169934641,
+      "grad_norm": 107.35762205675006,
+      "learning_rate": 8e-08,
+      "logits/chosen": -0.6024928092956543,
+      "logits/rejected": -0.6358267068862915,
+      "logps/chosen": -262.2758483886719,
+      "logps/rejected": -248.841796875,
+      "loss": 0.046,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.013199292123317719,
+      "rewards/margins": 0.022662917152047157,
+      "rewards/rejected": -0.03586220741271973,
+      "step": 16
+    },
+    {
+      "epoch": 0.058823529411764705,
+      "grad_norm": 137.49534383655268,
+      "learning_rate": 9e-08,
+      "logits/chosen": 0.047293394804000854,
+      "logits/rejected": 0.08893640339374542,
+      "logps/chosen": -173.2366485595703,
+      "logps/rejected": -179.84747314453125,
+      "loss": 0.0465,
+      "rewards/accuracies": 0.4375,
+      "rewards/chosen": -0.0029950374737381935,
+      "rewards/margins": -0.010749345645308495,
+      "rewards/rejected": 0.007754308171570301,
+      "step": 18
+    },
+    {
+      "epoch": 0.06535947712418301,
+      "grad_norm": 111.9434550888758,
+      "learning_rate": 1e-07,
+      "logits/chosen": -0.5138552784919739,
+      "logits/rejected": -0.459181547164917,
+      "logps/chosen": -158.83338928222656,
+      "logps/rejected": -190.1801300048828,
+      "loss": 0.0397,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.0033895811066031456,
+      "rewards/margins": 0.03482237458229065,
+      "rewards/rejected": -0.03143278881907463,
+      "step": 20
+    },
+    {
+      "epoch": 0.0718954248366013,
+      "grad_norm": 73.32756391400959,
+      "learning_rate": 1.0999999999999999e-07,
+      "logits/chosen": -0.22395852208137512,
+      "logits/rejected": -0.20229049026966095,
+      "logps/chosen": -127.13824462890625,
+      "logps/rejected": -162.4775390625,
+      "loss": 0.0572,
+      "rewards/accuracies": 0.25,
+      "rewards/chosen": -0.0054343463853001595,
+      "rewards/margins": -0.02307252213358879,
+      "rewards/rejected": 0.017638176679611206,
+      "step": 22
+    },
+    {
+      "epoch": 0.0784313725490196,
+      "grad_norm": 66.1892409756167,
+      "learning_rate": 1.2e-07,
+      "logits/chosen": -0.21700644493103027,
+      "logits/rejected": -0.26623234152793884,
+      "logps/chosen": -175.9444122314453,
+      "logps/rejected": -159.27264404296875,
+      "loss": 0.0296,
+      "rewards/accuracies": 0.4375,
+      "rewards/chosen": -0.014468155801296234,
+      "rewards/margins": 0.0006535393185913563,
+      "rewards/rejected": -0.015121696516871452,
+      "step": 24
+    },
+    {
+      "epoch": 0.08496732026143791,
+      "grad_norm": 83.289026575448,
+      "learning_rate": 1.3e-07,
+      "logits/chosen": -0.5614912509918213,
+      "logits/rejected": -0.5100910067558289,
+      "logps/chosen": -301.82861328125,
+      "logps/rejected": -297.5715637207031,
+      "loss": 0.0404,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": 0.0008494257926940918,
+      "rewards/margins": 0.00767078623175621,
+      "rewards/rejected": -0.0068213604390621185,
+      "step": 26
+    },
+    {
+      "epoch": 0.0915032679738562,
+      "grad_norm": 96.3505047955109,
+      "learning_rate": 1.4e-07,
+      "logits/chosen": -0.4129210412502289,
+      "logits/rejected": -0.3650543689727783,
+      "logps/chosen": -128.1119384765625,
+      "logps/rejected": -140.3528289794922,
+      "loss": 0.0521,
+      "rewards/accuracies": 0.4375,
+      "rewards/chosen": -0.023889705538749695,
+      "rewards/margins": -0.02461118996143341,
+      "rewards/rejected": 0.0007214842480607331,
+      "step": 28
+    },
+    {
+      "epoch": 0.09803921568627451,
+      "grad_norm": 73.69189558380467,
+      "learning_rate": 1.5e-07,
+      "logits/chosen": -0.4776341915130615,
+      "logits/rejected": -0.48878008127212524,
+      "logps/chosen": -126.70236206054688,
+      "logps/rejected": -106.85762023925781,
+      "loss": 0.0368,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.006672157905995846,
+      "rewards/margins": -0.004373257979750633,
+      "rewards/rejected": -0.0022989003919065,
+      "step": 30
+    },
+    {
+      "epoch": 0.10457516339869281,
+      "grad_norm": 498.4919423768916,
+      "learning_rate": 1.6e-07,
+      "logits/chosen": -0.3004995286464691,
+      "logits/rejected": -0.2671900689601898,
+      "logps/chosen": -183.04165649414062,
+      "logps/rejected": -222.8990936279297,
+      "loss": 0.0732,
+      "rewards/accuracies": 0.25,
+      "rewards/chosen": -0.03536735847592354,
+      "rewards/margins": -0.059441111981868744,
+      "rewards/rejected": 0.024073755368590355,
+      "step": 32
+    },
+    {
+      "epoch": 0.1111111111111111,
+      "grad_norm": 119.86287254476814,
+      "learning_rate": 1.7000000000000001e-07,
+      "logits/chosen": -0.48381978273391724,
+      "logits/rejected": -0.4859761893749237,
+      "logps/chosen": -179.51870727539062,
+      "logps/rejected": -180.90371704101562,
+      "loss": 0.0514,
+      "rewards/accuracies": 0.4375,
+      "rewards/chosen": 0.005223591811954975,
+      "rewards/margins": -0.0036696307361125946,
+      "rewards/rejected": 0.00889322254806757,
+      "step": 34
+    },
+    {
+      "epoch": 0.11764705882352941,
+      "grad_norm": 62.52050620887828,
+      "learning_rate": 1.8e-07,
+      "logits/chosen": -0.32540175318717957,
+      "logits/rejected": -0.22385042905807495,
+      "logps/chosen": -207.41830444335938,
+      "logps/rejected": -238.51629638671875,
+      "loss": 0.0458,
+      "rewards/accuracies": 0.4375,
+      "rewards/chosen": -0.023636601865291595,
+      "rewards/margins": -0.03656482696533203,
+      "rewards/rejected": 0.012928223237395287,
+      "step": 36
+    },
+    {
+      "epoch": 0.12418300653594772,
+      "grad_norm": 58.467689294858744,
+      "learning_rate": 1.8999999999999998e-07,
+      "logits/chosen": -0.4669422507286072,
+      "logits/rejected": -0.4995095729827881,
+      "logps/chosen": -226.8982391357422,
+      "logps/rejected": -177.6995849609375,
+      "loss": 0.0302,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.01279373373836279,
+      "rewards/margins": 0.02408938854932785,
+      "rewards/rejected": -0.011295657604932785,
+      "step": 38
+    },
+    {
+      "epoch": 0.13071895424836602,
+      "grad_norm": 177.2791858753744,
+      "learning_rate": 2e-07,
+      "logits/chosen": -0.04543227702379227,
+      "logits/rejected": -0.04743846505880356,
+      "logps/chosen": -210.29791259765625,
+      "logps/rejected": -166.44342041015625,
+      "loss": 0.0467,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": -0.012336494401097298,
+      "rewards/margins": 0.0049819243140518665,
+      "rewards/rejected": -0.017318416386842728,
+      "step": 40
+    },
+    {
+      "epoch": 0.13725490196078433,
+      "grad_norm": 91.13096452010151,
+      "learning_rate": 2.0999999999999997e-07,
+      "logits/chosen": -0.27671846747398376,
+      "logits/rejected": -0.25078991055488586,
+      "logps/chosen": -134.34136962890625,
+      "logps/rejected": -120.50598907470703,
+      "loss": 0.0444,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": -0.03240387886762619,
+      "rewards/margins": -0.012337028980255127,
+      "rewards/rejected": -0.02006685361266136,
+      "step": 42
+    },
+    {
+      "epoch": 0.1437908496732026,
+      "grad_norm": 120.79710591958457,
+      "learning_rate": 2.1999999999999998e-07,
+      "logits/chosen": -0.3876258134841919,
+      "logits/rejected": -0.38252392411231995,
+      "logps/chosen": -211.91107177734375,
+      "logps/rejected": -203.24740600585938,
+      "loss": 0.0459,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.01997688040137291,
+      "rewards/margins": -0.007075284607708454,
+      "rewards/rejected": -0.01290159486234188,
+      "step": 44
+    },
+    {
+      "epoch": 0.1503267973856209,
+      "grad_norm": 54.85253819005165,
+      "learning_rate": 2.3e-07,
+      "logits/chosen": -0.4527227282524109,
+      "logits/rejected": -0.38618728518486023,
+      "logps/chosen": -138.82118225097656,
+      "logps/rejected": -189.21810913085938,
+      "loss": 0.0341,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": -0.009205532260239124,
+      "rewards/margins": 0.05001336336135864,
+      "rewards/rejected": -0.05921889841556549,
+      "step": 46
+    },
+    {
+      "epoch": 0.1568627450980392,
+      "grad_norm": 75.93150224574772,
+      "learning_rate": 2.4e-07,
+      "logits/chosen": -0.2979278266429901,
+      "logits/rejected": -0.3578649163246155,
+      "logps/chosen": -234.79408264160156,
+      "logps/rejected": -232.48297119140625,
+      "loss": 0.0433,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.020882954820990562,
+      "rewards/margins": 0.01879079081118107,
+      "rewards/rejected": -0.03967374190688133,
+      "step": 48
+    },
+    {
+      "epoch": 0.16339869281045752,
+      "grad_norm": 100.28327769347527,
+      "learning_rate": 2.5e-07,
+      "logits/chosen": -0.24586978554725647,
+      "logits/rejected": -0.27809399366378784,
+      "logps/chosen": -162.1470947265625,
+      "logps/rejected": -194.7767333984375,
+      "loss": 0.0402,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.012141518294811249,
+      "rewards/margins": 0.006033387966454029,
+      "rewards/rejected": -0.018174903467297554,
+      "step": 50
+    },
+    {
+      "epoch": 0.16993464052287582,
+      "grad_norm": 96.76166919178296,
+      "learning_rate": 2.6e-07,
+      "logits/chosen": -0.6267715096473694,
+      "logits/rejected": -0.6358957886695862,
+      "logps/chosen": -237.6805419921875,
+      "logps/rejected": -227.60003662109375,
+      "loss": 0.0413,
+      "rewards/accuracies": 0.4375,
+      "rewards/chosen": -0.03248181566596031,
+      "rewards/margins": -0.004727806895971298,
+      "rewards/rejected": -0.027754006907343864,
+      "step": 52
+    },
+    {
+      "epoch": 0.17647058823529413,
+      "grad_norm": 79.91297169321828,
+      "learning_rate": 2.7e-07,
+      "logits/chosen": -0.2841065526008606,
+      "logits/rejected": -0.30247509479522705,
+      "logps/chosen": -173.89859008789062,
+      "logps/rejected": -162.05337524414062,
+      "loss": 0.0373,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": -0.01838557980954647,
+      "rewards/margins": 0.014434635639190674,
+      "rewards/rejected": -0.032820217311382294,
+      "step": 54
+    },
+    {
+      "epoch": 0.1830065359477124,
+      "grad_norm": 52.06876207133879,
+      "learning_rate": 2.8e-07,
+      "logits/chosen": -0.3873491883277893,
+      "logits/rejected": -0.3523615002632141,
+      "logps/chosen": -211.60292053222656,
+      "logps/rejected": -212.71141052246094,
+      "loss": 0.0331,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": -0.004088107496500015,
+      "rewards/margins": 0.03653626888990402,
+      "rewards/rejected": -0.04062437266111374,
+      "step": 56
+    },
+    {
+      "epoch": 0.1895424836601307,
+      "grad_norm": 84.75238438262372,
+      "learning_rate": 2.9e-07,
+      "logits/chosen": -0.4390150010585785,
+      "logits/rejected": -0.46629250049591064,
+      "logps/chosen": -91.2630615234375,
+      "logps/rejected": -91.14600372314453,
+      "loss": 0.0438,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.02989867329597473,
+      "rewards/margins": 0.01097220927476883,
+      "rewards/rejected": -0.04087088257074356,
+      "step": 58
+    },
+    {
+      "epoch": 0.19607843137254902,
+      "grad_norm": 75.50050908814517,
+      "learning_rate": 3e-07,
+      "logits/chosen": -0.5333156585693359,
+      "logits/rejected": -0.5472390651702881,
+      "logps/chosen": -124.8789291381836,
+      "logps/rejected": -127.45909118652344,
+      "loss": 0.0346,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.010859816335141659,
+      "rewards/margins": -0.011973461136221886,
+      "rewards/rejected": 0.0011136438697576523,
+      "step": 60
+    },
+    {
+      "epoch": 0.20261437908496732,
+      "grad_norm": 77.32680907408188,
+      "learning_rate": 3.1e-07,
+      "logits/chosen": -0.28590166568756104,
+      "logits/rejected": -0.31295061111450195,
+      "logps/chosen": -205.34341430664062,
+      "logps/rejected": -206.85134887695312,
+      "loss": 0.0375,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": -0.029062893241643906,
