wtxfrancise commited on
Commit
2f76b25
1 Parent(s): 80f843f
config.json ADDED
@@ -0,0 +1,42 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "/home/wangtianxiang/code/MINI_LLM/model_save/sft_v2/checkpoint-28500",
3
+ "architectures": [
4
+ "QWenLMHeadModel"
5
+ ],
6
+ "attn_dropout_prob": 0.0,
7
+ "auto_map": {
8
+ "AutoConfig": "configuration_qwen.QWenConfig",
9
+ "AutoModelForCausalLM": "modeling_qwen.QWenLMHeadModel"
10
+ },
11
+ "bf16": true,
12
+ "emb_dropout_prob": 0.0,
13
+ "fp16": false,
14
+ "fp32": false,
15
+ "hidden_size": 2048,
16
+ "initializer_range": 0.02,
17
+ "intermediate_size": 11008,
18
+ "kv_channels": 128,
19
+ "layer_norm_epsilon": 1e-06,
20
+ "max_position_embeddings": 8192,
21
+ "model_type": "qwen",
22
+ "no_bias": true,
23
+ "num_attention_heads": 16,
24
+ "num_hidden_layers": 16,
25
+ "onnx_safe": null,
26
+ "rotary_emb_base": 10000,
27
+ "rotary_pct": 1.0,
28
+ "scale_attn_weights": false,
29
+ "seq_length": 2048,
30
+ "softmax_in_fp32": false,
31
+ "tie_word_embeddings": false,
32
+ "tokenizer_class": "QWenTokenizer",
33
+ "torch_dtype": "bfloat16",
34
+ "transformers_version": "4.36.0",
35
+ "use_cache": true,
36
+ "use_cache_kernel": false,
37
+ "use_cache_quantization": false,
38
+ "use_dynamic_ntk": true,
39
+ "use_flash_attn": false,
40
+ "use_logn_attn": true,
41
+ "vocab_size": 151936
42
+ }
generation_config.json ADDED
@@ -0,0 +1,4 @@
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "transformers_version": "4.36.0"
4
+ }
optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c0b9263dc0e102599edbf9399cca93a86a51f7bda4bc3de8edc7bf8409c13636
3
+ size 4229498
qwen.tiktoken ADDED
The diff for this file is too large to render. See raw diff
 
rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e007934537bebfaaaf80205a3128a1054f728444dcf053f3267790401c632f0a
3
+ size 14308
scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:be6a78aca84ec00ea20de25e4513822f934135281e38063fffa2d07fba3d11b3
3
+ size 1064
special_tokens_map.json ADDED
@@ -0,0 +1,5 @@
 
 
 
 
 
 
1
+ {
2
+ "bos_token": "<|im_end|>",
3
+ "eos_token": "<|im_end|>",
4
+ "pad_token": "<|im_end|>"
5
+ }
tokenizer_config.json ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "added_tokens_decoder": {},
3
+ "auto_map": {
4
+ "AutoTokenizer": [
5
+ "tokenization_qwen.QWenTokenizer",
6
+ null
7
+ ]
8
+ },
9
+ "bos_token": "<|im_end|>",
10
+ "clean_up_tokenization_spaces": true,
11
+ "eos_token": "<|im_end|>",
12
+ "model_max_length": 8192,
13
+ "pad_token": "<|im_end|>",
14
+ "tokenizer_class": "QWenTokenizer"
15
+ }
trainer_state.json ADDED
@@ -0,0 +1,3115 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 3.9011415272082455,
5
+ "eval_steps": 500,
6
+ "global_step": 4400,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.0,
13
+ "learning_rate": 1e-08,
14
+ "logits/chosen": -2.25,
15
+ "logits/rejected": -2.421875,
16
+ "logps/chosen": -256.0,
17
+ "logps/rejected": -81.0,
18
+ "loss": 0.6971,
19
+ "rewards/accuracies": 0.453125,
20
+ "rewards/chosen": -0.011858206242322922,
21
+ "rewards/margins": -0.010589980520308018,
22
+ "rewards/rejected": -0.0012682259548455477,
23
+ "step": 1
24
+ },
25
+ {
26
+ "epoch": 0.02,
27
+ "learning_rate": 2.0000000000000002e-07,
28
+ "logits/chosen": -2.25,
29
+ "logits/rejected": -2.421875,
30
+ "logps/chosen": -248.0,
31
+ "logps/rejected": -71.0,
32
+ "loss": 0.6939,
33
+ "rewards/accuracies": 0.49177631735801697,
34
+ "rewards/chosen": -0.0004125148116145283,
35
+ "rewards/margins": 0.0006769677856937051,
36
+ "rewards/rejected": -0.0010894828010350466,
37
+ "step": 20
38
+ },
39
+ {
40
+ "epoch": 0.04,
41
+ "learning_rate": 4.0000000000000003e-07,
42
+ "logits/chosen": -2.25,
43
+ "logits/rejected": -2.40625,
44
+ "logps/chosen": -240.0,
45
+ "logps/rejected": -71.5,
46
+ "loss": 0.6942,
47
+ "rewards/accuracies": 0.5140625238418579,
48
+ "rewards/chosen": -0.001100730849429965,
49
+ "rewards/margins": -0.00020681759633589536,
50
+ "rewards/rejected": -0.0008939133840613067,
51
+ "step": 40
52
+ },
53
+ {
54
+ "epoch": 0.05,
55
+ "learning_rate": 6.000000000000001e-07,
56
+ "logits/chosen": -2.25,
57
+ "logits/rejected": -2.40625,
58
+ "logps/chosen": -256.0,
59
+ "logps/rejected": -73.0,
60
+ "loss": 0.6913,
61
+ "rewards/accuracies": 0.5179687738418579,
62
+ "rewards/chosen": 0.005297874100506306,
63
+ "rewards/margins": 0.007369426544755697,
64
+ "rewards/rejected": -0.002071552677080035,
65
+ "step": 60
66
+ },
67
+ {
68
+ "epoch": 0.07,
69
+ "learning_rate": 8.000000000000001e-07,
70
+ "logits/chosen": -2.234375,
71
+ "logits/rejected": -2.40625,
72
+ "logps/chosen": -252.0,
73
+ "logps/rejected": -68.0,
74
+ "loss": 0.6827,
75
+ "rewards/accuracies": 0.5992187261581421,
76
+ "rewards/chosen": 0.02091832458972931,
77
+ "rewards/margins": 0.02193695865571499,
78
+ "rewards/rejected": -0.001018635230138898,
79
+ "step": 80
80
+ },
81
+ {
82
+ "epoch": 0.09,
83
+ "learning_rate": 1.0000000000000002e-06,
84
+ "logits/chosen": -2.25,
85
+ "logits/rejected": -2.421875,
86
+ "logps/chosen": -249.0,
87
+ "logps/rejected": -69.5,
88
+ "loss": 0.6689,
89
+ "rewards/accuracies": 0.7437499761581421,
90
+ "rewards/chosen": 0.045813608914613724,
91
+ "rewards/margins": 0.05108557268977165,
92
+ "rewards/rejected": -0.005271967500448227,
93
+ "step": 100
94
+ },
95
+ {
96
+ "epoch": 0.11,
97
+ "learning_rate": 1.2000000000000002e-06,
98
+ "logits/chosen": -2.25,
99
+ "logits/rejected": -2.421875,
100
+ "logps/chosen": -254.0,
101
+ "logps/rejected": -69.5,
102
+ "loss": 0.6495,
103
+ "rewards/accuracies": 0.813281238079071,
104
+ "rewards/chosen": 0.07879128307104111,
105
+ "rewards/margins": 0.09384522587060928,
106
+ "rewards/rejected": -0.01505393348634243,
107
+ "step": 120
108
+ },
109
+ {
110
+ "epoch": 0.12,
111
+ "learning_rate": 1.4000000000000001e-06,
112
+ "logits/chosen": -2.265625,
113
+ "logits/rejected": -2.421875,
114
+ "logps/chosen": -239.0,
115
+ "logps/rejected": -71.5,
116
+ "loss": 0.6311,
117
+ "rewards/accuracies": 0.84375,
118
+ "rewards/chosen": 0.10534026473760605,
119
+ "rewards/margins": 0.13596875965595245,
120
+ "rewards/rejected": -0.03062848187983036,
121
+ "step": 140
122
+ },
123
+ {
124
+ "epoch": 0.14,
125
+ "learning_rate": 1.6000000000000001e-06,
126
+ "logits/chosen": -2.265625,
127
+ "logits/rejected": -2.421875,
128
+ "logps/chosen": -228.0,
129
+ "logps/rejected": -69.5,
130
+ "loss": 0.6082,
131
+ "rewards/accuracies": 0.8656250238418579,
132
+ "rewards/chosen": 0.13130317628383636,
133
+ "rewards/margins": 0.19439265131950378,
134
+ "rewards/rejected": -0.06308946758508682,
135
+ "step": 160
136
+ },
137
+ {
138
+ "epoch": 0.16,
139
+ "learning_rate": 1.8000000000000001e-06,
140
+ "logits/chosen": -2.265625,
141
+ "logits/rejected": -2.421875,
142
+ "logps/chosen": -238.0,
143
+ "logps/rejected": -68.5,
144
+ "loss": 0.5785,
145
+ "rewards/accuracies": 0.874218761920929,
146
+ "rewards/chosen": 0.17366917431354523,
147
+ "rewards/margins": 0.2717178463935852,
148
+ "rewards/rejected": -0.09804867208003998,
149
+ "step": 180
150
+ },
151
+ {
152
+ "epoch": 0.18,
153
+ "learning_rate": 2.0000000000000003e-06,
154
+ "logits/chosen": -2.28125,
155
+ "logits/rejected": -2.4375,
156
+ "logps/chosen": -234.0,
157
+ "logps/rejected": -70.0,
158
+ "loss": 0.548,
159
+ "rewards/accuracies": 0.8999999761581421,
160
+ "rewards/chosen": 0.18854746222496033,
161
+ "rewards/margins": 0.355171263217926,
162
+ "rewards/rejected": -0.1666238009929657,
163
+ "step": 200
164
+ },
165
+ {
166
+ "epoch": 0.2,
167
+ "learning_rate": 2.2e-06,