+      "rewards/margins": -2.8706155717372894e-05,
+      "rewards/rejected": -0.02903418242931366,
+      "step": 62
+    },
+    {
+      "epoch": 0.20915032679738563,
+      "grad_norm": 72.4176819662502,
+      "learning_rate": 3.2e-07,
+      "logits/chosen": -0.14959962666034698,
+      "logits/rejected": -0.12068141251802444,
+      "logps/chosen": -165.067138671875,
+      "logps/rejected": -160.7367401123047,
+      "loss": 0.0455,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.007513447664678097,
+      "rewards/margins": 0.03002190962433815,
+      "rewards/rejected": -0.03753535449504852,
+      "step": 64
+    },
+    {
+      "epoch": 0.21568627450980393,
+      "grad_norm": 133.92063814206392,
+      "learning_rate": 3.3e-07,
+      "logits/chosen": -0.250588059425354,
+      "logits/rejected": -0.2058851569890976,
+      "logps/chosen": -172.08871459960938,
+      "logps/rejected": -197.60906982421875,
+      "loss": 0.053,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.015155794098973274,
+      "rewards/margins": 0.0010369762312620878,
+      "rewards/rejected": -0.01619276963174343,
+      "step": 66
+    },
+    {
+      "epoch": 0.2222222222222222,
+      "grad_norm": 68.89385593605421,
+      "learning_rate": 3.4000000000000003e-07,
+      "logits/chosen": -0.43986210227012634,
+      "logits/rejected": -0.29949241876602173,
+      "logps/chosen": -99.88211059570312,
+      "logps/rejected": -117.11224365234375,
+      "loss": 0.0295,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.007752654142677784,
+      "rewards/margins": 0.022360891103744507,
+      "rewards/rejected": -0.014608235098421574,
+      "step": 68
+    },
+    {
+      "epoch": 0.22875816993464052,
+      "grad_norm": 108.26670263861944,
+      "learning_rate": 3.5e-07,
+      "logits/chosen": -0.30435511469841003,
+      "logits/rejected": -0.35443350672721863,
+      "logps/chosen": -121.57894897460938,
+      "logps/rejected": -123.25223541259766,
+      "loss": 0.0445,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.014977991580963135,
+      "rewards/margins": 0.03228951618075371,
+      "rewards/rejected": -0.04726750776171684,
+      "step": 70
+    },
+    {
+      "epoch": 0.23529411764705882,
+      "grad_norm": 45.528640494774166,
+      "learning_rate": 3.6e-07,
+      "logits/chosen": -0.31962263584136963,
+      "logits/rejected": -0.378132700920105,
+      "logps/chosen": -246.5164794921875,
+      "logps/rejected": -202.69044494628906,
+      "loss": 0.0367,
+      "rewards/accuracies": 0.4375,
+      "rewards/chosen": -0.0178424920886755,
+      "rewards/margins": 0.027004878968000412,
+      "rewards/rejected": -0.04484737291932106,
+      "step": 72
+    },
+    {
+      "epoch": 0.24183006535947713,
+      "grad_norm": 77.34685226226375,
+      "learning_rate": 3.7e-07,
+      "logits/chosen": -0.3981396555900574,
+      "logits/rejected": -0.2748948335647583,
+      "logps/chosen": -214.7650146484375,
+      "logps/rejected": -203.52685546875,
+      "loss": 0.055,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": -0.015284419059753418,
+      "rewards/margins": 0.0217432864010334,
+      "rewards/rejected": -0.03702770546078682,
+      "step": 74
+    },
+    {
+      "epoch": 0.24836601307189543,
+      "grad_norm": 51.38520168617918,
+      "learning_rate": 3.7999999999999996e-07,
+      "logits/chosen": -0.2653205990791321,
+      "logits/rejected": -0.2614271938800812,
+      "logps/chosen": -158.76300048828125,
+      "logps/rejected": -156.08438110351562,
+      "loss": 0.038,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.00472680851817131,
+      "rewards/margins": 0.02943401411175728,
+      "rewards/rejected": -0.03416082262992859,
+      "step": 76
+    },
+    {
+      "epoch": 0.2549019607843137,
+      "grad_norm": 61.83550419623966,
+      "learning_rate": 3.8999999999999997e-07,
+      "logits/chosen": -0.6505030989646912,
+      "logits/rejected": -0.6271724700927734,
+      "logps/chosen": -285.13299560546875,
+      "logps/rejected": -271.3460388183594,
+      "loss": 0.0351,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 3.300845855847001e-05,
+      "rewards/margins": 0.06149910390377045,
+      "rewards/rejected": -0.061466097831726074,
+      "step": 78
+    },
+    {
+      "epoch": 0.26143790849673204,
+      "grad_norm": 57.32619892247037,
+      "learning_rate": 4e-07,
+      "logits/chosen": -0.2050134837627411,
+      "logits/rejected": -0.3068835139274597,
+      "logps/chosen": -239.4825897216797,
+      "logps/rejected": -205.8413543701172,
+      "loss": 0.0363,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.030740953981876373,
+      "rewards/margins": -0.01743864268064499,
+      "rewards/rejected": -0.013302315026521683,
+      "step": 80
+    },
+    {
+      "epoch": 0.2679738562091503,
+      "grad_norm": 107.39339182579968,
+      "learning_rate": 4.0999999999999994e-07,
+      "logits/chosen": -0.5894187092781067,
+      "logits/rejected": -0.5882883071899414,
+      "logps/chosen": -226.10305786132812,
+      "logps/rejected": -227.6796417236328,
+      "loss": 0.0404,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.029567169025540352,
+      "rewards/margins": 0.023456299677491188,
+      "rewards/rejected": -0.05302346497774124,
+      "step": 82
+    },
+    {
+      "epoch": 0.27450980392156865,
+      "grad_norm": 51.3774223616183,
+      "learning_rate": 4.1999999999999995e-07,
+      "logits/chosen": -0.4666862487792969,
+      "logits/rejected": -0.4283989667892456,
+      "logps/chosen": -123.4700927734375,
+      "logps/rejected": -95.672119140625,
+      "loss": 0.0312,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.03146377205848694,
+      "rewards/margins": 0.015207940712571144,
+      "rewards/rejected": -0.04667171090841293,
+      "step": 84
+    },
+    {
+      "epoch": 0.28104575163398693,
+      "grad_norm": 44.15083575337392,
+      "learning_rate": 4.2999999999999996e-07,
+      "logits/chosen": -0.260882705450058,
+      "logits/rejected": -0.20230424404144287,
+      "logps/chosen": -167.0166015625,
+      "logps/rejected": -162.97259521484375,
+      "loss": 0.0236,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.024938005954027176,
+      "rewards/margins": 0.03646421059966087,
+      "rewards/rejected": -0.06140221655368805,
+      "step": 86
+    },
+    {
+      "epoch": 0.2875816993464052,
+      "grad_norm": 69.35215325120066,
+      "learning_rate": 4.3999999999999997e-07,
+      "logits/chosen": -0.5296764373779297,
+      "logits/rejected": -0.6039686799049377,
+      "logps/chosen": -258.93072509765625,
+      "logps/rejected": -254.47640991210938,
+      "loss": 0.0344,
+      "rewards/accuracies": 0.8125,
+      "rewards/chosen": 0.009829151444137096,
+      "rewards/margins": 0.05222008004784584,
+      "rewards/rejected": -0.04239093139767647,
+      "step": 88
+    },
+    {
+      "epoch": 0.29411764705882354,
+      "grad_norm": 104.16459847425452,
+      "learning_rate": 4.5e-07,
+      "logits/chosen": -0.31153103709220886,
+      "logits/rejected": -0.31315886974334717,
+      "logps/chosen": -178.79116821289062,
+      "logps/rejected": -254.7655487060547,
+      "loss": 0.0409,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": -0.014410759322345257,
+      "rewards/margins": -0.00972677581012249,
+      "rewards/rejected": -0.004683983977884054,
+      "step": 90
+    },
+    {
+      "epoch": 0.3006535947712418,
+      "grad_norm": 156.74384872842165,
+      "learning_rate": 4.6e-07,
+      "logits/chosen": -0.46352726221084595,
+      "logits/rejected": -0.48295092582702637,
+      "logps/chosen": -179.64422607421875,
+      "logps/rejected": -208.92225646972656,
+      "loss": 0.0473,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.006662917323410511,
+      "rewards/margins": 0.03271801769733429,
+      "rewards/rejected": -0.026055097579956055,
+      "step": 92
+    },
+    {
+      "epoch": 0.30718954248366015,
+      "grad_norm": 49.1765646913248,
+      "learning_rate": 4.6999999999999995e-07,
+      "logits/chosen": -0.3818736970424652,
+      "logits/rejected": -0.4321002960205078,
+      "logps/chosen": -120.14766693115234,
+      "logps/rejected": -171.5115966796875,
+      "loss": 0.0306,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.004514832515269518,
+      "rewards/margins": 0.034952811896800995,
+      "rewards/rejected": -0.030437976121902466,
+      "step": 94
+    },
+    {
+      "epoch": 0.3137254901960784,
+      "grad_norm": 79.96110026524003,
+      "learning_rate": 4.8e-07,
+      "logits/chosen": -0.34525632858276367,
+      "logits/rejected": -0.24373121559619904,
+      "logps/chosen": -157.12684631347656,
+      "logps/rejected": -168.47579956054688,
+      "loss": 0.0328,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.02434925176203251,
+      "rewards/margins": 0.029094142839312553,
+      "rewards/rejected": -0.05344339460134506,
+      "step": 96
+    },
+    {
+      "epoch": 0.3202614379084967,
+      "grad_norm": 53.26035737576449,
+      "learning_rate": 4.9e-07,
+      "logits/chosen": -0.2567426264286041,
+      "logits/rejected": -0.24333587288856506,
+      "logps/chosen": -149.03590393066406,
+      "logps/rejected": -152.7711181640625,
+      "loss": 0.034,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.023076260462403297,
+      "rewards/margins": 0.020750535652041435,
+      "rewards/rejected": -0.043826792389154434,
+      "step": 98
+    },
+    {
+      "epoch": 0.32679738562091504,
+      "grad_norm": 87.6736761406898,
+      "learning_rate": 5e-07,
+      "logits/chosen": -0.2764193117618561,
+      "logits/rejected": -0.2525123655796051,
+      "logps/chosen": -217.9014892578125,
+      "logps/rejected": -227.24899291992188,
+      "loss": 0.0424,
+      "rewards/accuracies": 0.8125,
+      "rewards/chosen": -0.007359824143350124,
+      "rewards/margins": 0.06547121703624725,
+      "rewards/rejected": -0.07283104211091995,
+      "step": 100
+    },
+    {
+      "epoch": 0.32679738562091504,
+      "eval_logits/chosen": -0.27211010456085205,
+      "eval_logits/rejected": -0.2623016834259033,
+      "eval_logps/chosen": -225.27517700195312,
+      "eval_logps/rejected": -241.56788635253906,
+      "eval_loss": 0.027157731354236603,
+      "eval_rewards/accuracies": 0.5384615659713745,
+      "eval_rewards/chosen": -0.003800948616117239,
+      "eval_rewards/margins": 0.040041591972112656,
+      "eval_rewards/rejected": -0.04384254664182663,
+      "eval_runtime": 14.2127,
+      "eval_samples_per_second": 7.036,
+      "eval_steps_per_second": 0.915,
+      "step": 100
+    },
+    {
+      "epoch": 0.3333333333333333,
+      "grad_norm": 101.28784465696837,
+      "learning_rate": 4.999811754597861e-07,
+      "logits/chosen": -0.24862715601921082,
+      "logits/rejected": -0.29575347900390625,
+      "logps/chosen": -176.89434814453125,
+      "logps/rejected": -157.25265502929688,
+      "loss": 0.0433,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.043580662459135056,
+      "rewards/margins": 0.045998454093933105,
+      "rewards/rejected": -0.08957912027835846,
+      "step": 102
+    },
+    {
+      "epoch": 0.33986928104575165,
+      "grad_norm": 71.89975535028833,
+      "learning_rate": 4.99924704674051e-07,
+      "logits/chosen": -0.4383275806903839,
+      "logits/rejected": -0.3172180652618408,
+      "logps/chosen": -287.0068664550781,
+      "logps/rejected": -242.01287841796875,
+      "loss": 0.0508,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.008010625839233398,
+      "rewards/margins": 0.04199879616498947,
+      "rewards/rejected": -0.05000942200422287,
+      "step": 104
+    },
+    {
+      "epoch": 0.3464052287581699,
+      "grad_norm": 51.25481149114216,
+      "learning_rate": 4.998305961470874e-07,
+      "logits/chosen": -0.6086812019348145,
+      "logits/rejected": -0.45340508222579956,
+      "logps/chosen": -135.10638427734375,
+      "logps/rejected": -145.10971069335938,
+      "loss": 0.0332,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.027037199586629868,
+      "rewards/margins": 0.03121129423379898,
+      "rewards/rejected": -0.05824849009513855,
+      "step": 106
+    },
+    {
+      "epoch": 0.35294117647058826,
+      "grad_norm": 91.71320259744597,