168
+ "logits/chosen": -2.296875,
169
+ "logits/rejected": -2.453125,
170
+ "logps/chosen": -247.0,
171
+ "logps/rejected": -75.5,
172
+ "loss": 0.5123,
173
+ "rewards/accuracies": 0.885937511920929,
174
+ "rewards/chosen": 0.20314271748065948,
175
+ "rewards/margins": 0.4698442816734314,
176
+ "rewards/rejected": -0.2667016088962555,
177
+ "step": 220
178
+ },
179
+ {
180
+ "epoch": 0.21,
181
+ "learning_rate": 2.4000000000000003e-06,
182
+ "logits/chosen": -2.3125,
183
+ "logits/rejected": -2.46875,
184
+ "logps/chosen": -243.0,
185
+ "logps/rejected": -75.5,
186
+ "loss": 0.4766,
187
+ "rewards/accuracies": 0.9125000238418579,
188
+ "rewards/chosen": 0.18197833001613617,
189
+ "rewards/margins": 0.5841859579086304,
190
+ "rewards/rejected": -0.4022076725959778,
191
+ "step": 240
192
+ },
193
+ {
194
+ "epoch": 0.23,
195
+ "learning_rate": 2.6e-06,
196
+ "logits/chosen": -2.3125,
197
+ "logits/rejected": -2.46875,
198
+ "logps/chosen": -243.0,
199
+ "logps/rejected": -77.0,
200
+ "loss": 0.4487,
201
+ "rewards/accuracies": 0.890625,
202
+ "rewards/chosen": 0.12376229465007782,
203
+ "rewards/margins": 0.7004261612892151,
204
+ "rewards/rejected": -0.5766638517379761,
205
+ "step": 260
206
+ },
207
+ {
208
+ "epoch": 0.25,
209
+ "learning_rate": 2.8000000000000003e-06,
210
+ "logits/chosen": -2.328125,
211
+ "logits/rejected": -2.5,
212
+ "logps/chosen": -238.0,
213
+ "logps/rejected": -76.0,
214
+ "loss": 0.4131,
215
+ "rewards/accuracies": 0.910937488079071,
216
+ "rewards/chosen": 0.08943221718072891,
217
+ "rewards/margins": 0.8512886762619019,
218
+ "rewards/rejected": -0.7618564963340759,
219
+ "step": 280
220
+ },
221
+ {
222
+ "epoch": 0.27,
223
+ "learning_rate": 3e-06,
224
+ "logits/chosen": -2.34375,
225
+ "logits/rejected": -2.515625,
226
+ "logps/chosen": -232.0,
227
+ "logps/rejected": -77.5,
228
+ "loss": 0.3867,
229
+ "rewards/accuracies": 0.9234374761581421,
230
+ "rewards/chosen": 3.435462713241577e-05,
231
+ "rewards/margins": 0.9718645215034485,
232
+ "rewards/rejected": -0.9718301892280579,
233
+ "step": 300
234
+ },
235
+ {
236
+ "epoch": 0.28,
237
+ "learning_rate": 3.2000000000000003e-06,
238
+ "logits/chosen": -2.359375,
239
+ "logits/rejected": -2.53125,
240
+ "logps/chosen": -246.0,
241
+ "logps/rejected": -82.0,
242
+ "loss": 0.3574,
243
+ "rewards/accuracies": 0.91796875,
244
+ "rewards/chosen": -0.028441106900572777,
245
+ "rewards/margins": 1.1482356786727905,
246
+ "rewards/rejected": -1.1766767501831055,
247
+ "step": 320
248
+ },
249
+ {
250
+ "epoch": 0.3,
251
+ "learning_rate": 3.4000000000000005e-06,
252
+ "logits/chosen": -2.359375,
253
+ "logits/rejected": -2.53125,
254
+ "logps/chosen": -247.0,
255
+ "logps/rejected": -84.0,
256
+ "loss": 0.3378,
257
+ "rewards/accuracies": 0.9156249761581421,
258
+ "rewards/chosen": -0.1456923484802246,
259
+ "rewards/margins": 1.2620505094528198,
260
+ "rewards/rejected": -1.4077427387237549,
261
+ "step": 340
262
+ },
263
+ {
264
+ "epoch": 0.32,
265
+ "learning_rate": 3.6000000000000003e-06,
266
+ "logits/chosen": -2.375,
267
+ "logits/rejected": -2.546875,
268
+ "logps/chosen": -245.0,
269
+ "logps/rejected": -86.0,
270
+ "loss": 0.313,
271
+ "rewards/accuracies": 0.91796875,
272
+ "rewards/chosen": -0.2751704156398773,
273
+ "rewards/margins": 1.4280754327774048,
274
+ "rewards/rejected": -1.703245759010315,
275
+ "step": 360
276
+ },
277
+ {
278
+ "epoch": 0.34,
279
+ "learning_rate": 3.8000000000000005e-06,
280
+ "logits/chosen": -2.359375,
281
+ "logits/rejected": -2.53125,
282
+ "logps/chosen": -244.0,
283
+ "logps/rejected": -91.0,
284
+ "loss": 0.2764,
285
+ "rewards/accuracies": 0.938281238079071,
286
+ "rewards/chosen": -0.4051324427127838,
287
+ "rewards/margins": 1.6381193399429321,
288
+ "rewards/rejected": -2.0432517528533936,
289
+ "step": 380
290
+ },
291
+ {
292
+ "epoch": 0.35,
293
+ "learning_rate": 4.000000000000001e-06,
294
+ "logits/chosen": -2.34375,
295
+ "logits/rejected": -2.53125,
296
+ "logps/chosen": -256.0,
297
+ "logps/rejected": -95.0,
298
+ "loss": 0.2592,
299
+ "rewards/accuracies": 0.936718761920929,
300
+ "rewards/chosen": -0.5756620168685913,
301
+ "rewards/margins": 1.8360493183135986,
302
+ "rewards/rejected": -2.4117112159729004,
303
+ "step": 400
304
+ },
305
+ {
306
+ "epoch": 0.37,
307
+ "learning_rate": 4.2000000000000004e-06,
308
+ "logits/chosen": -2.328125,
309
+ "logits/rejected": -2.5,
310
+ "logps/chosen": -249.0,
311
+ "logps/rejected": -97.5,
312
+ "loss": 0.2379,
313
+ "rewards/accuracies": 0.928906261920929,
314
+ "rewards/chosen": -0.6275719404220581,
315
+ "rewards/margins": 2.029207706451416,
316
+ "rewards/rejected": -2.6567800045013428,
317
+ "step": 420
318
+ },
319
+ {
320
+ "epoch": 0.39,
321
+ "learning_rate": 4.4e-06,
322
+ "logits/chosen": -2.296875,
323
+ "logits/rejected": -2.453125,
324
+ "logps/chosen": -244.0,
325
+ "logps/rejected": -95.0,
326
+ "loss": 0.2296,
327
+ "rewards/accuracies": 0.934374988079071,
328
+ "rewards/chosen": -0.5478817820549011,
329
+ "rewards/margins": 2.155457019805908,
330
+ "rewards/rejected": -2.703339099884033,
331
+ "step": 440
332
+ },
333
+ {
334
+ "epoch": 0.41,
335
+ "learning_rate": 4.600000000000001e-06,
336
+ "logits/chosen": -2.296875,
337
+ "logits/rejected": -2.453125,
338
+ "logps/chosen": -250.0,
339
+ "logps/rejected": -101.0,
340
+ "loss": 0.2055,
341
+ "rewards/accuracies": 0.9359375238418579,
342
+ "rewards/chosen": -0.6783720850944519,
343
+ "rewards/margins": 2.4690537452697754,
344
+ "rewards/rejected": -3.147425651550293,
345
+ "step": 460
346
+ },
347
+ {
348
+ "epoch": 0.43,
349
+ "learning_rate": 4.800000000000001e-06,
350
+ "logits/chosen": -2.265625,
351
+ "logits/rejected": -2.421875,
352
+ "logps/chosen": -254.0,
353
+ "logps/rejected": -108.5,
354
+ "loss": 0.1813,
355
+ "rewards/accuracies": 0.9476562738418579,
356
+ "rewards/chosen": -0.891997218132019,
357
+ "rewards/margins": 2.769514799118042,
358
+ "rewards/rejected": -3.6615123748779297,
359
+ "step": 480
360
+ },
361
+ {
362
+ "epoch": 0.44,
363
+ "learning_rate": 5e-06,
364
+ "logits/chosen": -2.265625,
365
+ "logits/rejected": -2.40625,
366
+ "logps/chosen": -258.0,
367
+ "logps/rejected": -111.5,
368
+ "loss": 0.1712,
369
+ "rewards/accuracies": 0.94140625,
370
+ "rewards/chosen": -1.160636305809021,
371
+ "rewards/margins": 2.947413682937622,
372
+ "rewards/rejected": -4.1080498695373535,
373
+ "step": 500
374
+ },
375
+ {
376
+ "epoch": 0.46,
377
+ "learning_rate": 5.2e-06,
378
+ "logits/chosen": -2.203125,
379
+ "logits/rejected": -2.34375,
380
+ "logps/chosen": -270.0,
381
+ "logps/rejected": -115.5,
382
+ "loss": 0.1571,
383
+ "rewards/accuracies": 0.953906238079071,
384
+ "rewards/chosen": -1.2190247774124146,
385
+ "rewards/margins": 3.16599702835083,
386
+ "rewards/rejected": -4.385021209716797,
387
+ "step": 520
388
+ },
389
+ {
390
+ "epoch": 0.48,
391
+ "learning_rate": 5.400000000000001e-06,
392
+ "logits/chosen": -2.203125,
393
+ "logits/rejected": -2.296875,
394
+ "logps/chosen": -247.0,
395
+ "logps/rejected": -117.5,
396
+ "loss": 0.1566,
397
+ "rewards/accuracies": 0.9507812261581421,
398
+ "rewards/chosen": -1.6336673498153687,
399
+ "rewards/margins": 3.272843837738037,
400
+ "rewards/rejected": -4.906511306762695,
401
+ "step": 540
402
+ },
403
+ {
404
+ "epoch": 0.5,
405
+ "learning_rate": 5.600000000000001e-06,
406
+ "logits/chosen": -2.140625,
407
+ "logits/rejected": -2.234375,
408
+ "logps/chosen": -270.0,
409
+ "logps/rejected": -120.5,
410
+ "loss": 0.1302,
411
+ "rewards/accuracies": 0.96484375,
412
+ "rewards/chosen": -1.5824471712112427,
413
+ "rewards/margins": 3.517754077911377,
414
+ "rewards/rejected": -5.10020112991333,
415
+ "step": 560
416
+ },
417
+ {
418
+ "epoch": 0.51,
419
+ "learning_rate": 5.8e-06,
420
+ "logits/chosen": -2.15625,
421
+ "logits/rejected": -2.203125,
422
+ "logps/chosen": -270.0,
423
+ "logps/rejected": -127.5,
424
+ "loss": 0.1328,
425
+ "rewards/accuracies": 0.957812488079071,
426
+ "rewards/chosen": -1.754668951034546,
427
+ "rewards/margins": 3.8442912101745605,
428
+ "rewards/rejected": -5.598960876464844,
429
+ "step": 580
430
+ },
431
+ {
432
+ "epoch": 0.53,
433
+ "learning_rate": 6e-06,
434
+ "logits/chosen": -2.046875,
435
+ "logits/rejected": -2.0625,
436
+ "logps/chosen": -266.0,
437
+ "logps/rejected": -127.5,
438
+ "loss": 0.11,
439
+ "rewards/accuracies": 0.9609375,
440
+ "rewards/chosen": -1.844238519668579,
441
+ "rewards/margins": 4.067028522491455,
442
+ "rewards/rejected": -5.911267280578613,
443
+ "step": 600
444
+ },
445
+ {
446
+ "epoch": 0.55,
447
+ "learning_rate": 6.200000000000001e-06,
448
+ "logits/chosen": -2.015625,
449
+ "logits/rejected": -1.9921875,
450
+ "logps/chosen": -262.0,
451
+ "logps/rejected": -133.0,
452
+ "loss": 0.1116,
453
+ "rewards/accuracies": 0.9609375,
454
+ "rewards/chosen": -2.0047552585601807,
455
+ "rewards/margins": 4.232540130615234,
456
+ "rewards/rejected": -6.237295150756836,
457
+ "step": 620
458
+ },
459
+ {
460
+ "epoch": 0.57,
461
+ "learning_rate": 6.4000000000000006e-06,
462
+ "logits/chosen": -1.9921875,
463
+ "logits/rejected": -1.984375,
464
+ "logps/chosen": -256.0,
465
+ "logps/rejected": -134.0,
466
+ "loss": 0.1099,
467
+ "rewards/accuracies": 0.95703125,
468
+ "rewards/chosen": -1.988957405090332,
469
+ "rewards/margins": 4.430800437927246,
470
+ "rewards/rejected": -6.4197587966918945,
471
+ "step": 640
472
+ },
473
+ {
474
+ "epoch": 0.59,
475
+ "learning_rate": 6.600000000000001e-06,
476
+ "logits/chosen": -1.984375,
477
+ "logits/rejected": -1.9765625,
478
+ "logps/chosen": -268.0,
479
+ "logps/rejected": -134.0,
480
+ "loss": 0.1005,
481
+ "rewards/accuracies": 0.9632812738418579,
482
+ "rewards/chosen": -1.875688910484314,
483
+ "rewards/margins": 4.4820122718811035,
484
+ "rewards/rejected": -6.357701301574707,
485
+ "step": 660
486
+ },
487
+ {
488
+ "epoch": 0.6,
489
+ "learning_rate": 6.800000000000001e-06,
490
+ "logits/chosen": -1.96875,
491
+ "logits/rejected": -1.953125,
492
+ "logps/chosen": -278.0,
493
+ "logps/rejected": -143.0,
494
+ "loss": 0.0932,
495
+ "rewards/accuracies": 0.967968761920929,
496
+ "rewards/chosen": -2.173671245574951,
497
+ "rewards/margins": 4.973406791687012,
498
+ "rewards/rejected": -7.1470770835876465,
499
+ "step": 680
500
+ },
501
+ {
502
+ "epoch": 0.62,
503
+ "learning_rate": 7e-06,
504
+ "logits/chosen": -1.9765625,
505
+ "logits/rejected": -1.9609375,
506
+ "logps/chosen": -258.0,
507
+ "logps/rejected": -138.0,
508
+ "loss": 0.0932,
509
+ "rewards/accuracies": 0.9671875238418579,
510
+ "rewards/chosen": -2.037585735321045,
511
+ "rewards/margins": 4.935227394104004,
512
+ "rewards/rejected": -6.972813606262207,
513
+ "step": 700
514
+ },
515
+ {
516
+ "epoch": 0.64,
517
+ "learning_rate": 7.2000000000000005e-06,
518
+ "logits/chosen": -1.8828125,
519
+ "logits/rejected": -1.84375,
520
+ "logps/chosen": -268.0,
521
+ "logps/rejected": -142.0,
522
+ "loss": 0.0866,
523
+ "rewards/accuracies": 0.9750000238418579,
524
+ "rewards/chosen": -2.1355109214782715,
525
+ "rewards/margins": 5.11392879486084,
526
+ "rewards/rejected": -7.2494401931762695,
527
+ "step": 720
528
+ },
529
+ {
530
+ "epoch": 0.66,
531
+ "learning_rate": 7.4e-06,
532
+ "logits/chosen": -1.8125,
533
+ "logits/rejected": -1.7578125,
534
+ "logps/chosen": -266.0,
535
+ "logps/rejected": -141.0,
536
+ "loss": 0.0873,
537
+ "rewards/accuracies": 0.9664062261581421,
538
+ "rewards/chosen": -2.000317335128784,
539
+ "rewards/margins": 5.19692325592041,
540
+ "rewards/rejected": -7.197241306304932,
541
+ "step": 740
542
+ },
543
+ {
544
+ "epoch": 0.67,
545
+ "learning_rate": 7.600000000000001e-06,
546
+ "logits/chosen": -1.765625,
547
+ "logits/rejected": -1.6875,
548
+ "logps/chosen": -260.0,
549
+ "logps/rejected": -143.0,
550
+ "loss": 0.0798,
551
+ "rewards/accuracies": 0.97265625,
552
+ "rewards/chosen": -1.756317138671875,
553
+ "rewards/margins": 5.629761219024658,
554
+ "rewards/rejected": -7.386077880859375,
555
+ "step": 760
556
+ },
557
+ {
558
+ "epoch": 0.69,
559
+ "learning_rate": 7.800000000000002e-06,
560
+ "logits/chosen": -1.78125,
561
+ "logits/rejected": -1.7265625,
562
+ "logps/chosen": -272.0,
563
+ "logps/rejected": -151.0,
564
+ "loss": 0.0707,
565
+ "rewards/accuracies": 0.9710937738418579,
566
+ "rewards/chosen": -2.1809704303741455,
567
+ "rewards/margins": 5.904406547546387,
568
+ "rewards/rejected": -8.08537769317627,
569
+ "step": 780
570
+ },
571
+ {
572
+ "epoch": 0.71,
573
+ "learning_rate": 8.000000000000001e-06,
574
+ "logits/chosen": -1.71875,
575
+ "logits/rejected": -1.6640625,
576
+ "logps/chosen": -260.0,
577
+ "logps/rejected": -147.0,
578
+ "loss": 0.0753,
579
+ "rewards/accuracies": 0.967968761920929,
580
+ "rewards/chosen": -1.8227026462554932,
581
+ "rewards/margins": 5.907409191131592,
582
+ "rewards/rejected": -7.730111598968506,
583
+ "step": 800
584
+ },
585
+ {
586
+ "epoch": 0.73,
587
+ "learning_rate": 8.2e-06,
588
+ "logits/chosen": -1.7265625,
589
+ "logits/rejected": -1.6953125,
590
+ "logps/chosen": -255.0,
591
+ "logps/rejected": -147.0,
592
+ "loss": 0.0651,
593
+ "rewards/accuracies": 0.9765625,
594
+ "rewards/chosen": -1.8092275857925415,
595
+ "rewards/margins": 6.15098237991333,
596
+ "rewards/rejected": -7.960209846496582,
597
+ "step": 820
598
+ },
599
+ {
600
+ "epoch": 0.74,
601
+ "learning_rate": 8.400000000000001e-06,
602
+ "logits/chosen": -1.75,
603
+ "logits/rejected": -1.7421875,
604
+ "logps/chosen": -255.0,
605
+ "logps/rejected": -147.0,
606
+ "loss": 0.1021,
607
+ "rewards/accuracies": 0.969531238079071,
608
+ "rewards/chosen": -1.6101219654083252,
609
+ "rewards/margins": 6.002054691314697,
610
+ "rewards/rejected": -7.612176418304443,
611
+ "step": 840
612
+ },
613
+ {
614
+ "epoch": 0.76,
615
+ "learning_rate": 8.6e-06,
616
+ "logits/chosen": -1.734375,
617
+ "logits/rejected": -1.6953125,
618
+ "logps/chosen": -253.0,
619
+ "logps/rejected": -146.0,
620
+ "loss": 0.0796,
621
+ "rewards/accuracies": 0.9609375,
622
+ "rewards/chosen": -1.733062505722046,
623
+ "rewards/margins": 5.910679817199707,
624
+ "rewards/rejected": -7.643742561340332,
625
+ "step": 860
626
+ },
627
+ {
628
+ "epoch": 0.78,
629
+ "learning_rate": 8.8e-06,
630
+ "logits/chosen": -1.703125,
631
+ "logits/rejected": -1.6796875,
632
+ "logps/chosen": -258.0,
633
+ "logps/rejected": -148.0,
634
+ "loss": 0.099,
635
+ "rewards/accuracies": 0.9585937261581421,
636
+ "rewards/chosen": -1.9996089935302734,
637
+ "rewards/margins": 5.942723751068115,
638
+ "rewards/rejected": -7.9423322677612305,
639
+ "step": 880
640
+ },
641
+ {
642
+ "epoch": 0.8,
643
+ "learning_rate": 9e-06,
644
+ "logits/chosen": -1.6015625,
645
+ "logits/rejected": -1.5703125,
646
+ "logps/chosen": -260.0,
647
+ "logps/rejected": -149.0,
648
+ "loss": 0.0651,
649
+ "rewards/accuracies": 0.9750000238418579,
650
+ "rewards/chosen": -1.6655864715576172,
651
+ "rewards/margins": 6.289559364318848,
652
+ "rewards/rejected": -7.955146789550781,
653
+ "step": 900
654
+ },
655
+ {
656
+ "epoch": 0.82,
657
+ "learning_rate": 9.200000000000002e-06,
658
+ "logits/chosen": -1.6953125,
659