+      "learning_rate": 4.996988640512931e-07,
+      "logits/chosen": -0.5552752017974854,
+      "logits/rejected": -0.6182447671890259,
+      "logps/chosen": -206.78866577148438,
+      "logps/rejected": -191.0224609375,
+      "loss": 0.0433,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.03465491160750389,
+      "rewards/margins": 0.037451740354299545,
+      "rewards/rejected": -0.07210665196180344,
+      "step": 108
+    },
+    {
+      "epoch": 0.35947712418300654,
+      "grad_norm": 197.12501333305508,
+      "learning_rate": 4.995295282250373e-07,
+      "logits/chosen": -0.28855523467063904,
+      "logits/rejected": -0.27654799818992615,
+      "logps/chosen": -250.15733337402344,
+      "logps/rejected": -297.04241943359375,
+      "loss": 0.0491,
+      "rewards/accuracies": 0.4375,
+      "rewards/chosen": -0.03602990135550499,
+      "rewards/margins": -0.016173377633094788,
+      "rewards/rejected": -0.019856523722410202,
+      "step": 110
+    },
+    {
+      "epoch": 0.3660130718954248,
+      "grad_norm": 80.5529377011555,
+      "learning_rate": 4.993226141696726e-07,
+      "logits/chosen": -0.12453371286392212,
+      "logits/rejected": -0.23867610096931458,
+      "logps/chosen": -139.8778076171875,
+      "logps/rejected": -109.93470001220703,
+      "loss": 0.0349,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.007715956307947636,
+      "rewards/margins": 0.03523428738117218,
+      "rewards/rejected": -0.02751832827925682,
+      "step": 112
+    },
+    {
+      "epoch": 0.37254901960784315,
+      "grad_norm": 2507.9704461396027,
+      "learning_rate": 4.990781530456944e-07,
+      "logits/chosen": -0.37073391675949097,
+      "logits/rejected": -0.2975941002368927,
+      "logps/chosen": -204.9236297607422,
+      "logps/rejected": -177.5823974609375,
+      "loss": 0.0667,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.012357506901025772,
+      "rewards/margins": 0.028699388727545738,
+      "rewards/rejected": -0.041056886315345764,
+      "step": 114
+    },
+    {
+      "epoch": 0.3790849673202614,
+      "grad_norm": 47.31331007031533,
+      "learning_rate": 4.987961816680492e-07,
+      "logits/chosen": -0.3617817163467407,
+      "logits/rejected": -0.34306129813194275,
+      "logps/chosen": -170.80978393554688,
+      "logps/rejected": -159.3184814453125,
+      "loss": 0.0299,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.024177467450499535,
+      "rewards/margins": 0.03145816549658775,
+      "rewards/rejected": -0.05563563480973244,
+      "step": 116
+    },
+    {
+      "epoch": 0.38562091503267976,
+      "grad_norm": 32.94671909903442,
+      "learning_rate": 4.98476742500589e-07,
+      "logits/chosen": -0.34017905592918396,
+      "logits/rejected": -0.3857191205024719,
+      "logps/chosen": -219.93994140625,
+      "logps/rejected": -198.02703857421875,
+      "loss": 0.0328,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": 0.0071497405879199505,
+      "rewards/margins": 0.061277568340301514,
+      "rewards/rejected": -0.054127827286720276,
+      "step": 118
+    },
+    {
+      "epoch": 0.39215686274509803,
+      "grad_norm": 86.3225379014016,
+      "learning_rate": 4.981198836496775e-07,
+      "logits/chosen": -0.216067373752594,
+      "logits/rejected": -0.21636635065078735,
+      "logps/chosen": -179.15823364257812,
+      "logps/rejected": -157.9337158203125,
+      "loss": 0.0381,
+      "rewards/accuracies": 0.375,
+      "rewards/chosen": -0.02674124762415886,
+      "rewards/margins": 0.0077613405883312225,
+      "rewards/rejected": -0.03450259193778038,
+      "step": 120
+    },
+    {
+      "epoch": 0.39869281045751637,
+      "grad_norm": 44.051678427226555,
+      "learning_rate": 4.97725658856945e-07,
+      "logits/chosen": -0.6343599557876587,
+      "logits/rejected": -0.6327584981918335,
+      "logps/chosen": -156.98074340820312,
+      "logps/rejected": -147.32308959960938,
+      "loss": 0.0475,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.015004327520728111,
+      "rewards/margins": 0.05022800713777542,
+      "rewards/rejected": -0.06523232907056808,
+      "step": 122
+    },
+    {
+      "epoch": 0.40522875816993464,
+      "grad_norm": 149.75578987209673,
+      "learning_rate": 4.972941274911952e-07,
+      "logits/chosen": -0.3033314049243927,
+      "logits/rejected": -0.29633399844169617,
+      "logps/chosen": -152.39173889160156,
+      "logps/rejected": -180.32638549804688,
+      "loss": 0.0524,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": 0.0030812136828899384,
+      "rewards/margins": 0.06714190542697906,
+      "rewards/rejected": -0.06406069546937943,
+      "step": 124
+    },
+    {
+      "epoch": 0.4117647058823529,
+      "grad_norm": 82.18299609552417,
+      "learning_rate": 4.968253545394646e-07,
+      "logits/chosen": -0.40410923957824707,
+      "logits/rejected": -0.412802517414093,
+      "logps/chosen": -220.5089111328125,
+      "logps/rejected": -187.06781005859375,
+      "loss": 0.0433,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.011833786964416504,
+      "rewards/margins": 0.0539216473698616,
+      "rewards/rejected": -0.0657554343342781,
+      "step": 126
+    },
+    {
+      "epoch": 0.41830065359477125,
+      "grad_norm": 47.009993331202814,
+      "learning_rate": 4.963194105972352e-07,
+      "logits/chosen": -0.2891400456428528,
+      "logits/rejected": -0.3041932284832001,
+      "logps/chosen": -182.587158203125,
+      "logps/rejected": -183.41351318359375,
+      "loss": 0.0246,
+      "rewards/accuracies": 0.4375,
+      "rewards/chosen": -0.058166660368442535,
+      "rewards/margins": -0.012963611632585526,
+      "rewards/rejected": -0.04520304873585701,
+      "step": 128
+    },
+    {
+      "epoch": 0.42483660130718953,
+      "grad_norm": 40.70703478038728,
+      "learning_rate": 4.957763718578041e-07,
+      "logits/chosen": -0.4367174208164215,
+      "logits/rejected": -0.45004910230636597,
+      "logps/chosen": -164.34942626953125,
+      "logps/rejected": -143.730224609375,
+      "loss": 0.0356,
+      "rewards/accuracies": 0.8125,
+      "rewards/chosen": 0.005354715045541525,
+      "rewards/margins": 0.061466995626688004,
+      "rewards/rejected": -0.05611227825284004,
+      "step": 130
+    },
+    {
+      "epoch": 0.43137254901960786,
+      "grad_norm": 99.16096953498,
+      "learning_rate": 4.951963201008075e-07,
+      "logits/chosen": -0.39199769496917725,
+      "logits/rejected": -0.3033060133457184,
+      "logps/chosen": -214.502685546875,
+      "logps/rejected": -230.5396270751953,
+      "loss": 0.0415,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.018370263278484344,
+      "rewards/margins": 0.0535210557281971,
+      "rewards/rejected": -0.07189132273197174,
+      "step": 132
+    },
+    {
+      "epoch": 0.43790849673202614,
+      "grad_norm": 71.15780255302757,
+      "learning_rate": 4.945793426799069e-07,
+      "logits/chosen": -0.4300433099269867,
+      "logits/rejected": -0.42421796917915344,
+      "logps/chosen": -153.15821838378906,
+      "logps/rejected": -145.508056640625,
+      "loss": 0.0588,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": -0.014510566368699074,
+      "rewards/margins": 0.013640512712299824,
+      "rewards/rejected": -0.028151076287031174,
+      "step": 134
+    },
+    {
+      "epoch": 0.4444444444444444,
+      "grad_norm": 41.069098185508146,
+      "learning_rate": 4.939255325096321e-07,
+      "logits/chosen": -0.19357608258724213,
+      "logits/rejected": -0.14385178685188293,
+      "logps/chosen": -146.409423828125,
+      "logps/rejected": -131.0671844482422,
+      "loss": 0.0255,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": -0.028030507266521454,
+      "rewards/margins": 0.04699843376874924,
+      "rewards/rejected": -0.0750289335846901,
+      "step": 136
+    },
+    {
+      "epoch": 0.45098039215686275,
+      "grad_norm": 197.8558794085048,
+      "learning_rate": 4.9323498805139e-07,
+      "logits/chosen": -0.265545129776001,
+      "logits/rejected": -0.2893003821372986,
+      "logps/chosen": -170.95066833496094,
+      "logps/rejected": -163.1459197998047,
+      "loss": 0.0526,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": -0.001728493720293045,
+      "rewards/margins": 0.059184931218624115,
+      "rewards/rejected": -0.06091342493891716,
+      "step": 138
+    },
+    {
+      "epoch": 0.45751633986928103,
+      "grad_norm": 44.24806873978243,
+      "learning_rate": 4.92507813298636e-07,
+      "logits/chosen": -0.6117287874221802,
+      "logits/rejected": -0.5951799750328064,
+      "logps/chosen": -240.50216674804688,
+      "logps/rejected": -280.4054260253906,
+      "loss": 0.0313,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.02655884250998497,
+      "rewards/margins": 0.038479749113321304,
+      "rewards/rejected": -0.06503859162330627,
+      "step": 140
+    },
+    {
+      "epoch": 0.46405228758169936,
+      "grad_norm": 45.74670203199478,
+      "learning_rate": 4.91744117761213e-07,
+      "logits/chosen": -0.20483921468257904,
+      "logits/rejected": -0.16523896157741547,
+      "logps/chosen": -242.33460998535156,
+      "logps/rejected": -212.6498565673828,
+      "loss": 0.028,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -0.018263787031173706,
+      "rewards/margins": 0.040821466594934464,
+      "rewards/rejected": -0.05908524990081787,
+      "step": 142
+    },
+    {
+      "epoch": 0.47058823529411764,
+      "grad_norm": 72.51220814031906,
+      "learning_rate": 4.9094401644886e-07,
+      "logits/chosen": -0.3257930278778076,
+      "logits/rejected": -0.3842613101005554,
+      "logps/chosen": -285.24542236328125,
+      "logps/rejected": -247.49020385742188,
+      "loss": 0.0358,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.0266116913408041,
+      "rewards/margins": 0.018548624590039253,
+      "rewards/rejected": -0.04516031593084335,
+      "step": 144
+    },
+    {
+      "epoch": 0.477124183006536,
+      "grad_norm": 75.24889159086233,
+      "learning_rate": 4.901076298538914e-07,
+      "logits/chosen": -0.3696897625923157,
+      "logits/rejected": -0.3628237545490265,
+      "logps/chosen": -167.63885498046875,
+      "logps/rejected": -157.33889770507812,
+      "loss": 0.0364,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.00012063398025929928,
+      "rewards/margins": 0.010545263066887856,
+      "rewards/rejected": -0.0106658935546875,
+      "step": 146
+    },
+    {
+      "epoch": 0.48366013071895425,
+      "grad_norm": 52.43980068921001,
+      "learning_rate": 4.892350839330522e-07,
+      "logits/chosen": -0.4322928190231323,
+      "logits/rejected": -0.4630933701992035,
+      "logps/chosen": -260.55731201171875,
+      "logps/rejected": -246.49502563476562,
+      "loss": 0.0419,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": -0.044187724590301514,
+      "rewards/margins": -0.002027117647230625,
+      "rewards/rejected": -0.04216060787439346,
+      "step": 148
+    },
+    {
+      "epoch": 0.49019607843137253,
+      "grad_norm": 73.09555998601864,
+      "learning_rate": 4.883265100885484e-07,
+      "logits/chosen": -0.12294097244739532,
+      "logits/rejected": -0.18963906168937683,
+      "logps/chosen": -212.29010009765625,
+      "logps/rejected": -189.12506103515625,
+      "loss": 0.0443,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.012035032734274864,
+      "rewards/margins": 0.0958094447851181,
+      "rewards/rejected": -0.10784447938203812,
+      "step": 150
+    },
+    {
+      "epoch": 0.49673202614379086,
+      "grad_norm": 76.72854426346223,
+      "learning_rate": 4.873820451482591e-07,
+      "logits/chosen": -0.4885264039039612,
+      "logits/rejected": -0.4644479751586914,
+      "logps/chosen": -185.98782348632812,
+      "logps/rejected": -173.14358520507812,
+      "loss": 0.0354,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": -0.004793064668774605,
+      "rewards/margins": 0.05165287107229233,
+      "rewards/rejected": -0.05644593387842178,
+      "step": 152
+    },
+    {
+      "epoch": 0.5032679738562091,
+      "grad_norm": 54.67449853479546,
+      "learning_rate": 4.864018313451303e-07,
+      "logits/chosen": -0.42130377888679504,