+ "logits/rejected": -1.703125,
660
+ "logps/chosen": -252.0,
661
+ "logps/rejected": -148.0,
662
+ "loss": 0.0676,
663
+ "rewards/accuracies": 0.973437488079071,
664
+ "rewards/chosen": -1.2225464582443237,
665
+ "rewards/margins": 6.570587158203125,
666
+ "rewards/rejected": -7.7931342124938965,
667
+ "step": 920
668
+ },
669
+ {
670
+ "epoch": 0.83,
671
+ "learning_rate": 9.4e-06,
672
+ "logits/chosen": -1.6484375,
673
+ "logits/rejected": -1.6171875,
674
+ "logps/chosen": -264.0,
675
+ "logps/rejected": -154.0,
676
+ "loss": 0.0582,
677
+ "rewards/accuracies": 0.984375,
678
+ "rewards/chosen": -1.5915133953094482,
679
+ "rewards/margins": 6.880612850189209,
680
+ "rewards/rejected": -8.472126007080078,
681
+ "step": 940
682
+ },
683
+ {
684
+ "epoch": 0.85,
685
+ "learning_rate": 9.600000000000001e-06,
686
+ "logits/chosen": -1.6796875,
687
+ "logits/rejected": -1.6796875,
688
+ "logps/chosen": -242.0,
689
+ "logps/rejected": -148.0,
690
+ "loss": 0.0659,
691
+ "rewards/accuracies": 0.97265625,
692
+ "rewards/chosen": -1.3238825798034668,
693
+ "rewards/margins": 6.604360103607178,
694
+ "rewards/rejected": -7.928241729736328,
695
+ "step": 960
696
+ },
697
+ {
698
+ "epoch": 0.87,
699
+ "learning_rate": 9.800000000000001e-06,
700
+ "logits/chosen": -1.5859375,
701
+ "logits/rejected": -1.546875,
702
+ "logps/chosen": -264.0,
703
+ "logps/rejected": -159.0,
704
+ "loss": 0.0621,
705
+ "rewards/accuracies": 0.9742187261581421,
706
+ "rewards/chosen": -1.903455376625061,
707
+ "rewards/margins": 6.904341220855713,
708
+ "rewards/rejected": -8.807796478271484,
709
+ "step": 980
710
+ },
711
+ {
712
+ "epoch": 0.89,
713
+ "learning_rate": 1e-05,
714
+ "logits/chosen": -1.53125,
715
+ "logits/rejected": -1.4765625,
716
+ "logps/chosen": -270.0,
717
+ "logps/rejected": -161.0,
718
+ "loss": 0.0598,
719
+ "rewards/accuracies": 0.969531238079071,
720
+ "rewards/chosen": -1.7653465270996094,
721
+ "rewards/margins": 7.352142333984375,
722
+ "rewards/rejected": -9.117487907409668,
723
+ "step": 1000
724
+ },
725
+ {
726
+ "epoch": 0.9,
727
+ "learning_rate": 9.942987457240593e-06,
728
+ "logits/chosen": -1.515625,
729
+ "logits/rejected": -1.4765625,
730
+ "logps/chosen": -258.0,
731
+ "logps/rejected": -166.0,
732
+ "loss": 0.0725,
733
+ "rewards/accuracies": 0.9710937738418579,
734
+ "rewards/chosen": -2.206141948699951,
735
+ "rewards/margins": 7.433462619781494,
736
+ "rewards/rejected": -9.639605522155762,
737
+ "step": 1020
738
+ },
739
+ {
740
+ "epoch": 0.92,
741
+ "learning_rate": 9.885974914481187e-06,
742
+ "logits/chosen": -1.4921875,
743
+ "logits/rejected": -1.5,
744
+ "logps/chosen": -264.0,
745
+ "logps/rejected": -165.0,
746
+ "loss": 0.0579,
747
+ "rewards/accuracies": 0.973437488079071,
748
+ "rewards/chosen": -1.83426034450531,
749
+ "rewards/margins": 7.645236015319824,
750
+ "rewards/rejected": -9.479497909545898,
751
+ "step": 1040
752
+ },
753
+ {
754
+ "epoch": 0.94,
755
+ "learning_rate": 9.828962371721779e-06,
756
+ "logits/chosen": -1.3359375,
757
+ "logits/rejected": -1.28125,
758
+ "logps/chosen": -260.0,
759
+ "logps/rejected": -165.0,
760
+ "loss": 0.0569,
761
+ "rewards/accuracies": 0.9789062738418579,
762
+ "rewards/chosen": -2.0447144508361816,
763
+ "rewards/margins": 7.545592308044434,
764
+ "rewards/rejected": -9.590306282043457,
765
+ "step": 1060
766
+ },
767
+ {
768
+ "epoch": 0.96,
769
+ "learning_rate": 9.771949828962373e-06,
770
+ "logits/chosen": -1.4296875,
771
+ "logits/rejected": -1.390625,
772
+ "logps/chosen": -270.0,
773
+ "logps/rejected": -162.0,
774
+ "loss": 0.071,
775
+ "rewards/accuracies": 0.975781261920929,
776
+ "rewards/chosen": -2.062221050262451,
777
+ "rewards/margins": 7.198528289794922,
778
+ "rewards/rejected": -9.260747909545898,
779
+ "step": 1080
780
+ },
781
+ {
782
+ "epoch": 0.98,
783
+ "learning_rate": 9.714937286202965e-06,
784
+ "logits/chosen": -1.4296875,
785
+ "logits/rejected": -1.359375,
786
+ "logps/chosen": -260.0,
787
+ "logps/rejected": -156.0,
788
+ "loss": 0.0608,
789
+ "rewards/accuracies": 0.973437488079071,
790
+ "rewards/chosen": -1.3428866863250732,
791
+ "rewards/margins": 7.246352195739746,
792
+ "rewards/rejected": -8.589239120483398,
793
+ "step": 1100
794
+ },
795
+ {
796
+ "epoch": 0.99,
797
+ "learning_rate": 9.657924743443559e-06,
798
+ "logits/chosen": -1.4140625,
799
+ "logits/rejected": -1.34375,
800
+ "logps/chosen": -258.0,
801
+ "logps/rejected": -163.0,
802
+ "loss": 0.0705,
803
+ "rewards/accuracies": 0.971875011920929,
804
+ "rewards/chosen": -2.211174488067627,
805
+ "rewards/margins": 7.069491386413574,
806
+ "rewards/rejected": -9.280667304992676,
807
+ "step": 1120
808
+ },
809
+ {
810
+ "epoch": 1.01,
811
+ "learning_rate": 9.600912200684151e-06,
812
+ "logits/chosen": -1.4453125,
813
+ "logits/rejected": -1.3671875,
814
+ "logps/chosen": -251.0,
815
+ "logps/rejected": -162.0,
816
+ "loss": 0.0424,
817
+ "rewards/accuracies": 0.9859374761581421,
818
+ "rewards/chosen": -1.804861307144165,
819
+ "rewards/margins": 7.515401363372803,
820
+ "rewards/rejected": -9.32026195526123,
821
+ "step": 1140
822
+ },
823
+ {
824
+ "epoch": 1.03,
825
+ "learning_rate": 9.543899657924745e-06,
826
+ "logits/chosen": -1.4375,
827
+ "logits/rejected": -1.3828125,
828
+ "logps/chosen": -272.0,
829
+ "logps/rejected": -172.0,
830
+ "loss": 0.0362,
831
+ "rewards/accuracies": 0.987500011920929,
832
+ "rewards/chosen": -2.1582987308502197,
833
+ "rewards/margins": 7.993732452392578,
834
+ "rewards/rejected": -10.152030944824219,
835
+ "step": 1160
836
+ },
837
+ {
838
+ "epoch": 1.05,
839
+ "learning_rate": 9.486887115165337e-06,
840
+ "logits/chosen": -1.359375,
841
+ "logits/rejected": -1.3203125,
842
+ "logps/chosen": -270.0,
843
+ "logps/rejected": -183.0,
844
+ "loss": 0.0376,
845
+ "rewards/accuracies": 0.987500011920929,
846
+ "rewards/chosen": -2.6995084285736084,
847
+ "rewards/margins": 8.586763381958008,
848
+ "rewards/rejected": -11.286273002624512,
849
+ "step": 1180
850
+ },
851
+ {
852
+ "epoch": 1.06,
853
+ "learning_rate": 9.429874572405931e-06,
854
+ "logits/chosen": -1.34375,
855
+ "logits/rejected": -1.2734375,
856
+ "logps/chosen": -260.0,
857
+ "logps/rejected": -171.0,
858
+ "loss": 0.0336,
859
+ "rewards/accuracies": 0.984375,
860
+ "rewards/chosen": -1.5670948028564453,
861
+ "rewards/margins": 8.42174243927002,
862
+ "rewards/rejected": -9.988837242126465,
863
+ "step": 1200
864
+ },
865
+ {
866
+ "epoch": 1.08,
867
+ "learning_rate": 9.372862029646523e-06,
868
+ "logits/chosen": -1.421875,
869
+ "logits/rejected": -1.375,
870
+ "logps/chosen": -258.0,
871
+ "logps/rejected": -172.0,
872
+ "loss": 0.029,
873
+ "rewards/accuracies": 0.9921875,
874
+ "rewards/chosen": -1.5389244556427002,
875
+ "rewards/margins": 8.682077407836914,
876
+ "rewards/rejected": -10.221002578735352,
877
+ "step": 1220
878
+ },
879
+ {
880
+ "epoch": 1.1,
881
+ "learning_rate": 9.315849486887117e-06,
882
+ "logits/chosen": -1.4609375,
883
+ "logits/rejected": -1.390625,
884
+ "logps/chosen": -262.0,
885
+ "logps/rejected": -177.0,
886
+ "loss": 0.0381,
887
+ "rewards/accuracies": 0.987500011920929,
888
+ "rewards/chosen": -2.027346134185791,
889
+ "rewards/margins": 8.802923202514648,
890
+ "rewards/rejected": -10.830268859863281,
891
+ "step": 1240
892
+ },
893
+ {
894
+ "epoch": 1.12,
895
+ "learning_rate": 9.25883694412771e-06,
896
+ "logits/chosen": -1.4296875,
897
+ "logits/rejected": -1.3125,
898
+ "logps/chosen": -264.0,
899
+ "logps/rejected": -173.0,
900
+ "loss": 0.0342,
901
+ "rewards/accuracies": 0.9859374761581421,