+      "logits/rejected": -0.47208788990974426,
+      "logps/chosen": -170.6588592529297,
+      "logps/rejected": -183.85504150390625,
+      "loss": 0.039,
+      "rewards/accuracies": 0.375,
+      "rewards/chosen": -0.041496820747852325,
+      "rewards/margins": -0.005996996536850929,
+      "rewards/rejected": -0.03549982234835625,
+      "step": 154
+    },
+    {
+      "epoch": 0.5098039215686274,
+      "grad_norm": 65.36834957905036,
+      "learning_rate": 4.853860162957551e-07,
+      "logits/chosen": -0.2886989712715149,
+      "logits/rejected": -0.3145550787448883,
+      "logps/chosen": -191.28952026367188,
+      "logps/rejected": -176.920166015625,
+      "loss": 0.0305,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -0.00622145552188158,
+      "rewards/margins": 0.04355767369270325,
+      "rewards/rejected": -0.04977912828326225,
+      "step": 156
+    },
+    {
+      "epoch": 0.5163398692810458,
+      "grad_norm": 48.67995551910959,
+      "learning_rate": 4.843347529781437e-07,
+      "logits/chosen": -0.30549341440200806,
+      "logits/rejected": -0.23713193833827972,
+      "logps/chosen": -266.6553039550781,
+      "logps/rejected": -215.36293029785156,
+      "loss": 0.0384,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.01247333362698555,
+      "rewards/margins": 0.04777282476425171,
+      "rewards/rejected": -0.06024615466594696,
+      "step": 158
+    },
+    {
+      "epoch": 0.5228758169934641,
+      "grad_norm": 73.00953672926772,
+      "learning_rate": 4.832481997086846e-07,
+      "logits/chosen": -0.2555863559246063,
+      "logits/rejected": -0.2645967900753021,
+      "logps/chosen": -132.61776733398438,
+      "logps/rejected": -107.6927261352539,
+      "loss": 0.0384,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.009637578390538692,
+      "rewards/margins": 0.08541408181190491,
+      "rewards/rejected": -0.09505165368318558,
+      "step": 160
+    },
+    {
+      "epoch": 0.5294117647058824,
+      "grad_norm": 50.14020376104284,
+      "learning_rate": 4.821265201183039e-07,
+      "logits/chosen": 0.08016139268875122,
+      "logits/rejected": 0.06998459994792938,
+      "logps/chosen": -159.1204833984375,
+      "logps/rejected": -145.2503662109375,
+      "loss": 0.0261,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.02143080346286297,
+      "rewards/margins": 0.024232879281044006,
+      "rewards/rejected": -0.045663684606552124,
+      "step": 162
+    },
+    {
+      "epoch": 0.5359477124183006,
+      "grad_norm": 72.2101373334251,
+      "learning_rate": 4.809698831278217e-07,
+      "logits/chosen": -0.33866626024246216,
+      "logits/rejected": -0.29605868458747864,
+      "logps/chosen": -280.1914367675781,
+      "logps/rejected": -306.0768737792969,
+      "loss": 0.0442,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.03203167766332626,
+      "rewards/margins": 0.025068627670407295,
+      "rewards/rejected": -0.05710030719637871,
+      "step": 164
+    },
+    {
+      "epoch": 0.5424836601307189,
+      "grad_norm": 115.71150458809988,
+      "learning_rate": 4.797784629225144e-07,
+      "logits/chosen": -0.40549731254577637,
+      "logits/rejected": -0.4242163300514221,
+      "logps/chosen": -206.74249267578125,
+      "logps/rejected": -208.47763061523438,
+      "loss": 0.0413,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.006779552437365055,
+      "rewards/margins": 0.00495811365544796,
+      "rewards/rejected": -0.011737669818103313,
+      "step": 166
+    },
+    {
+      "epoch": 0.5490196078431373,
+      "grad_norm": 95.9435911527007,
+      "learning_rate": 4.785524389258826e-07,
+      "logits/chosen": -0.43239825963974,
+      "logits/rejected": -0.3713723123073578,
+      "logps/chosen": -235.73826599121094,
+      "logps/rejected": -292.4533386230469,
+      "loss": 0.0399,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.006934458389878273,
+      "rewards/margins": 0.03468269854784012,
+      "rewards/rejected": -0.027748236432671547,
+      "step": 168
+    },
+    {
+      "epoch": 0.5555555555555556,
+      "grad_norm": 89.04577196324648,
+      "learning_rate": 4.772919957726306e-07,
+      "logits/chosen": -0.17360466718673706,
+      "logits/rejected": -0.15008430182933807,
+      "logps/chosen": -215.46627807617188,
+      "logps/rejected": -144.87901306152344,
+      "loss": 0.0389,
+      "rewards/accuracies": 0.3125,
+      "rewards/chosen": -0.06131269037723541,
+      "rewards/margins": -0.011144360527396202,
+      "rewards/rejected": -0.05016832798719406,
+      "step": 170
+    },
+    {
+      "epoch": 0.5620915032679739,
+      "grad_norm": 111.83704879746078,
+      "learning_rate": 4.7599732328086083e-07,
+      "logits/chosen": -0.6134898066520691,
+      "logits/rejected": -0.6926597356796265,
+      "logps/chosen": -223.89962768554688,
+      "logps/rejected": -227.33554077148438,
+      "loss": 0.0354,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.030636107549071312,
+      "rewards/margins": 0.006057367660105228,
+      "rewards/rejected": -0.036693476140499115,
+      "step": 172
+    },
+    {
+      "epoch": 0.5686274509803921,
+      "grad_norm": 64.74576975925206,
+      "learning_rate": 4.746686164234884e-07,
+      "logits/chosen": -0.23288995027542114,
+      "logits/rejected": -0.31232950091362,
+      "logps/chosen": -232.74766540527344,
+      "logps/rejected": -245.40292358398438,
+      "loss": 0.0408,
+      "rewards/accuracies": 0.4375,
+      "rewards/chosen": -0.0636177808046341,
+      "rewards/margins": -0.02461876906454563,
+      "rewards/rejected": -0.03899901360273361,
+      "step": 174
+    },
+    {
+      "epoch": 0.5751633986928104,
+      "grad_norm": 57.24720513707116,
+      "learning_rate": 4.733060752988788e-07,
+      "logits/chosen": -0.5741746425628662,
+      "logits/rejected": -0.5928477048873901,
+      "logps/chosen": -320.5159606933594,
+      "logps/rejected": -310.0211181640625,
+      "loss": 0.0316,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -0.00849897414445877,
+      "rewards/margins": 0.03513266518712044,
+      "rewards/rejected": -0.04363163560628891,
+      "step": 176
+    },
+    {
+      "epoch": 0.5816993464052288,
+      "grad_norm": 50.76505066609826,
+      "learning_rate": 4.719099051007135e-07,
+      "logits/chosen": -0.26551538705825806,
+      "logits/rejected": -0.25012826919555664,
+      "logps/chosen": -185.27513122558594,
+      "logps/rejected": -179.50245666503906,
+      "loss": 0.0387,
+      "rewards/accuracies": 0.4375,
+      "rewards/chosen": -0.012902450747787952,
+      "rewards/margins": 0.0013320930302143097,
+      "rewards/rejected": -0.014234543778002262,
+      "step": 178
+    },
+    {
+      "epoch": 0.5882352941176471,
+      "grad_norm": 86.34465226933693,
+      "learning_rate": 4.704803160870887e-07,
+      "logits/chosen": -0.11030279099941254,
+      "logits/rejected": -0.2218640148639679,
+      "logps/chosen": -178.433349609375,
+      "logps/rejected": -165.95046997070312,
+      "loss": 0.0439,
+      "rewards/accuracies": 0.4375,
+      "rewards/chosen": -0.011190449818968773,
+      "rewards/margins": -0.027342043817043304,
+      "rewards/rejected": 0.01615159586071968,
+      "step": 180
+    },
+    {
+      "epoch": 0.5947712418300654,
+      "grad_norm": 71.18638979427448,
+      "learning_rate": 4.690175235488516e-07,
+      "logits/chosen": -0.39331820607185364,
+      "logits/rejected": -0.3075729012489319,
+      "logps/chosen": -127.24760437011719,
+      "logps/rejected": -143.00271606445312,
+      "loss": 0.0418,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": -0.017544306814670563,
+      "rewards/margins": 0.05406621843576431,
+      "rewards/rejected": -0.07161052525043488,
+      "step": 182
+    },
+    {
+      "epoch": 0.6013071895424836,
+      "grad_norm": 59.553742198448234,
+      "learning_rate": 4.6752174777717785e-07,
+      "logits/chosen": -0.32959118485450745,
+      "logits/rejected": -0.29011330008506775,
+      "logps/chosen": -204.7603759765625,
+      "logps/rejected": -208.90843200683594,
+      "loss": 0.0296,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": 0.03137066587805748,
+      "rewards/margins": 0.07829579710960388,
+      "rewards/rejected": -0.046925127506256104,
+      "step": 184
+    },
+    {
+      "epoch": 0.6078431372549019,
+      "grad_norm": 57.38399229323279,
+      "learning_rate": 4.659932140303967e-07,
+      "logits/chosen": -0.2671470642089844,
+      "logits/rejected": -0.21147973835468292,
+      "logps/chosen": -302.9839172363281,
+      "logps/rejected": -340.75701904296875,
+      "loss": 0.0366,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.02384043112397194,
+      "rewards/margins": -0.00432724691927433,
+      "rewards/rejected": -0.01951317861676216,
+      "step": 186
+    },
+    {
+      "epoch": 0.6143790849673203,
+      "grad_norm": 40.59058661884519,
+      "learning_rate": 4.64432152500068e-07,
+      "logits/chosen": -0.2717140018939972,
+      "logits/rejected": -0.2219124287366867,
+      "logps/chosen": -132.78054809570312,
+      "logps/rejected": -133.28677368164062,
+      "loss": 0.0262,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.021570589393377304,
+      "rewards/margins": 0.01576700061559677,
+      "rewards/rejected": -0.03733758628368378,
+      "step": 188
+    },
+    {
+      "epoch": 0.6209150326797386,
+      "grad_norm": 106.0468033889429,
+      "learning_rate": 4.628387982763163e-07,
+      "logits/chosen": -0.3003775179386139,
+      "logits/rejected": -0.301864892244339,
+      "logps/chosen": -254.27407836914062,
+      "logps/rejected": -261.75933837890625,
+      "loss": 0.032,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.007413983810693026,
+      "rewards/margins": 0.03127045929431915,
+      "rewards/rejected": -0.023856475949287415,
+      "step": 190
+    },
+    {
+      "epoch": 0.6274509803921569,
+      "grad_norm": 52.73567765269575,
+      "learning_rate": 4.612133913124268e-07,
+      "logits/chosen": -0.3008057475090027,
+      "logits/rejected": -0.23563653230667114,
+      "logps/chosen": -193.44033813476562,
+      "logps/rejected": -172.4871826171875,
+      "loss": 0.0455,
+      "rewards/accuracies": 0.375,
+      "rewards/chosen": -0.0540606714785099,
+      "rewards/margins": -0.030924921855330467,
+      "rewards/rejected": -0.023135751485824585,
+      "step": 192
+    },
+    {
+      "epoch": 0.6339869281045751,
+      "grad_norm": 54.33703419101653,
+      "learning_rate": 4.5955617638870947e-07,
+      "logits/chosen": -0.4892820715904236,
+      "logits/rejected": -0.4010857939720154,
+      "logps/chosen": -247.21307373046875,
+      "logps/rejected": -242.25018310546875,
+      "loss": 0.0331,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.024371083825826645,
+      "rewards/margins": 0.0071059828624129295,
+      "rewards/rejected": -0.03147706389427185,
+      "step": 194
+    },
+    {
+      "epoch": 0.6405228758169934,
+      "grad_norm": 50.228886507401455,
+      "learning_rate": 4.578674030756363e-07,
+      "logits/chosen": -0.3148457109928131,
+      "logits/rejected": -0.3138217031955719,
+      "logps/chosen": -268.1454162597656,
+      "logps/rejected": -253.48989868164062,
+      "loss": 0.0336,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.015978574752807617,
+      "rewards/margins": 0.011177252978086472,
+      "rewards/rejected": -0.02715582773089409,
+      "step": 196
+    },
+    {
+      "epoch": 0.6470588235294118,
+      "grad_norm": 66.92458647982313,
+      "learning_rate": 4.561473256962563e-07,
+      "logits/chosen": -0.2708606719970703,
+      "logits/rejected": -0.26236456632614136,
+      "logps/chosen": -147.4287567138672,
+      "logps/rejected": -144.36834716796875,
+      "loss": 0.0308,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.010315479710698128,
+      "rewards/margins": 0.03954504430294037,
+      "rewards/rejected": -0.02922956272959709,
+      "step": 198
+    },
+    {
+      "epoch": 0.6535947712418301,
+      "grad_norm": 36.72249643086387,
+      "learning_rate": 4.543962032878959e-07,
+      "logits/chosen": -0.202681764960289,
+      "logits/rejected": -0.2514805793762207,
+      "logps/chosen": -129.49627685546875,
+      "logps/rejected": -135.28347778320312,