902
+ "rewards/chosen": -1.638588309288025,
903
+ "rewards/margins": 8.736043930053711,
904
+ "rewards/rejected": -10.374631881713867,
905
+ "step": 1260
906
+ },
907
+ {
908
+ "epoch": 1.13,
909
+ "learning_rate": 9.201824401368301e-06,
910
+ "logits/chosen": -1.34375,
911
+ "logits/rejected": -1.2421875,
912
+ "logps/chosen": -255.0,
913
+ "logps/rejected": -176.0,
914
+ "loss": 0.0338,
915
+ "rewards/accuracies": 0.9898437261581421,
916
+ "rewards/chosen": -1.8327977657318115,
917
+ "rewards/margins": 9.06210708618164,
918
+ "rewards/rejected": -10.894905090332031,
919
+ "step": 1280
920
+ },
921
+ {
922
+ "epoch": 1.15,
923
+ "learning_rate": 9.144811858608895e-06,
924
+ "logits/chosen": -1.375,
925
+ "logits/rejected": -1.296875,
926
+ "logps/chosen": -268.0,
927
+ "logps/rejected": -178.0,
928
+ "loss": 0.0301,
929
+ "rewards/accuracies": 0.987500011920929,
930
+ "rewards/chosen": -1.7202352285385132,
931
+ "rewards/margins": 9.098135948181152,
932
+ "rewards/rejected": -10.81837272644043,
933
+ "step": 1300
934
+ },
935
+ {
936
+ "epoch": 1.17,
937
+ "learning_rate": 9.087799315849487e-06,
938
+ "logits/chosen": -1.4296875,
939
+ "logits/rejected": -1.34375,
940
+ "logps/chosen": -264.0,
941
+ "logps/rejected": -178.0,
942
+ "loss": 0.0312,
943
+ "rewards/accuracies": 0.991406261920929,
944
+ "rewards/chosen": -1.8896671533584595,
945
+ "rewards/margins": 8.800851821899414,
946
+ "rewards/rejected": -10.69051742553711,
947
+ "step": 1320
948
+ },
949
+ {
950
+ "epoch": 1.19,
951
+ "learning_rate": 9.03078677309008e-06,
952
+ "logits/chosen": -1.46875,
953
+ "logits/rejected": -1.4453125,
954
+ "logps/chosen": -254.0,
955
+ "logps/rejected": -177.0,
956
+ "loss": 0.0318,
957
+ "rewards/accuracies": 0.991406261920929,
958
+ "rewards/chosen": -1.7370885610580444,
959
+ "rewards/margins": 9.022928237915039,
960
+ "rewards/rejected": -10.760017395019531,
961
+ "step": 1340
962
+ },
963
+ {
964
+ "epoch": 1.21,
965
+ "learning_rate": 8.973774230330673e-06,
966
+ "logits/chosen": -1.46875,
967
+ "logits/rejected": -1.4375,
968
+ "logps/chosen": -270.0,
969
+ "logps/rejected": -184.0,
970
+ "loss": 0.0296,
971
+ "rewards/accuracies": 0.9906250238418579,
972
+ "rewards/chosen": -2.1952197551727295,
973
+ "rewards/margins": 9.119722366333008,
974
+ "rewards/rejected": -11.314942359924316,
975
+ "step": 1360
976
+ },
977
+ {
978
+ "epoch": 1.22,
979
+ "learning_rate": 8.916761687571266e-06,
980
+ "logits/chosen": -1.40625,
981
+ "logits/rejected": -1.3125,
982
+ "logps/chosen": -266.0,
983
+ "logps/rejected": -179.0,
984
+ "loss": 0.0341,
985
+ "rewards/accuracies": 0.9859374761581421,
986
+ "rewards/chosen": -1.6254527568817139,
987
+ "rewards/margins": 9.138445854187012,
988
+ "rewards/rejected": -10.763897895812988,
989
+ "step": 1380
990
+ },
991
+ {
992
+ "epoch": 1.24,
993
+ "learning_rate": 8.85974914481186e-06,
994
+ "logits/chosen": -1.3515625,
995
+ "logits/rejected": -1.203125,
996
+ "logps/chosen": -255.0,
997
+ "logps/rejected": -171.0,
998
+ "loss": 0.0301,
999
+ "rewards/accuracies": 0.98828125,
1000
+ "rewards/chosen": -1.489253282546997,
1001
+ "rewards/margins": 8.813684463500977,
1002
+ "rewards/rejected": -10.302938461303711,
1003
+ "step": 1400
1004
+ },
1005
+ {
1006
+ "epoch": 1.26,
1007
+ "learning_rate": 8.802736602052452e-06,
1008
+ "logits/chosen": -1.3515625,
1009
+ "logits/rejected": -1.234375,
1010
+ "logps/chosen": -270.0,
1011
+ "logps/rejected": -183.0,
1012
+ "loss": 0.0293,
1013
+ "rewards/accuracies": 0.991406261920929,
1014
+ "rewards/chosen": -1.8335425853729248,
1015
+ "rewards/margins": 9.455236434936523,
1016
+ "rewards/rejected": -11.288779258728027,
1017
+ "step": 1420
1018
+ },
1019
+ {
1020
+ "epoch": 1.28,
1021
+ "learning_rate": 8.745724059293046e-06,
1022
+ "logits/chosen": -1.3359375,
1023
+ "logits/rejected": -1.15625,
1024
+ "logps/chosen": -270.0,
1025
+ "logps/rejected": -185.0,
1026
+ "loss": 0.0321,
1027
+ "rewards/accuracies": 0.9898437261581421,
1028
+ "rewards/chosen": -2.0949649810791016,
1029
+ "rewards/margins": 9.466029167175293,
1030
+ "rewards/rejected": -11.560995101928711,
1031
+ "step": 1440
1032
+ },
1033
+ {
1034
+ "epoch": 1.29,
1035
+ "learning_rate": 8.688711516533638e-06,
1036
+ "logits/chosen": -1.40625,
1037
+ "logits/rejected": -1.2890625,
1038
+ "logps/chosen": -262.0,
1039
+ "logps/rejected": -184.0,
1040
+ "loss": 0.0364,
1041
+ "rewards/accuracies": 0.98828125,
1042
+ "rewards/chosen": -2.0965704917907715,
1043
+ "rewards/margins": 9.323990821838379,
1044
+ "rewards/rejected": -11.420560836791992,
1045
+ "step": 1460
1046
+ },
1047
+ {
1048
+ "epoch": 1.31,
1049
+ "learning_rate": 8.63169897377423e-06,
1050
+ "logits/chosen": -1.453125,
1051
+ "logits/rejected": -1.3125,
1052
+ "logps/chosen": -268.0,
1053
+ "logps/rejected": -184.0,
1054
+ "loss": 0.0286,
1055
+ "rewards/accuracies": 0.98828125,
1056
+ "rewards/chosen": -2.029646396636963,
1057
+ "rewards/margins": 9.428507804870605,
1058
+ "rewards/rejected": -11.458154678344727,
1059
+ "step": 1480
1060
+ },
1061
+ {
1062
+ "epoch": 1.33,
1063
+ "learning_rate": 8.574686431014824e-06,
1064
+ "logits/chosen": -1.4765625,
1065
+ "logits/rejected": -1.328125,
1066
+ "logps/chosen": -254.0,
1067
+ "logps/rejected": -179.0,
1068
+ "loss": 0.0286,
1069
+ "rewards/accuracies": 0.9921875,
1070
+ "rewards/chosen": -1.9806649684906006,
1071
+ "rewards/margins": 9.070594787597656,
1072
+ "rewards/rejected": -11.051259994506836,
1073
+ "step": 1500
1074
+ },
1075
+ {
1076
+ "epoch": 1.35,
1077
+ "learning_rate": 8.517673888255416e-06,
1078
+ "logits/chosen": -1.4609375,
1079
+ "logits/rejected": -1.3671875,
1080
+ "logps/chosen": -268.0,
1081
+ "logps/rejected": -186.0,
1082
+ "loss": 0.0428,
1083
+ "rewards/accuracies": 0.985156238079071,
1084
+ "rewards/chosen": -2.330333709716797,
1085
+ "rewards/margins": 9.129961013793945,
1086
+ "rewards/rejected": -11.460294723510742,
1087
+ "step": 1520
1088
+ },
1089
+ {
1090
+ "epoch": 1.37,
1091
+ "learning_rate": 8.46066134549601e-06,
1092
+ "logits/chosen": -1.3984375,
1093
+ "logits/rejected": -1.265625,
1094
+ "logps/chosen": -260.0,
1095
+ "logps/rejected": -188.0,
1096
+ "loss": 0.045,
1097
+ "rewards/accuracies": 0.9820312261581421,
1098
+ "rewards/chosen": -2.5738627910614014,
1099
+ "rewards/margins": 9.508840560913086,
1100
+ "rewards/rejected": -12.082701683044434,
1101
+ "step": 1540
1102
+ },
1103
+ {
1104
+ "epoch": 1.38,
1105
+ "learning_rate": 8.403648802736602e-06,
1106
+ "logits/chosen": -1.4765625,
1107
+ "logits/rejected": -1.3515625,
1108
+ "logps/chosen": -256.0,
1109
+ "logps/rejected": -179.0,
1110
+ "loss": 0.0215,
1111
+ "rewards/accuracies": 0.9906250238418579,
1112
+ "rewards/chosen": -1.6458170413970947,
1113
+ "rewards/margins": 9.304633140563965,
1114
+ "rewards/rejected": -10.95044994354248,
1115
+ "step": 1560
1116
+ },
1117
+ {
1118
+ "epoch": 1.4,
1119
+ "learning_rate": 8.346636259977196e-06,
1120
+ "logits/chosen": -1.3984375,
1121
+ "logits/rejected": -1.234375,
1122
+ "logps/chosen": -272.0,
1123
+ "logps/rejected": -185.0,
1124
+ "loss": 0.034,
1125
+ "rewards/accuracies": 0.985156238079071,
1126
+ "rewards/chosen": -1.9806787967681885,
1127
+ "rewards/margins": 9.267131805419922,
1128
+ "rewards/rejected": -11.247811317443848,
1129
+ "step": 1580
1130
+ },
1131
+ {
1132
+ "epoch": 1.42,
1133
+ "learning_rate": 8.289623717217788e-06,
1134
+ "logits/chosen": -1.328125,
1135
+ "logits/rejected": -1.1328125,
1136
+ "logps/chosen": -272.0,
1137
+ "logps/rejected": -196.0,
1138
+ "loss": 0.0282,
1139