+      "loss": 0.0337,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.01739976741373539,
+      "rewards/margins": 0.07026023417711258,
+      "rewards/rejected": -0.05286046117544174,
+      "step": 200
+    },
+    {
+      "epoch": 0.6535947712418301,
+      "eval_logits/chosen": -0.2794332504272461,
+      "eval_logits/rejected": -0.26968610286712646,
+      "eval_logps/chosen": -225.51043701171875,
+      "eval_logps/rejected": -241.7299041748047,
+      "eval_loss": 0.03847924992442131,
+      "eval_rewards/accuracies": 0.6153846383094788,
+      "eval_rewards/chosen": -0.027326412498950958,
+      "eval_rewards/margins": 0.03271753340959549,
+      "eval_rewards/rejected": -0.06004394590854645,
+      "eval_runtime": 13.0388,
+      "eval_samples_per_second": 7.669,
+      "eval_steps_per_second": 0.997,
+      "step": 200
+    },
+    {
+      "epoch": 0.6601307189542484,
+      "grad_norm": 71.09751023846168,
+      "learning_rate": 4.526142995631487e-07,
+      "logits/chosen": -0.2152756005525589,
+      "logits/rejected": -0.20909318327903748,
+      "logps/chosen": -141.82937622070312,
+      "logps/rejected": -150.34214782714844,
+      "loss": 0.0314,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.016430635005235672,
+      "rewards/margins": 0.04131507873535156,
+      "rewards/rejected": -0.057745713740587234,
+      "step": 202
+    },
+    {
+      "epoch": 0.6666666666666666,
+      "grad_norm": 45.63193666533576,
+      "learning_rate": 4.5080188287016117e-07,
+      "logits/chosen": -0.23087212443351746,
+      "logits/rejected": -0.2568607032299042,
+      "logps/chosen": -224.44241333007812,
+      "logps/rejected": -199.0589599609375,
+      "loss": 0.0312,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": -0.050122812390327454,
+      "rewards/margins": 0.022095417603850365,
+      "rewards/rejected": -0.07221823185682297,
+      "step": 204
+    },
+    {
+      "epoch": 0.673202614379085,
+      "grad_norm": 61.54745715139486,
+      "learning_rate": 4.489592261522209e-07,
+      "logits/chosen": -0.6421210765838623,
+      "logits/rejected": -0.6634953022003174,
+      "logps/chosen": -167.06668090820312,
+      "logps/rejected": -177.42169189453125,
+      "loss": 0.0297,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.025261811912059784,
+      "rewards/margins": 0.053875647485256195,
+      "rewards/rejected": -0.07913745939731598,
+      "step": 206
+    },
+    {
+      "epoch": 0.6797385620915033,
+      "grad_norm": 109.36595449396489,
+      "learning_rate": 4.4708660690665156e-07,
+      "logits/chosen": -0.4967430830001831,
+      "logits/rejected": -0.5115347504615784,
+      "logps/chosen": -163.57867431640625,
+      "logps/rejected": -196.04672241210938,
+      "loss": 0.039,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.012439524754881859,
+      "rewards/margins": 0.03624848276376724,
+      "rewards/rejected": -0.04868800938129425,
+      "step": 208
+    },
+    {
+      "epoch": 0.6862745098039216,
+      "grad_norm": 57.48245653207028,
+      "learning_rate": 4.4518430714302355e-07,
+      "logits/chosen": -0.586164116859436,
+      "logits/rejected": -0.6060546636581421,
+      "logps/chosen": -249.234375,
+      "logps/rejected": -183.01060485839844,
+      "loss": 0.0346,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": -0.05379612743854523,
+      "rewards/margins": -0.004242984112352133,
+      "rewards/rejected": -0.049553144723176956,
+      "step": 210
+    },
+    {
+      "epoch": 0.6928104575163399,
+      "grad_norm": 74.13088548828297,
+      "learning_rate": 4.432526133406842e-07,
+      "logits/chosen": -0.17366769909858704,
+      "logits/rejected": -0.15363453328609467,
+      "logps/chosen": -200.74636840820312,
+      "logps/rejected": -206.4248046875,
+      "loss": 0.0443,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -0.013298360630869865,
+      "rewards/margins": 0.025727882981300354,
+      "rewards/rejected": -0.03902624547481537,
+      "step": 212
+    },
+    {
+      "epoch": 0.6993464052287581,
+      "grad_norm": 49.27085141317152,
+      "learning_rate": 4.412918164056147e-07,
+      "logits/chosen": -0.14803485572338104,
+      "logits/rejected": -0.13686254620552063,
+      "logps/chosen": -222.88531494140625,
+      "logps/rejected": -236.4432373046875,
+      "loss": 0.0371,
+      "rewards/accuracies": 0.4375,
+      "rewards/chosen": -0.035159457474946976,
+      "rewards/margins": 0.0094193359836936,
+      "rewards/rejected": -0.04457879066467285,
+      "step": 214
+    },
+    {
+      "epoch": 0.7058823529411765,
+      "grad_norm": 61.83146677871601,
+      "learning_rate": 4.393022116266211e-07,
+      "logits/chosen": -0.15769526362419128,
+      "logits/rejected": -0.11800147593021393,
+      "logps/chosen": -116.457763671875,
+      "logps/rejected": -113.97442626953125,
+      "loss": 0.0382,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -0.05673939734697342,
+      "rewards/margins": 0.01641450822353363,
+      "rewards/rejected": -0.07315390557050705,
+      "step": 216
+    },
+    {
+      "epoch": 0.7124183006535948,
+      "grad_norm": 79.10512417305885,
+      "learning_rate": 4.372840986308648e-07,
+      "logits/chosen": -0.42518341541290283,
+      "logits/rejected": -0.38883841037750244,
+      "logps/chosen": -209.81771850585938,
+      "logps/rejected": -181.5315704345703,
+      "loss": 0.0468,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.001939011737704277,
+      "rewards/margins": 0.02207384631037712,
+      "rewards/rejected": -0.0240128543227911,
+      "step": 218
+    },
+    {
+      "epoch": 0.7189542483660131,
+      "grad_norm": 51.72259919781669,
+      "learning_rate": 4.352377813387398e-07,
+      "logits/chosen": -0.3397374153137207,
+      "logits/rejected": -0.29077762365341187,
+      "logps/chosen": -164.26441955566406,
+      "logps/rejected": -163.83558654785156,
+      "loss": 0.034,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -0.0027443100698292255,
+      "rewards/margins": 0.049380671232938766,
+      "rewards/rejected": -0.052124984562397,
+      "step": 220
+    },
+    {
+      "epoch": 0.7254901960784313,
+      "grad_norm": 65.21410741841781,
+      "learning_rate": 4.3316356791810316e-07,
+      "logits/chosen": -0.2071482539176941,
+      "logits/rejected": -0.2656726539134979,
+      "logps/chosen": -175.44873046875,
+      "logps/rejected": -145.8885040283203,
+      "loss": 0.0389,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.029756590723991394,
+      "rewards/margins": 0.005180124193429947,
+      "rewards/rejected": -0.03493671864271164,
+      "step": 222
+    },
+    {
+      "epoch": 0.7320261437908496,
+      "grad_norm": 67.0407435952103,
+      "learning_rate": 4.3106177073786677e-07,
+      "logits/chosen": -0.35352203249931335,
+      "logits/rejected": -0.41906818747520447,
+      "logps/chosen": -186.884521484375,
+      "logps/rejected": -132.28582763671875,
+      "loss": 0.0335,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.016441544517874718,
+      "rewards/margins": 0.035818230360746384,
+      "rewards/rejected": -0.019376687705516815,
+      "step": 224
+    },
+    {
+      "epoch": 0.738562091503268,
+      "grad_norm": 66.92329399527597,
+      "learning_rate": 4.2893270632095464e-07,
+      "logits/chosen": -0.8845791816711426,
+      "logits/rejected": -0.8822619915008545,
+      "logps/chosen": -117.72002410888672,
+      "logps/rejected": -124.98562622070312,
+      "loss": 0.0435,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.02073250338435173,
+      "rewards/margins": 0.03629453852772713,
+      "rewards/rejected": -0.057027045637369156,
+      "step": 226
+    },
+    {
+      "epoch": 0.7450980392156863,
+      "grad_norm": 47.62189510754749,
+      "learning_rate": 4.2677669529663686e-07,
+      "logits/chosen": -0.3312215209007263,
+      "logits/rejected": -0.3501621186733246,
+      "logps/chosen": -267.466064453125,
+      "logps/rejected": -224.93753051757812,
+      "loss": 0.034,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.03357229009270668,
+      "rewards/margins": 0.018656831234693527,
+      "rewards/rejected": -0.05222911387681961,
+      "step": 228
+    },
+    {
+      "epoch": 0.7516339869281046,
+      "grad_norm": 47.02834194262754,
+      "learning_rate": 4.245940623522432e-07,
+      "logits/chosen": -0.12694722414016724,
+      "logits/rejected": -0.11925915628671646,
+      "logps/chosen": -177.51560974121094,
+      "logps/rejected": -162.05763244628906,
+      "loss": 0.0258,
+      "rewards/accuracies": 0.8125,
+      "rewards/chosen": -0.012887739576399326,
+      "rewards/margins": 0.0833028256893158,
+      "rewards/rejected": -0.09619056433439255,
+      "step": 230
+    },
+    {
+      "epoch": 0.7581699346405228,
+      "grad_norm": 46.3351330282676,
+      "learning_rate": 4.223851361842667e-07,
+      "logits/chosen": -0.26479241251945496,
+      "logits/rejected": -0.16933494806289673,
+      "logps/chosen": -241.49549865722656,
+      "logps/rejected": -250.08544921875,
+      "loss": 0.0394,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -0.009336483664810658,
+      "rewards/margins": 0.05946476757526398,
+      "rewards/rejected": -0.06880125403404236,
+      "step": 232
+    },
+    {
+      "epoch": 0.7647058823529411,
+      "grad_norm": 70.6602205540441,
+      "learning_rate": 4.201502494488632e-07,
+      "logits/chosen": -0.4951695203781128,
+      "logits/rejected": -0.41871514916419983,
+      "logps/chosen": -217.7791290283203,
+      "logps/rejected": -214.5328369140625,
+      "loss": 0.042,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.027298089116811752,
+      "rewards/margins": 0.030486909672617912,
+      "rewards/rejected": -0.057784996926784515,
+      "step": 234
+    },
+    {
+      "epoch": 0.7712418300653595,
+      "grad_norm": 35.15888058368259,
+      "learning_rate": 4.178897387117546e-07,
+      "logits/chosen": -0.3321375250816345,
+      "logits/rejected": -0.2632332742214203,
+      "logps/chosen": -195.45669555664062,
+      "logps/rejected": -178.31564331054688,
+      "loss": 0.0313,
+      "rewards/accuracies": 0.4375,
+      "rewards/chosen": -0.037370648235082626,
+      "rewards/margins": -0.01588914729654789,
+      "rewards/rejected": -0.021481499075889587,
+      "step": 236
+    },
+    {
+      "epoch": 0.7777777777777778,
+      "grad_norm": 82.8158592087351,
+      "learning_rate": 4.1560394439754295e-07,
+      "logits/chosen": -0.27334389090538025,
+      "logits/rejected": -0.30037882924079895,
+      "logps/chosen": -244.21656799316406,
+      "logps/rejected": -231.90805053710938,
+      "loss": 0.0538,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.03506515547633171,
+      "rewards/margins": 0.033455055207014084,
+      "rewards/rejected": -0.06852021813392639,
+      "step": 238
+    },
+    {
+      "epoch": 0.7843137254901961,
+      "grad_norm": 51.86406034696497,
+      "learning_rate": 4.1329321073844413e-07,
+      "logits/chosen": -0.2863461375236511,
+      "logits/rejected": -0.3510417938232422,
+      "logps/chosen": -175.3660888671875,
+      "logps/rejected": -147.9562225341797,
+      "loss": 0.0263,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.030355989933013916,
+      "rewards/margins": 0.06688905507326126,
+      "rewards/rejected": -0.09724503755569458,
+      "step": 240
+    },
+    {
+      "epoch": 0.7908496732026143,
+      "grad_norm": 40.129828469241076,
+      "learning_rate": 4.109578857224478e-07,
+      "logits/chosen": -0.4185572564601898,
+      "logits/rejected": -0.43934208154678345,
+      "logps/chosen": -114.40471649169922,
+      "logps/rejected": -109.1476821899414,
+      "loss": 0.0362,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.03967870771884918,
+      "rewards/margins": 0.06878667324781418,
+      "rewards/rejected": -0.029107961803674698,
+      "step": 242
+    },
+    {
+      "epoch": 0.7973856209150327,
+      "grad_norm": 48.74623786533979,
+      "learning_rate": 4.0859832104091136e-07,
+      "logits/chosen": -0.1064753532409668,
+      "logits/rejected": -0.07447938621044159,
+      "logps/chosen": -128.144775390625,
+      "logps/rejected": -137.69802856445312,
+      "loss": 0.0347,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": -0.0157692339271307,