+ "rewards/accuracies": 0.991406261920929,
1140
+ "rewards/chosen": -2.805903434753418,
1141
+ "rewards/margins": 9.905074119567871,
1142
+ "rewards/rejected": -12.710977554321289,
1143
+ "step": 1600
1144
+ },
1145
+ {
1146
+ "epoch": 1.44,
1147
+ "learning_rate": 8.232611174458382e-06,
1148
+ "logits/chosen": -1.3203125,
1149
+ "logits/rejected": -1.1015625,
1150
+ "logps/chosen": -264.0,
1151
+ "logps/rejected": -187.0,
1152
+ "loss": 0.0377,
1153
+ "rewards/accuracies": 0.984375,
1154
+ "rewards/chosen": -1.9745209217071533,
1155
+ "rewards/margins": 9.847769737243652,
1156
+ "rewards/rejected": -11.822290420532227,
1157
+ "step": 1620
1158
+ },
1159
+ {
1160
+ "epoch": 1.45,
1161
+ "learning_rate": 8.175598631698974e-06,
1162
+ "logits/chosen": -1.3203125,
1163
+ "logits/rejected": -1.15625,
1164
+ "logps/chosen": -266.0,
1165
+ "logps/rejected": -190.0,
1166
+ "loss": 0.0324,
1167
+ "rewards/accuracies": 0.989062488079071,
1168
+ "rewards/chosen": -1.9384052753448486,
1169
+ "rewards/margins": 10.039579391479492,
1170
+ "rewards/rejected": -11.977984428405762,
1171
+ "step": 1640
1172
+ },
1173
+ {
1174
+ "epoch": 1.47,
1175
+ "learning_rate": 8.118586088939568e-06,
1176
+ "logits/chosen": -1.3671875,
1177
+ "logits/rejected": -1.203125,
1178
+ "logps/chosen": -276.0,
1179
+ "logps/rejected": -191.0,
1180
+ "loss": 0.0222,
1181
+ "rewards/accuracies": 0.991406261920929,
1182
+ "rewards/chosen": -2.1441657543182373,
1183
+ "rewards/margins": 9.81259822845459,
1184
+ "rewards/rejected": -11.956764221191406,
1185
+ "step": 1660
1186
+ },
1187
+ {
1188
+ "epoch": 1.49,
1189
+ "learning_rate": 8.06157354618016e-06,
1190
+ "logits/chosen": -1.3046875,
1191
+ "logits/rejected": -1.15625,
1192
+ "logps/chosen": -256.0,
1193
+ "logps/rejected": -189.0,
1194
+ "loss": 0.0236,
1195
+ "rewards/accuracies": 0.9921875,
1196
+ "rewards/chosen": -2.252146005630493,
1197
+ "rewards/margins": 9.743573188781738,
1198
+ "rewards/rejected": -11.995719909667969,
1199
+ "step": 1680
1200
+ },
1201
+ {
1202
+ "epoch": 1.51,
1203
+ "learning_rate": 8.004561003420754e-06,
1204
+ "logits/chosen": -1.3203125,
1205
+ "logits/rejected": -1.1015625,
1206
+ "logps/chosen": -262.0,
1207
+ "logps/rejected": -190.0,
1208
+ "loss": 0.0292,
1209
+ "rewards/accuracies": 0.987500011920929,
1210
+ "rewards/chosen": -1.8711124658584595,
1211
+ "rewards/margins": 10.123231887817383,
1212
+ "rewards/rejected": -11.994344711303711,
1213
+ "step": 1700
1214
+ },
1215
+ {
1216
+ "epoch": 1.52,
1217
+ "learning_rate": 7.947548460661346e-06,
1218
+ "logits/chosen": -1.3984375,
1219
+ "logits/rejected": -1.171875,
1220
+ "logps/chosen": -255.0,
1221
+ "logps/rejected": -176.0,
1222
+ "loss": 0.0359,
1223
+ "rewards/accuracies": 0.9820312261581421,
1224
+ "rewards/chosen": -1.2544156312942505,
1225
+ "rewards/margins": 9.407190322875977,
1226
+ "rewards/rejected": -10.661605834960938,
1227
+ "step": 1720
1228
+ },
1229
+ {
1230
+ "epoch": 1.54,
1231
+ "learning_rate": 7.89053591790194e-06,
1232
+ "logits/chosen": -1.421875,
1233
+ "logits/rejected": -1.2109375,
1234
+ "logps/chosen": -272.0,
1235
+ "logps/rejected": -189.0,
1236
+ "loss": 0.0313,
1237
+ "rewards/accuracies": 0.9921875,
1238
+ "rewards/chosen": -1.9942920207977295,
1239
+ "rewards/margins": 9.837443351745605,
1240
+ "rewards/rejected": -11.831735610961914,
1241
+ "step": 1740
1242
+ },
1243
+ {
1244
+ "epoch": 1.56,
1245
+ "learning_rate": 7.833523375142532e-06,
1246
+ "logits/chosen": -1.390625,
1247
+ "logits/rejected": -1.1875,
1248
+ "logps/chosen": -260.0,
1249
+ "logps/rejected": -186.0,
1250
+ "loss": 0.0325,
1251
+ "rewards/accuracies": 0.9867187738418579,
1252
+ "rewards/chosen": -1.7988389730453491,
1253
+ "rewards/margins": 9.885848999023438,
1254
+ "rewards/rejected": -11.684687614440918,
1255
+ "step": 1760
1256
+ },
1257
+ {
1258
+ "epoch": 1.58,
1259
+ "learning_rate": 7.776510832383124e-06,
1260
+ "logits/chosen": -1.40625,
1261
+ "logits/rejected": -1.2109375,
1262
+ "logps/chosen": -260.0,
1263
+ "logps/rejected": -182.0,
1264
+ "loss": 0.0261,
1265
+ "rewards/accuracies": 0.9906250238418579,
1266
+ "rewards/chosen": -1.5674057006835938,
1267
+ "rewards/margins": 9.868083000183105,
1268
+ "rewards/rejected": -11.435487747192383,
1269
+ "step": 1780
1270
+ },
1271
+ {
1272
+ "epoch": 1.6,
1273
+ "learning_rate": 7.719498289623718e-06,
1274
+ "logits/chosen": -1.390625,
1275
+ "logits/rejected": -1.203125,
1276
+ "logps/chosen": -268.0,
1277
+ "logps/rejected": -196.0,
1278
+ "loss": 0.0439,
1279
+ "rewards/accuracies": 0.985156238079071,
1280
+ "rewards/chosen": -2.2471280097961426,
1281
+ "rewards/margins": 10.478254318237305,
1282
+ "rewards/rejected": -12.725381851196289,
1283
+ "step": 1800
1284
+ },
1285
+ {
1286
+ "epoch": 1.61,
1287
+ "learning_rate": 7.66248574686431e-06,
1288
+ "logits/chosen": -1.3984375,
1289
+ "logits/rejected": -1.2265625,
1290
+ "logps/chosen": -260.0,
1291
+ "logps/rejected": -192.0,
1292
+ "loss": 0.0255,
1293
+ "rewards/accuracies": 0.9906250238418579,
1294
+ "rewards/chosen": -2.0418763160705566,
1295
+ "rewards/margins": 10.087868690490723,
1296
+ "rewards/rejected": -12.129744529724121,
1297
+ "step": 1820
1298
+ },
1299
+ {
1300
+ "epoch": 1.63,
1301
+ "learning_rate": 7.605473204104904e-06,
1302
+ "logits/chosen": -1.359375,
1303
+ "logits/rejected": -1.1875,
1304
+ "logps/chosen": -254.0,
1305
+ "logps/rejected": -187.0,
1306
+ "loss": 0.0326,
1307
+ "rewards/accuracies": 0.98828125,
1308
+ "rewards/chosen": -1.7770217657089233,
1309
+ "rewards/margins": 9.914891242980957,
1310
+ "rewards/rejected": -11.691912651062012,
1311
+ "step": 1840
1312
+ },
1313
+ {
1314
+ "epoch": 1.65,
1315
+ "learning_rate": 7.548460661345497e-06,
1316
+ "logits/chosen": -1.375,
1317
+ "logits/rejected": -1.171875,
1318
+ "logps/chosen": -264.0,
1319
+ "logps/rejected": -190.0,
1320
+ "loss": 0.0368,
1321
+ "rewards/accuracies": 0.989062488079071,
1322
+ "rewards/chosen": -1.742910385131836,
1323
+ "rewards/margins": 9.918044090270996,
1324
+ "rewards/rejected": -11.660953521728516,
1325
+ "step": 1860
1326
+ },
1327
+ {
1328
+ "epoch": 1.67,
1329
+ "learning_rate": 7.49144811858609e-06,
1330
+ "logits/chosen": -1.3359375,
1331
+ "logits/rejected": -1.1328125,
1332
+ "logps/chosen": -258.0,
1333
+ "logps/rejected": -186.0,
1334
+ "loss": 0.0306,
1335
+ "rewards/accuracies": 0.98828125,
1336
+ "rewards/chosen": -1.813442587852478,
1337
+ "rewards/margins": 9.687917709350586,
1338
+ "rewards/rejected": -11.501361846923828,
1339
+ "step": 1880
1340
+ },
1341
+ {
1342
+ "epoch": 1.68,
1343
+ "learning_rate": 7.4344355758266825e-06,
1344
+ "logits/chosen": -1.3515625,
1345
+ "logits/rejected": -1.15625,
1346
+ "logps/chosen": -260.0,
1347
+ "logps/rejected": -188.0,
1348
+ "loss": 0.0322,
1349
+ "rewards/accuracies": 0.987500011920929,
1350
+ "rewards/chosen": -2.153613328933716,
1351
+ "rewards/margins": 9.764740943908691,
1352
+ "rewards/rejected": -11.918354034423828,
1353
+ "step": 1900
1354
+ },
1355
+ {
1356
+ "epoch": 1.7,
1357
+ "learning_rate": 7.3774230330672755e-06,
1358
+ "logits/chosen": -1.3125,
1359
+ "logits/rejected": -1.171875,
1360
+ "logps/chosen": -264.0,
1361
+ "logps/rejected": -201.0,
1362
+ "loss": 0.0276,
1363
+ "rewards/accuracies": 0.991406261920929,
1364
+ "rewards/chosen": -2.7011101245880127,
1365
+ "rewards/margins": 10.471219062805176,
1366
+ "rewards/rejected": -13.172329902648926,
1367
+ "step": 1920
1368
+ },
1369
+ {
1370
+ "epoch": 1.72,
1371
+ "learning_rate": 7.3204104903078685e-06,
1372
+ "logits/chosen": -1.3828125,