+      "rewards/margins": 0.030010271817445755,
+      "rewards/rejected": -0.0457795076072216,
+      "step": 244
+    },
+    {
+      "epoch": 0.803921568627451,
+      "grad_norm": 52.84978728808052,
+      "learning_rate": 4.062148720355966e-07,
+      "logits/chosen": -0.6164553165435791,
+      "logits/rejected": -0.6747774481773376,
+      "logps/chosen": -181.6231689453125,
+      "logps/rejected": -142.87408447265625,
+      "loss": 0.0321,
+      "rewards/accuracies": 0.8125,
+      "rewards/chosen": -0.010438485071063042,
+      "rewards/margins": 0.0474529042840004,
+      "rewards/rejected": -0.05789139121770859,
+      "step": 246
+    },
+    {
+      "epoch": 0.8104575163398693,
+      "grad_norm": 119.61904406288232,
+      "learning_rate": 4.038078976451567e-07,
+      "logits/chosen": -0.2829161584377289,
+      "logits/rejected": -0.30661067366600037,
+      "logps/chosen": -162.88607788085938,
+      "logps/rejected": -173.26712036132812,
+      "loss": 0.051,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.04022559896111488,
+      "rewards/margins": 0.026598503813147545,
+      "rewards/rejected": 0.013627094216644764,
+      "step": 248
+    },
+    {
+      "epoch": 0.8169934640522876,
+      "grad_norm": 101.19196335734438,
+      "learning_rate": 4.013777603510814e-07,
+      "logits/chosen": -0.29677027463912964,
+      "logits/rejected": -0.2817401885986328,
+      "logps/chosen": -257.9617919921875,
+      "logps/rejected": -204.421875,
+      "loss": 0.0355,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.026150740683078766,
+      "rewards/margins": 0.0313868410885334,
+      "rewards/rejected": -0.005236100405454636,
+      "step": 250
+    },
+    {
+      "epoch": 0.8235294117647058,
+      "grad_norm": 114.0349972637749,
+      "learning_rate": 3.9892482612310836e-07,
+      "logits/chosen": -0.19675743579864502,
+      "logits/rejected": -0.23067927360534668,
+      "logps/chosen": -236.50201416015625,
+      "logps/rejected": -244.11070251464844,
+      "loss": 0.0407,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": 4.799594171345234e-05,
+      "rewards/margins": 0.045811135321855545,
+      "rewards/rejected": -0.04576313495635986,
+      "step": 252
+    },
+    {
+      "epoch": 0.8300653594771242,
+      "grad_norm": 61.84648031355568,
+      "learning_rate": 3.964494643641097e-07,
+      "logits/chosen": -0.5482683777809143,
+      "logits/rejected": -0.45914798974990845,
+      "logps/chosen": -146.50405883789062,
+      "logps/rejected": -189.81564331054688,
+      "loss": 0.0333,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.002045202534645796,
+      "rewards/margins": 0.019666600972414017,
+      "rewards/rejected": -0.02171180211007595,
+      "step": 254
+    },
+    {
+      "epoch": 0.8366013071895425,
+      "grad_norm": 65.83749315052893,
+      "learning_rate": 3.9395204785446134e-07,
+      "logits/chosen": -0.6100823879241943,
+      "logits/rejected": -0.6729919910430908,
+      "logps/chosen": -249.09225463867188,
+      "logps/rejected": -252.1439666748047,
+      "loss": 0.0478,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": -0.03489159792661667,
+      "rewards/margins": 0.02592843770980835,
+      "rewards/rejected": -0.06082003936171532,
+      "step": 256
+    },
+    {
+      "epoch": 0.8431372549019608,
+      "grad_norm": 42.0032142833148,
+      "learning_rate": 3.914329526959033e-07,
+      "logits/chosen": -0.32031214237213135,
+      "logits/rejected": -0.35006630420684814,
+      "logps/chosen": -179.93194580078125,
+      "logps/rejected": -158.91160583496094,
+      "loss": 0.0357,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": -0.03062274679541588,
+      "rewards/margins": 0.021233588457107544,
+      "rewards/rejected": -0.05185633897781372,
+      "step": 258
+    },
+    {
+      "epoch": 0.8496732026143791,
+      "grad_norm": 55.52271334396093,
+      "learning_rate": 3.8889255825490053e-07,
+      "logits/chosen": -0.5715893507003784,
+      "logits/rejected": -0.633870542049408,
+      "logps/chosen": -109.69697570800781,
+      "logps/rejected": -117.57046508789062,
+      "loss": 0.0365,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -0.005186846479773521,
+      "rewards/margins": 0.06275759637355804,
+      "rewards/rejected": -0.06794443726539612,
+      "step": 260
+    },
+    {
+      "epoch": 0.8562091503267973,
+      "grad_norm": 54.614118908985105,
+      "learning_rate": 3.863312471055116e-07,
+      "logits/chosen": -0.4618644118309021,
+      "logits/rejected": -0.45565295219421387,
+      "logps/chosen": -285.0732421875,
+      "logps/rejected": -259.30438232421875,
+      "loss": 0.0339,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": -0.01997671276330948,
+      "rewards/margins": 0.05328190326690674,
+      "rewards/rejected": -0.07325860857963562,
+      "step": 262
+    },
+    {
+      "epoch": 0.8627450980392157,
+      "grad_norm": 86.59748573844875,
+      "learning_rate": 3.837494049717743e-07,
+      "logits/chosen": -0.34464627504348755,
+      "logits/rejected": -0.39752787351608276,
+      "logps/chosen": -128.7744140625,
+      "logps/rejected": -106.2349853515625,
+      "loss": 0.0541,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.0070334551855921745,
+      "rewards/margins": 0.04889511689543724,
+      "rewards/rejected": -0.05592857301235199,
+      "step": 264
+    },
+    {
+      "epoch": 0.869281045751634,
+      "grad_norm": 48.39173924230384,
+      "learning_rate": 3.811474206696172e-07,
+      "logits/chosen": -0.3531397581100464,
+      "logits/rejected": -0.38672247529029846,
+      "logps/chosen": -186.7664794921875,
+      "logps/rejected": -231.62869262695312,
+      "loss": 0.0332,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": -0.012070789001882076,
+      "rewards/margins": 0.034914784133434296,
+      "rewards/rejected": -0.04698557034134865,
+      "step": 266
+    },
+    {
+      "epoch": 0.8758169934640523,
+      "grad_norm": 46.90161013550772,
+      "learning_rate": 3.7852568604830535e-07,
+      "logits/chosen": -0.6219074726104736,
+      "logits/rejected": -0.6186295747756958,
+      "logps/chosen": -251.63230895996094,
+      "logps/rejected": -268.7247314453125,
+      "loss": 0.0312,
+      "rewards/accuracies": 0.375,
+      "rewards/chosen": -0.02365129068493843,
+      "rewards/margins": -0.0044717611744999886,
+      "rewards/rejected": -0.019179528579115868,
+      "step": 268
+    },
+    {
+      "epoch": 0.8823529411764706,
+      "grad_norm": 33.053818795227286,
+      "learning_rate": 3.758845959314294e-07,
+      "logits/chosen": -0.5435028076171875,
+      "logits/rejected": -0.5069835782051086,
+      "logps/chosen": -139.86465454101562,
+      "logps/rejected": -186.67251586914062,
+      "loss": 0.0391,
+      "rewards/accuracies": 0.8125,
+      "rewards/chosen": -0.006164162419736385,
+      "rewards/margins": 0.06311607360839844,
+      "rewards/rejected": -0.0692802369594574,
+      "step": 270
+    },
+    {
+      "epoch": 0.8888888888888888,
+      "grad_norm": 105.81914894276271,
+      "learning_rate": 3.73224548057446e-07,
+      "logits/chosen": -0.6307239532470703,
+      "logits/rejected": -0.6075409054756165,
+      "logps/chosen": -145.82635498046875,
+      "logps/rejected": -121.76011657714844,
+      "loss": 0.0404,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.000289565185084939,
+      "rewards/margins": 0.03613666445016861,
+      "rewards/rejected": -0.03584709018468857,
+      "step": 272
+    },
+    {
+      "epoch": 0.8954248366013072,
+      "grad_norm": 89.95781222613417,
+      "learning_rate": 3.705459430197807e-07,
+      "logits/chosen": -0.2674006223678589,
+      "logits/rejected": -0.16192153096199036,
+      "logps/chosen": -202.25843811035156,
+      "logps/rejected": -185.20831298828125,
+      "loss": 0.0418,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.048374757170677185,
+      "rewards/margins": 0.027705367654561996,
+      "rewards/rejected": -0.07608012109994888,
+      "step": 274
+    },
+    {
+      "epoch": 0.9019607843137255,
+      "grad_norm": 113.280337336444,
+      "learning_rate": 3.6784918420649944e-07,
+      "logits/chosen": -0.3857190012931824,
+      "logits/rejected": -0.2640799283981323,
+      "logps/chosen": -116.33189392089844,
+      "logps/rejected": -149.42462158203125,
+      "loss": 0.0475,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.004665530286729336,
+      "rewards/margins": 0.04346862807869911,
+      "rewards/rejected": -0.0388030931353569,
+      "step": 276
+    },
+    {
+      "epoch": 0.9084967320261438,
+      "grad_norm": 111.47562456293704,
+      "learning_rate": 3.6513467773956e-07,
+      "logits/chosen": -0.2151796817779541,
+      "logits/rejected": -0.12956911325454712,
+      "logps/chosen": -216.22149658203125,
+      "logps/rejected": -201.15672302246094,
+      "loss": 0.0419,
+      "rewards/accuracies": 0.8125,
+      "rewards/chosen": -0.009489751420915127,
+      "rewards/margins": 0.06132902204990387,
+      "rewards/rejected": -0.07081876695156097,
+      "step": 278
+    },
+    {
+      "epoch": 0.9150326797385621,
+      "grad_norm": 54.460011174623446,
+      "learning_rate": 3.624028324136517e-07,
+      "logits/chosen": -0.12757417559623718,
+      "logits/rejected": -0.11652399599552155,
+      "logps/chosen": -154.72991943359375,
+      "logps/rejected": -124.26289367675781,
+      "loss": 0.0466,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": -0.030394166707992554,
+      "rewards/margins": -0.00027284957468509674,
+      "rewards/rejected": -0.030121318995952606,
+      "step": 280
+    },
+    {
+      "epoch": 0.9215686274509803,
+      "grad_norm": 103.81076751840281,
+      "learning_rate": 3.596540596346319e-07,
+      "logits/chosen": -0.6131604909896851,
+      "logits/rejected": -0.6404291391372681,
+      "logps/chosen": -166.4450225830078,
+      "logps/rejected": -165.696533203125,
+      "loss": 0.04,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.040318600833415985,
+      "rewards/margins": 0.025455022230744362,
+      "rewards/rejected": -0.0657736212015152,
+      "step": 282
+    },
+    {
+      "epoch": 0.9281045751633987,
+      "grad_norm": 79.44757332375497,
+      "learning_rate": 3.568887733575705e-07,
+      "logits/chosen": -0.20878827571868896,
+      "logits/rejected": -0.07686343789100647,
+      "logps/chosen": -183.89395141601562,
+      "logps/rejected": -166.82208251953125,
+      "loss": 0.0395,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.0037750527262687683,
+      "rewards/margins": 0.07864300906658173,
+      "rewards/rejected": -0.07486795634031296,
+      "step": 284
+    },
+    {
+      "epoch": 0.934640522875817,
+      "grad_norm": 137.97164733930262,
+      "learning_rate": 3.541073900244093e-07,
+      "logits/chosen": -0.4165429472923279,
+      "logits/rejected": -0.5227095484733582,
+      "logps/chosen": -143.21482849121094,
+      "logps/rejected": -146.760498046875,
+      "loss": 0.0385,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.019058387726545334,
+      "rewards/margins": 0.05776282027363777,
+      "rewards/rejected": -0.0768212080001831,
+      "step": 286
+    },
+    {
+      "epoch": 0.9411764705882353,
+      "grad_norm": 53.964568437000956,
+      "learning_rate": 3.5131032850124743e-07,
+      "logits/chosen": -0.24787794053554535,
+      "logits/rejected": -0.12958963215351105,
+      "logps/chosen": -139.16610717773438,
+      "logps/rejected": -177.59097290039062,
+      "loss": 0.0364,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.006884424947202206,
+      "rewards/margins": 0.03813309222459793,
+      "rewards/rejected": -0.04501751810312271,
+      "step": 288
+    },
+    {
+      "epoch": 0.9477124183006536,
+      "grad_norm": 44.52522827038271,
+      "learning_rate": 3.48498010015262e-07,
+      "logits/chosen": -0.1434451788663864,
+      "logits/rejected": -0.17539609968662262,
+      "logps/chosen": -128.35287475585938,
+      "logps/rejected": -139.58091735839844,
+      "loss": 0.041,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": -0.013128826394677162,
+      "rewards/margins": 0.02540736459195614,
+      "rewards/rejected": -0.0385361909866333,