1373
+ "logits/rejected": -1.203125,
1374
+ "logps/chosen": -268.0,
1375
+ "logps/rejected": -189.0,
1376
+ "loss": 0.0355,
1377
+ "rewards/accuracies": 0.985156238079071,
1378
+ "rewards/chosen": -1.7472997903823853,
1379
+ "rewards/margins": 10.209493637084961,
1380
+ "rewards/rejected": -11.956793785095215,
1381
+ "step": 1940
1382
+ },
1383
+ {
1384
+ "epoch": 1.74,
1385
+ "learning_rate": 7.2633979475484615e-06,
1386
+ "logits/chosen": -1.375,
1387
+ "logits/rejected": -1.25,
1388
+ "logps/chosen": -268.0,
1389
+ "logps/rejected": -209.0,
1390
+ "loss": 0.0304,
1391
+ "rewards/accuracies": 0.985156238079071,
1392
+ "rewards/chosen": -2.909478187561035,
1393
+ "rewards/margins": 10.954832077026367,
1394
+ "rewards/rejected": -13.864311218261719,
1395
+ "step": 1960
1396
+ },
1397
+ {
1398
+ "epoch": 1.76,
1399
+ "learning_rate": 7.2063854047890545e-06,
1400
+ "logits/chosen": -1.3125,
1401
+ "logits/rejected": -1.1640625,
1402
+ "logps/chosen": -272.0,
1403
+ "logps/rejected": -204.0,
1404
+ "loss": 0.0285,
1405
+ "rewards/accuracies": 0.9921875,
1406
+ "rewards/chosen": -2.4843478202819824,
1407
+ "rewards/margins": 11.059233665466309,
1408
+ "rewards/rejected": -13.54358196258545,
1409
+ "step": 1980
1410
+ },
1411
+ {
1412
+ "epoch": 1.77,
1413
+ "learning_rate": 7.1493728620296476e-06,
1414
+ "logits/chosen": -1.3828125,
1415
+ "logits/rejected": -1.2421875,
1416
+ "logps/chosen": -264.0,
1417
+ "logps/rejected": -194.0,
1418
+ "loss": 0.0225,
1419
+ "rewards/accuracies": 0.9921875,
1420
+ "rewards/chosen": -1.8076410293579102,
1421
+ "rewards/margins": 10.52374267578125,
1422
+ "rewards/rejected": -12.331381797790527,
1423
+ "step": 2000
1424
+ },
1425
+ {
1426
+ "epoch": 1.79,
1427
+ "learning_rate": 7.09236031927024e-06,
1428
+ "logits/chosen": -1.375,
1429
+ "logits/rejected": -1.234375,
1430
+ "logps/chosen": -251.0,
1431
+ "logps/rejected": -189.0,
1432
+ "loss": 0.0268,
1433
+ "rewards/accuracies": 0.98828125,
1434
+ "rewards/chosen": -1.8604707717895508,
1435
+ "rewards/margins": 10.113370895385742,
1436
+ "rewards/rejected": -11.973841667175293,
1437
+ "step": 2020
1438
+ },
1439
+ {
1440
+ "epoch": 1.81,
1441
+ "learning_rate": 7.035347776510833e-06,
1442
+ "logits/chosen": -1.359375,
1443
+ "logits/rejected": -1.2109375,
1444
+ "logps/chosen": -266.0,
1445
+ "logps/rejected": -192.0,
1446
+ "loss": 0.0304,
1447
+ "rewards/accuracies": 0.987500011920929,
1448
+ "rewards/chosen": -1.831870436668396,
1449
+ "rewards/margins": 10.311990737915039,
1450
+ "rewards/rejected": -12.143861770629883,
1451
+ "step": 2040
1452
+ },
1453
+ {
1454
+ "epoch": 1.83,
1455
+ "learning_rate": 6.978335233751426e-06,
1456
+ "logits/chosen": -1.4375,
1457
+ "logits/rejected": -1.296875,
1458
+ "logps/chosen": -256.0,
1459
+ "logps/rejected": -183.0,
1460
+ "loss": 0.0308,
1461
+ "rewards/accuracies": 0.98828125,
1462
+ "rewards/chosen": -1.3896019458770752,
1463
+ "rewards/margins": 10.015677452087402,
1464
+ "rewards/rejected": -11.405279159545898,
1465
+ "step": 2060
1466
+ },
1467
+ {
1468
+ "epoch": 1.84,
1469
+ "learning_rate": 6.921322690992019e-06,
1470
+ "logits/chosen": -1.375,
1471
+ "logits/rejected": -1.2265625,
1472
+ "logps/chosen": -253.0,
1473
+ "logps/rejected": -186.0,
1474
+ "loss": 0.0253,
1475
+ "rewards/accuracies": 0.9921875,
1476
+ "rewards/chosen": -1.5604383945465088,
1477
+ "rewards/margins": 10.14197063446045,
1478
+ "rewards/rejected": -11.702408790588379,
1479
+ "step": 2080
1480
+ },
1481
+ {
1482
+ "epoch": 1.86,
1483
+ "learning_rate": 6.864310148232611e-06,
1484
+ "logits/chosen": -1.3203125,
1485
+ "logits/rejected": -1.1796875,
1486
+ "logps/chosen": -266.0,
1487
+ "logps/rejected": -198.0,
1488
+ "loss": 0.0259,
1489
+ "rewards/accuracies": 0.9898437261581421,
1490
+ "rewards/chosen": -2.0974209308624268,
1491
+ "rewards/margins": 10.54310131072998,
1492
+ "rewards/rejected": -12.640522003173828,
1493
+ "step": 2100
1494
+ },
1495
+ {
1496
+ "epoch": 1.88,
1497
+ "learning_rate": 6.807297605473204e-06,
1498
+ "logits/chosen": -1.28125,
1499
+ "logits/rejected": -1.15625,
1500
+ "logps/chosen": -266.0,
1501
+ "logps/rejected": -208.0,
1502
+ "loss": 0.0348,
1503
+ "rewards/accuracies": 0.9859374761581421,
1504
+ "rewards/chosen": -2.662431240081787,
1505
+ "rewards/margins": 11.303977966308594,
1506
+ "rewards/rejected": -13.966409683227539,
1507
+ "step": 2120
1508
+ },
1509
+ {
1510
+ "epoch": 1.9,
1511
+ "learning_rate": 6.750285062713797e-06,
1512
+ "logits/chosen": -1.3203125,
1513
+ "logits/rejected": -1.1640625,
1514
+ "logps/chosen": -262.0,
1515
+ "logps/rejected": -198.0,
1516
+ "loss": 0.0248,
1517
+ "rewards/accuracies": 0.9937499761581421,
1518
+ "rewards/chosen": -2.2072205543518066,
1519
+ "rewards/margins": 10.692474365234375,
1520
+ "rewards/rejected": -12.899694442749023,
1521
+ "step": 2140
1522
+ },
1523
+ {
1524
+ "epoch": 1.92,
1525
+ "learning_rate": 6.69327251995439e-06,
1526
+ "logits/chosen": -1.2890625,
1527
+ "logits/rejected": -1.140625,
1528
+ "logps/chosen": -256.0,
1529
+ "logps/rejected": -196.0,
1530
+ "loss": 0.0264,
1531
+ "rewards/accuracies": 0.989062488079071,
1532
+ "rewards/chosen": -1.9567207098007202,
1533
+ "rewards/margins": 10.699260711669922,
1534
+ "rewards/rejected": -12.655980110168457,
1535
+ "step": 2160
1536
+ },
1537
+ {
1538
+ "epoch": 1.93,
1539
+ "learning_rate": 6.636259977194983e-06,
1540
+ "logits/chosen": -1.2578125,
1541
+ "logits/rejected": -1.1171875,
1542
+ "logps/chosen": -268.0,
1543
+ "logps/rejected": -195.0,
1544
+ "loss": 0.0269,
1545
+ "rewards/accuracies": 0.9906250238418579,
1546
+ "rewards/chosen": -2.2203476428985596,
1547
+ "rewards/margins": 10.40222454071045,
1548
+ "rewards/rejected": -12.622572898864746,
1549
+ "step": 2180
1550
+ },
1551
+ {
1552
+ "epoch": 1.95,
1553
+ "learning_rate": 6.579247434435576e-06,
1554
+ "logits/chosen": -1.25,
1555
+ "logits/rejected": -1.0546875,
1556
+ "logps/chosen": -260.0,
1557
+ "logps/rejected": -188.0,
1558
+ "loss": 0.0203,
1559
+ "rewards/accuracies": 0.9921875,
1560
+ "rewards/chosen": -1.528204321861267,
1561
+ "rewards/margins": 10.256698608398438,
1562
+ "rewards/rejected": -11.784903526306152,
1563
+ "step": 2200
1564
+ },
1565
+ {
1566
+ "epoch": 1.97,
1567
+ "learning_rate": 6.522234891676169e-06,
1568
+ "logits/chosen": -1.2109375,
1569
+ "logits/rejected": -1.03125,
1570
+ "logps/chosen": -272.0,
1571
+ "logps/rejected": -199.0,
1572
+ "loss": 0.0235,
1573
+ "rewards/accuracies": 0.9921875,
1574
+ "rewards/chosen": -2.464181423187256,
1575
+ "rewards/margins": 10.21046257019043,
1576
+ "rewards/rejected": -12.674644470214844,
1577
+ "step": 2220
1578
+ },
1579
+ {
1580
+ "epoch": 1.99,
1581
+ "learning_rate": 6.465222348916762e-06,
1582
+ "logits/chosen": -1.2265625,
1583
+ "logits/rejected": -1.0546875,
1584
+ "logps/chosen": -264.0,
1585
+ "logps/rejected": -194.0,
1586
+ "loss": 0.0304,
1587
+ "rewards/accuracies": 0.9867187738418579,
1588
+ "rewards/chosen": -2.2391433715820312,
1589
+ "rewards/margins": 10.337366104125977,
1590
+ "rewards/rejected": -12.576510429382324,
1591
+ "step": 2240
1592
+ },
1593
+ {
1594
+ "epoch": 2.0,
1595
+ "learning_rate": 6.408209806157355e-06,
1596
+ "logits/chosen": -1.1796875,
1597
+ "logits/rejected": -0.9453125,
1598
+ "logps/chosen": -274.0,
1599
+ "logps/rejected": -191.0,
1600
+ "loss": 0.023,
1601
+ "rewards/accuracies": 0.991406261920929,
1602
+ "rewards/chosen": -2.1470155715942383,
1603
+ "rewards/margins": 10.05528450012207,
1604
+ "rewards/rejected": -12.202299118041992,
1605
+ "step": 2260