+      "step": 290
+    },
+    {
+      "epoch": 0.954248366013072,
+      "grad_norm": 71.07759184618978,
+      "learning_rate": 3.4567085809127245e-07,
+      "logits/chosen": -0.32530540227890015,
+      "logits/rejected": -0.28533849120140076,
+      "logps/chosen": -128.11807250976562,
+      "logps/rejected": -138.7263641357422,
+      "loss": 0.0453,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": -0.014445440843701363,
+      "rewards/margins": 0.01683320477604866,
+      "rewards/rejected": -0.031278643757104874,
+      "step": 292
+    },
+    {
+      "epoch": 0.9607843137254902,
+      "grad_norm": 78.25744070118225,
+      "learning_rate": 3.428292984879594e-07,
+      "logits/chosen": -0.5519564151763916,
+      "logits/rejected": -0.5303467512130737,
+      "logps/chosen": -179.47283935546875,
+      "logps/rejected": -220.14447021484375,
+      "loss": 0.034,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.0216342993080616,
+      "rewards/margins": 0.03530774265527725,
+      "rewards/rejected": -0.05694204568862915,
+      "step": 294
+    },
+    {
+      "epoch": 0.9673202614379085,
+      "grad_norm": 60.10719130500669,
+      "learning_rate": 3.3997375913374705e-07,
+      "logits/chosen": -0.5333956480026245,
+      "logits/rejected": -0.6568067669868469,
+      "logps/chosen": -226.55911254882812,
+      "logps/rejected": -218.00527954101562,
+      "loss": 0.037,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": -0.026494504883885384,
+      "rewards/margins": 0.022691309452056885,
+      "rewards/rejected": -0.04918581247329712,
+      "step": 296
+    },
+    {
+      "epoch": 0.9738562091503268,
+      "grad_norm": 144.12564771661113,
+      "learning_rate": 3.371046700623586e-07,
+      "logits/chosen": -0.3042282462120056,
+      "logits/rejected": -0.3441520929336548,
+      "logps/chosen": -202.99655151367188,
+      "logps/rejected": -185.23617553710938,
+      "loss": 0.0474,
+      "rewards/accuracies": 0.4375,
+      "rewards/chosen": -0.029465947300195694,
+      "rewards/margins": 0.015123619697988033,
+      "rewards/rejected": -0.0445895679295063,
+      "step": 298
+    },
+    {
+      "epoch": 0.9803921568627451,
+      "grad_norm": 47.18163387782249,
+      "learning_rate": 3.3422246334805497e-07,
+      "logits/chosen": -0.3935033082962036,
+      "logits/rejected": -0.4153602719306946,
+      "logps/chosen": -186.66470336914062,
+      "logps/rejected": -193.68746948242188,
+      "loss": 0.0351,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.0020712590776383877,
+      "rewards/margins": 0.07452717423439026,
+      "rewards/rejected": -0.07659843564033508,
+      "step": 300
+    },
+    {
+      "epoch": 0.9803921568627451,
+      "eval_logits/chosen": -0.27245596051216125,
+      "eval_logits/rejected": -0.2619582712650299,
+      "eval_logps/chosen": -225.4788360595703,
+      "eval_logps/rejected": -241.60293579101562,
+      "eval_loss": 0.08713487535715103,
+      "eval_rewards/accuracies": 0.692307710647583,
+      "eval_rewards/chosen": -0.024169014766812325,
+      "eval_rewards/margins": 0.02317720651626587,
+      "eval_rewards/rejected": -0.047346215695142746,
+      "eval_runtime": 14.2235,
+      "eval_samples_per_second": 7.031,
+      "eval_steps_per_second": 0.914,
+      "step": 300
+    }
+  ],
+  "logging_steps": 2,
+  "max_steps": 612,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 2,
+  "save_steps": 100,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 0.0,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}

GRPO-Qwen2.5-7B/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a2ce3404beebbc8dfd4654745c82801487d8ddcb5e0ecb07e2948967bb0ce623
+size 7697

GRPO-Qwen2.5-7B/zero_to_fp32.py

@@ -0,0 +1,760 @@
+#!/usr/bin/env python
+
+# Copyright (c) Microsoft Corporation.
+# SPDX-License-Identifier: Apache-2.0
+
+# DeepSpeed Team
+
+# This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
+# copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
+# the future. Once extracted, the weights don't require DeepSpeed and can be used in any
+# application.
+#
+# example:
+#   python zero_to_fp32.py . output_dir/
+#   or
+#   python zero_to_fp32.py . output_dir/ --safe_serialization
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+import gc
+import json
+import numpy as np
+from tqdm import tqdm
+from collections import OrderedDict
+from dataclasses import dataclass
+
+# while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
+# DeepSpeed data structures it has to be available in the current python environment.
+from deepspeed.utils import logger
+from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
+                                            FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
+                                            FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
+
+
+@dataclass
+class zero_model_state:
+    buffers: dict()
+    param_shapes: dict()
+    shared_params: list
+    ds_version: int
+    frozen_param_shapes: dict()
+    frozen_param_fragments: dict()
+
+
+debug = 0
+
+# load to cpu
+device = torch.device('cpu')
+
+
+def atoi(text):
+    return int(text) if text.isdigit() else text
+
+
+def natural_keys(text):
+    '''
+    alist.sort(key=natural_keys) sorts in human order
+    http://nedbatchelder.com/blog/200712/human_sorting.html
+    (See Toothy's implementation in the comments)
+    '''
+    return [atoi(c) for c in re.split(r'(\d+)', text)]
+
+
+def get_model_state_file(checkpoint_dir, zero_stage):
+    if not os.path.isdir(checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
+
+    # there should be only one file
+    if zero_stage <= 2:
+        file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
+    elif zero_stage == 3:
+        file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
+
+    if not os.path.exists(file):
+        raise FileNotFoundError(f"can't find model states file at '{file}'")
+
+    return file
+
+
+def get_checkpoint_files(checkpoint_dir, glob_pattern):
+    # XXX: need to test that this simple glob rule works for multi-node setup too
+    ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
+
+    if len(ckpt_files) == 0:
+        raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
+
+    return ckpt_files
+
+
+def get_optim_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
+
+
+def get_model_state_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
+
+
+def parse_model_states(files):
+    zero_model_states = []
+    for file in files:
+        state_dict = torch.load(file, map_location=device, weights_only=False)
+
+        if BUFFER_NAMES not in state_dict:
+            raise ValueError(f"{file} is not a model state checkpoint")
+        buffer_names = state_dict[BUFFER_NAMES]
+        if debug:
+            print("Found buffers:", buffer_names)
+
+        # recover just the buffers while restoring them to fp32 if they were saved in fp16
+        buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
+        param_shapes = state_dict[PARAM_SHAPES]
+
+        # collect parameters that are included in param_shapes
+        param_names = []
+        for s in param_shapes:
+            for name in s.keys():
+                param_names.append(name)
+
+        # update with frozen parameters
+        frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
+        if frozen_param_shapes is not None:
+            if debug:
+                print(f"Found frozen_param_shapes: {frozen_param_shapes}")
+            param_names += list(frozen_param_shapes.keys())
+
+        # handle shared params
+        shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
+
+        ds_version = state_dict.get(DS_VERSION, None)
+
+        frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
+
+        z_model_state = zero_model_state(buffers=buffers,
+                                         param_shapes=param_shapes,
+                                         shared_params=shared_params,
+                                         ds_version=ds_version,
+                                         frozen_param_shapes=frozen_param_shapes,
+                                         frozen_param_fragments=frozen_param_fragments)
+        zero_model_states.append(z_model_state)
+
+    return zero_model_states
+
+
+def parse_optim_states(files, ds_checkpoint_dir):
+    total_files = len(files)
+    state_dicts = []
+    for f in tqdm(files, desc='Loading checkpoint shards'):
+        state_dict = torch.load(f, map_location=device, mmap=True, weights_only=False)
+        # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
+        # and also handle the case where it was already removed by another helper script
+        state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
+        state_dicts.append(state_dict)
+
+    if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
+        raise ValueError(f"{files[0]} is not a zero checkpoint")
+    zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
+    world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
+
+    # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
+    # parameters can be different from data parallelism for non-expert parameters. So we can just
+    # use the max of the partition_count to get the dp world_size.
+
+    if type(world_size) is list:
+        world_size = max(world_size)
+
+    if world_size != total_files:
+        raise ValueError(
+            f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
+            "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
+        )
+
+    # the groups are named differently in each stage
+    if zero_stage <= 2:
+        fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
+    elif zero_stage == 3:
+        fp32_groups_key = FP32_FLAT_GROUPS
+    else:
+        raise ValueError(f"unknown zero stage {zero_stage}")
+
+    fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
+    return zero_stage, world_size, fp32_flat_groups
+
+
+def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters):
+    """
+    Returns fp32 state_dict reconstructed from ds checkpoint
+
+    Args:
+        - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
+
+    """
+    print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
+
+    optim_files = get_optim_files(ds_checkpoint_dir)
+    zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
+    print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
+
+    model_files = get_model_state_files(ds_checkpoint_dir)
+
+    zero_model_states = parse_model_states(model_files)
+    print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
+
+    if zero_stage <= 2:
+        return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+
+
+def _zero2_merge_frozen_params(state_dict, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+    frozen_param_fragments = zero_model_states[0].frozen_param_fragments
+
+    if debug:
+        num_elem = sum(s.numel() for s in frozen_param_shapes.values())
+        print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        state_dict[name] = frozen_param_fragments[name]
+
+        if debug:
+            print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _has_callable(obj, fn):
+    attr = getattr(obj, fn, None)
+    return callable(attr)
+
+
+def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+
+    # Reconstruction protocol:
+    #
+    # XXX: document this
+
+    if debug:
+        for i in range(world_size):
+            for j in range(len(fp32_flat_groups[0])):
+                print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
+
+    # XXX: memory usage doubles here (zero2)
+    num_param_groups = len(fp32_flat_groups[0])
+    merged_single_partition_of_fp32_groups = []
+    for i in range(num_param_groups):
+        merged_partitions = [sd[i] for sd in fp32_flat_groups]
+        full_single_fp32_vector = torch.cat(merged_partitions, 0)
+        merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
+    avail_numel = sum(
+        [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
+
+    if debug:
+        wanted_params = sum([len(shapes) for shapes in param_shapes])
+        wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
+        # not asserting if there is a mismatch due to possible padding
+        print(f"Have {avail_numel} numels to process.")
+        print(f"Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    total_numel = 0
+    total_params = 0
+    for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
+        offset = 0
+        avail_numel = full_single_fp32_vector.numel()
+        for name, shape in shapes.items():
+
+            unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
+            total_numel += unpartitioned_numel
+            total_params += 1
+
+            if debug:
+                print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+            state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
+            offset += unpartitioned_numel
+
+        # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
+        # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
+        # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
+        # live optimizer object, so we are checking that the numbers are within the right range
+        align_to = 2 * world_size
+
+        def zero2_align(x):
+            return align_to * math.ceil(x / align_to)
+
+        if debug:
+            print(f"original offset={offset}, avail_numel={avail_numel}")
+
+        offset = zero2_align(offset)
+        avail_numel = zero2_align(avail_numel)
+
+        if debug:
+            print(f"aligned  offset={offset}, avail_numel={avail_numel}")
+
+        # Sanity check
+        if offset != avail_numel:
+            raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero2_merge_frozen_params(state_dict, zero_model_states)
+
+    _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def zero3_partitioned_param_info(unpartitioned_numel, world_size):
+    remainder = unpartitioned_numel % world_size
+    padding_numel = (world_size - remainder) if remainder else 0
+    partitioned_numel = math.ceil(unpartitioned_numel / world_size)
+    return partitioned_numel, padding_numel
+
+
+def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    if debug:
+        for i in range(world_size):
+            num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
+            print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in zero_model_states[0].frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
+        state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+class GatheredTensor:
+    """
+    A pseudo tensor that collects partitioned weights.
+    It is more memory efficient when there are multiple groups.
+    """
+
+    def __init__(self, flat_groups, flat_groups_offset, offset, partitioned_numel, shape):
+        self.flat_groups = flat_groups
+        self.flat_groups_offset = flat_groups_offset
+        self.offset = offset
+        self.partitioned_numel = partitioned_numel
+        self.shape = shape
+        self.dtype = self.flat_groups[0][0].dtype
+
+    def contiguous(self):
+        """
+        Merge partitioned weights from flat_groups into a single tensor.
+        """
+        end_idx = self.offset + self.partitioned_numel
+        world_size = len(self.flat_groups)
+        pad_flat_param_chunks = []
+
+        for rank_i in range(world_size):
+            # for each rank, we need to collect weights from related group/groups
+            flat_groups_at_rank_i = self.flat_groups[rank_i]
+            start_group_id = None
+            end_group_id = None
+            for group_id in range(len(self.flat_groups_offset)):
+                if self.flat_groups_offset[group_id] <= self.offset < self.flat_groups_offset[group_id + 1]:
+                    start_group_id = group_id
+                if self.flat_groups_offset[group_id] < end_idx <= self.flat_groups_offset[group_id + 1]:
+                    end_group_id = group_id
+                    break
+            # collect weights from related group/groups
+            for group_id in range(start_group_id, end_group_id + 1):
+                flat_tensor = flat_groups_at_rank_i[group_id]
+                start_offset = self.offset - self.flat_groups_offset[group_id]
+                end_offset = min(end_idx, self.flat_groups_offset[group_id + 1]) - self.flat_groups_offset[group_id]
+                pad_flat_param_chunks.append(flat_tensor[start_offset:end_offset])
+
+        # collect weights from all ranks
+        pad_flat_param = torch.cat(pad_flat_param_chunks, dim=0)
+        param = pad_flat_param[:self.shape.numel()].view(self.shape).contiguous()
+        return param
+
+
+def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+    avail_numel = sum([flat_group.numel() for flat_group in fp32_flat_groups[0]]) * world_size
+
+    # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
+    # param, re-consolidating each param, while dealing with padding if any
+
+    # merge list of dicts, preserving order
+    param_shapes = {k: v for d in param_shapes for k, v in d.items()}
+
+    if debug:
+        for i in range(world_size):
+            print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
+
+        wanted_params = len(param_shapes)
+        wanted_numel = sum(shape.numel() for shape in param_shapes.values())
+        # not asserting if there is a mismatch due to possible padding
+        avail_numel = fp32_flat_groups[0].numel() * world_size
+        print(f"Trainable params: Have {avail_numel} numels to process.")
+        print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    offset = 0
+    total_numel = 0
+    total_params = 0
+    flat_groups_offset = [0] + list(np.cumsum([flat_tensor.numel() for flat_tensor in fp32_flat_groups[0]]))
+    for name, shape in tqdm(param_shapes.items(), desc='Gathering sharded weights'):
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+        total_params += 1
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+        # memory efficient tensor
+        tensor = GatheredTensor(fp32_flat_groups, flat_groups_offset, offset, partitioned_numel, shape)
+        state_dict[name] = tensor
+        offset += partitioned_numel
+
+    offset *= world_size
+
+    # Sanity check
+    if offset != avail_numel:
+        raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
+
+    _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def to_torch_tensor(state_dict, return_empty_tensor=False):
+    """
+    Convert state_dict of GatheredTensor to torch tensor
+    """
+    torch_state_dict = {}
+    converted_tensors = {}
+    for name, tensor in state_dict.items():
+        tensor_id = id(tensor)
+        if tensor_id in converted_tensors:  # shared tensors
+            shared_tensor = torch_state_dict[converted_tensors[tensor_id]]
+            torch_state_dict[name] = shared_tensor
+        else:
+            converted_tensors[tensor_id] = name
+            if return_empty_tensor:
+                torch_state_dict[name] = torch.empty(tensor.shape, dtype=tensor.dtype)
+            else:
+                torch_state_dict[name] = tensor.contiguous()
+    return torch_state_dict
+
+
+def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir,
+                                             tag=None,
+                                             exclude_frozen_parameters=False,
+                                             lazy_mode=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
+    ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
+    via a model hub.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+        - ``lazy_mode``: get state_dict in lazy mode. It returns a dict of pesduo tensor instead of torch tensor, which is more memory efficient.
+          Convert the pesduo tensor to torch tensor by ``.contiguous()``
+
+    Returns:
+        - pytorch ``state_dict``
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        # do the training and checkpoint saving
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
+        model = model.cpu() # move to cpu
+        model.load_state_dict(state_dict)
+        # submit to model hub or save the model to share with others
+
+    In this example the ``model`` will no longer be usable in the deepspeed context of the same
+    application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
+
+    Note: the above usage may not work if your application doesn't have sufficient free CPU memory.
+    You may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
+    the checkpoint. Or you can load state_dict in lazy mode ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, lazy_mode=True) # not on cpu
+        for name, lazy_tensor in state_dict.item():
+            tensor = lazy_tensor.contiguous()  # to cpu
+            print(name, tensor)
+            # del tensor to release memory if it no longer in use
+    """
+    if tag is None:
+        latest_path = os.path.join(checkpoint_dir, 'latest')
+        if os.path.isfile(latest_path):
+            with open(latest_path, 'r') as fd:
+                tag = fd.read().strip()
+        else:
+            raise ValueError(f"Unable to find 'latest' file at {latest_path}")
+
+    ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
+
+    if not os.path.isdir(ds_checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
+
+    state_dict = _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
+    if lazy_mode:
+        return state_dict
+    else:
+        return to_torch_tensor(state_dict)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir,
+                                               output_dir,
+                                               max_shard_size="5GB",
+                                               safe_serialization=False,
+                                               tag=None,
+                                               exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
+    loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``output_dir``: directory to the pytorch fp32 state_dict output files
+        - ``max_shard_size``: the maximum size for a checkpoint before being sharded, default value is 5GB
+        - ``safe_serialization``:  whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+    """
+
+    # Dependency pre-check
+    if safe_serialization:
+        try:
+            from safetensors.torch import save_file
+        except ImportError:
+            print('If you want to use `safe_serialization`, please `pip install safetensors`')
+            raise
+    if max_shard_size is not None:
+        try:
+            from huggingface_hub import split_torch_state_dict_into_shards
+        except ImportError:
+            print('If you want to use `max_shard_size`, please `pip install huggingface_hub`')
+            raise
+
+    # Convert zero checkpoint to state_dict
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir,
+                                                          tag,
+                                                          exclude_frozen_parameters,
+                                                          lazy_mode=True)
+
+    # Shard the model if it is too big.
+    weights_name = "model.safetensors" if safe_serialization else "pytorch_model.bin"
+    if max_shard_size is not None:
+        filename_pattern = weights_name.replace(".bin", "{suffix}.bin").replace(".safetensors", "{suffix}.safetensors")
+        # an memory-efficient approach for sharding
+        empty_state_dict = to_torch_tensor(state_dict, return_empty_tensor=True)
+        state_dict_split = split_torch_state_dict_into_shards(empty_state_dict,
+                                                              filename_pattern=filename_pattern,
+                                                              max_shard_size=max_shard_size)
+    else:
+        from collections import namedtuple
+        StateDictSplit = namedtuple("StateDictSplit", ["is_sharded", "filename_to_tensors"])
+        state_dict_split = StateDictSplit(is_sharded=False,
+                                          filename_to_tensors={weights_name: list(state_dict.keys())})
+
+    # Save the model by shard
+    os.makedirs(output_dir, exist_ok=True)
+    filename_to_tensors = state_dict_split.filename_to_tensors.items()
+    for shard_file, tensors in tqdm(filename_to_tensors, desc="Saving checkpoint shards"):
+        shard_state_dict = {tensor_name: state_dict[tensor_name] for tensor_name in tensors}
+        shard_state_dict = to_torch_tensor(shard_state_dict)
+        output_path = os.path.join(output_dir, shard_file)
+        if safe_serialization:
+            save_file(shard_state_dict, output_path, metadata={"format": "pt"})
+        else:
+            torch.save(shard_state_dict, output_path)
+        # release the memory of current shard
+        for tensor_name in list(shard_state_dict.keys()):
+            del state_dict[tensor_name]
+            del shard_state_dict[tensor_name]
+        del shard_state_dict
+        gc.collect()
+
+    # Save index if sharded
+    if state_dict_split.is_sharded:
+        index = {
+            "metadata": state_dict_split.metadata,
+            "weight_map": state_dict_split.tensor_to_filename,
+        }
+        save_index_file = "model.safetensors.index.json" if safe_serialization else "pytorch_model.bin.index.json"
+        save_index_file = os.path.join(output_dir, save_index_file)
+        with open(save_index_file, "w", encoding="utf-8") as f:
+            content = json.dumps(index, indent=2, sort_keys=True) + "\n"
+            f.write(content)
+
+
+def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
+    """
+    1. Put the provided model to cpu
+    2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
+    3. Load it into the provided model
+
+    Args:
+        - ``model``: the model object to update
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+
+    Returns:
+        - ``model`: modified model
+
+    Make sure you have plenty of CPU memory available before you call this function. If you don't
+    have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
+    conveniently placed for you in the checkpoint folder.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
+        model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
+        # submit to model hub or save the model to share with others
+
+    Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
+    of the same application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    """
+    logger.info(f"Extracting fp32 weights")
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+
+    logger.info(f"Overwriting model with fp32 weights")
+    model = model.cpu()
+    model.load_state_dict(state_dict, strict=False)
+
+    return model
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("checkpoint_dir",
+                        type=str,
+                        help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
+    parser.add_argument("output_dir",
+                        type=str,
+                        help="directory to the pytorch fp32 state_dict output files"
+                        "(e.g. path/checkpoint-12-output/)")
+    parser.add_argument(
+        "--max_shard_size",
+        type=str,
+        default="5GB",
+        help="The maximum size for a checkpoint before being sharded. Checkpoints shard will then be each of size"
+        "lower than this size. If expressed as a string, needs to be digits followed by a unit (like `5MB`"
+        "We default it to 5GB in order for models to be able to run easily on free-tier google colab instances"
+        "without CPU OOM issues.")
+    parser.add_argument(
+        "--safe_serialization",
+        default=False,
+        action='store_true',
+        help="Whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).")
+    parser.add_argument("-t",
+                        "--tag",
+                        type=str,
+                        default=None,
+                        help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
+    parser.add_argument("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
+    parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
+    args = parser.parse_args()
+
+    debug = args.debug
+
+    convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir,
+                                               args.output_dir,
+                                               max_shard_size=args.max_shard_size,
+                                               safe_serialization=args.safe_serialization,
+                                               tag=args.tag,
+                                               exclude_frozen_parameters=args.exclude_frozen_parameters)