1606
+ },
1607
+ {
1608
+ "epoch": 2.02,
1609
+ "learning_rate": 6.351197263397948e-06,
1610
+ "logits/chosen": -1.171875,
1611
+ "logits/rejected": -0.95703125,
1612
+ "logps/chosen": -264.0,
1613
+ "logps/rejected": -205.0,
1614
+ "loss": 0.0041,
1615
+ "rewards/accuracies": 0.999218761920929,
1616
+ "rewards/chosen": -2.200409173965454,
1617
+ "rewards/margins": 11.429448127746582,
1618
+ "rewards/rejected": -13.629857063293457,
1619
+ "step": 2280
1620
+ },
1621
+ {
1622
+ "epoch": 2.04,
1623
+ "learning_rate": 6.294184720638541e-06,
1624
+ "logits/chosen": -1.203125,
1625
+ "logits/rejected": -0.984375,
1626
+ "logps/chosen": -260.0,
1627
+ "logps/rejected": -203.0,
1628
+ "loss": 0.0057,
1629
+ "rewards/accuracies": 0.999218761920929,
1630
+ "rewards/chosen": -1.8475700616836548,
1631
+ "rewards/margins": 11.384028434753418,
1632
+ "rewards/rejected": -13.231597900390625,
1633
+ "step": 2300
1634
+ },
1635
+ {
1636
+ "epoch": 2.06,
1637
+ "learning_rate": 6.237172177879134e-06,
1638
+ "logits/chosen": -1.1328125,
1639
+ "logits/rejected": -0.8984375,
1640
+ "logps/chosen": -258.0,
1641
+ "logps/rejected": -211.0,
1642
+ "loss": 0.0067,
1643
+ "rewards/accuracies": 0.9976562261581421,
1644
+ "rewards/chosen": -2.3629112243652344,
1645
+ "rewards/margins": 11.951788902282715,
1646
+ "rewards/rejected": -14.314702033996582,
1647
+ "step": 2320
1648
+ },
1649
+ {
1650
+ "epoch": 2.07,
1651
+ "learning_rate": 6.180159635119727e-06,
1652
+ "logits/chosen": -1.125,
1653
+ "logits/rejected": -0.80078125,
1654
+ "logps/chosen": -258.0,
1655
+ "logps/rejected": -219.0,
1656
+ "loss": 0.0048,
1657
+ "rewards/accuracies": 0.9984375238418579,
1658
+ "rewards/chosen": -2.6320786476135254,
1659
+ "rewards/margins": 12.362083435058594,
1660
+ "rewards/rejected": -14.994163513183594,
1661
+ "step": 2340
1662
+ },
1663
+ {
1664
+ "epoch": 2.09,
1665
+ "learning_rate": 6.123147092360319e-06,
1666
+ "logits/chosen": -1.125,
1667
+ "logits/rejected": -0.8125,
1668
+ "logps/chosen": -266.0,
1669
+ "logps/rejected": -218.0,
1670
+ "loss": 0.0044,
1671
+ "rewards/accuracies": 1.0,
1672
+ "rewards/chosen": -2.58856463432312,
1673
+ "rewards/margins": 12.54808521270752,
1674
+ "rewards/rejected": -15.136648178100586,
1675
+ "step": 2360
1676
+ },
1677
+ {
1678
+ "epoch": 2.11,
1679
+ "learning_rate": 6.066134549600912e-06,
1680
+ "logits/chosen": -1.0625,
1681
+ "logits/rejected": -0.71875,
1682
+ "logps/chosen": -270.0,
1683
+ "logps/rejected": -226.0,
1684
+ "loss": 0.0046,
1685
+ "rewards/accuracies": 0.999218761920929,
1686
+ "rewards/chosen": -2.7806572914123535,
1687
+ "rewards/margins": 13.015957832336426,
1688
+ "rewards/rejected": -15.796615600585938,
1689
+ "step": 2380
1690
+ },
1691
+ {
1692
+ "epoch": 2.13,
1693
+ "learning_rate": 6.009122006841505e-06,
1694
+ "logits/chosen": -0.9765625,
1695
+ "logits/rejected": -0.671875,
1696
+ "logps/chosen": -274.0,
1697
+ "logps/rejected": -228.0,
1698
+ "loss": 0.0133,
1699
+ "rewards/accuracies": 0.99609375,
1700
+ "rewards/chosen": -3.3743209838867188,
1701
+ "rewards/margins": 12.621271133422852,
1702
+ "rewards/rejected": -15.995593070983887,
1703
+ "step": 2400
1704
+ },
1705
+ {
1706
+ "epoch": 2.15,
1707
+ "learning_rate": 5.9521094640820984e-06,
1708
+ "logits/chosen": -1.09375,
1709
+ "logits/rejected": -0.671875,
1710
+ "logps/chosen": -274.0,
1711
+ "logps/rejected": -219.0,
1712
+ "loss": 0.0023,
1713
+ "rewards/accuracies": 0.999218761920929,
1714
+ "rewards/chosen": -2.076186418533325,
1715
+ "rewards/margins": 12.830595016479492,
1716
+ "rewards/rejected": -14.906782150268555,
1717
+ "step": 2420
1718
+ },
1719
+ {
1720
+ "epoch": 2.16,
1721
+ "learning_rate": 5.8950969213226914e-06,
1722
+ "logits/chosen": -1.109375,
1723
+ "logits/rejected": -0.6484375,
1724
+ "logps/chosen": -268.0,
1725
+ "logps/rejected": -228.0,
1726
+ "loss": 0.0028,
1727
+ "rewards/accuracies": 0.9984375238418579,
1728
+ "rewards/chosen": -2.168813467025757,
1729
+ "rewards/margins": 13.737543106079102,
1730
+ "rewards/rejected": -15.906353950500488,
1731
+ "step": 2440
1732
+ },
1733
+ {
1734
+ "epoch": 2.18,
1735
+ "learning_rate": 5.8380843785632845e-06,
1736
+ "logits/chosen": -1.109375,
1737
+ "logits/rejected": -0.6640625,
1738
+ "logps/chosen": -266.0,
1739
+ "logps/rejected": -227.0,
1740
+ "loss": 0.005,
1741
+ "rewards/accuracies": 0.999218761920929,
1742
+ "rewards/chosen": -2.49792742729187,
1743
+ "rewards/margins": 13.276643753051758,
1744
+ "rewards/rejected": -15.774572372436523,
1745
+ "step": 2460
1746
+ },
1747
+ {
1748
+ "epoch": 2.2,
1749
+ "learning_rate": 5.7810718358038775e-06,
1750
+ "logits/chosen": -1.125,
1751
+ "logits/rejected": -0.7265625,
1752
+ "logps/chosen": -264.0,
1753
+ "logps/rejected": -224.0,
1754
+ "loss": 0.008,
1755
+ "rewards/accuracies": 0.9984375238418579,
1756
+ "rewards/chosen": -2.374281167984009,
1757
+ "rewards/margins": 13.046087265014648,
1758
+ "rewards/rejected": -15.420369148254395,
1759
+ "step": 2480
1760
+ },
1761
+ {
1762
+ "epoch": 2.22,
1763
+ "learning_rate": 5.7240592930444705e-06,
1764
+ "logits/chosen": -1.0859375,
1765
+ "logits/rejected": -0.6171875,
1766
+ "logps/chosen": -264.0,
1767
+ "logps/rejected": -226.0,
1768
+ "loss": 0.0036,
1769
+ "rewards/accuracies": 0.9984375238418579,
1770
+ "rewards/chosen": -2.243548631668091,
1771
+ "rewards/margins": 13.532651901245117,
1772
+ "rewards/rejected": -15.776202201843262,
1773
+ "step": 2500
1774
+ },
1775
+ {
1776
+ "epoch": 2.23,
1777
+ "learning_rate": 5.6670467502850635e-06,
1778
+ "logits/chosen": -1.109375,
1779
+ "logits/rejected": -0.64453125,
1780
+ "logps/chosen": -280.0,
1781
+ "logps/rejected": -234.0,
1782
+ "loss": 0.004,
1783
+ "rewards/accuracies": 0.999218761920929,
1784
+ "rewards/chosen": -2.683474063873291,
1785
+ "rewards/margins": 13.599886894226074,
1786
+ "rewards/rejected": -16.28335952758789,
1787
+ "step": 2520
1788
+ },
1789
+ {
1790
+ "epoch": 2.25,
1791
+ "learning_rate": 5.6100342075256565e-06,
1792
+ "logits/chosen": -1.1484375,
1793
+ "logits/rejected": -0.71875,
1794
+ "logps/chosen": -272.0,
1795
+ "logps/rejected": -234.0,
1796
+ "loss": 0.0042,
1797
+ "rewards/accuracies": 0.999218761920929,
1798
+ "rewards/chosen": -2.5661137104034424,
1799
+ "rewards/margins": 13.591761589050293,
1800
+ "rewards/rejected": -16.157875061035156,
1801
+ "step": 2540
1802
+ },
1803
+ {
1804
+ "epoch": 2.27,
1805
+ "learning_rate": 5.5530216647662496e-06,
1806
+ "logits/chosen": -1.125,
1807
+ "logits/rejected": -0.63671875,
1808
+ "logps/chosen": -282.0,
1809
+ "logps/rejected": -241.0,
1810
+ "loss": 0.0035,
1811
+ "rewards/accuracies": 0.999218761920929,
1812
+ "rewards/chosen": -2.7359392642974854,
1813
+ "rewards/margins": 14.275691032409668,
1814
+ "rewards/rejected": -17.01163101196289,
1815
+ "step": 2560
1816
+ },
1817
+ {
1818
+ "epoch": 2.29,
1819
+ "learning_rate": 5.496009122006843e-06,
1820
+ "logits/chosen": -1.15625,
1821
+ "logits/rejected": -0.625,
1822
+ "logps/chosen": -280.0,
1823
+ "logps/rejected": -241.0,
1824
+ "loss": 0.0057,
1825
+ "rewards/accuracies": 0.9976562261581421,
1826
+ "rewards/chosen": -2.646916389465332,
1827
+ "rewards/margins": 14.39399242401123,
1828
+ "rewards/rejected": -17.040910720825195,
1829
+ "step": 2580
1830
+ },
1831
+ {
1832
+ "epoch": 2.31,
1833
+ "learning_rate": 5.438996579247436e-06,
1834
+ "logits/chosen": -1.171875,
1835
+ "logits/rejected": -0.71484375,
1836
+ "logps/chosen": -266.0,
1837
+ "logps/rejected": -242.0,
1838
+ "loss": 0.0073,
1839
+ "rewards/accuracies": 0.9984375238418579,
1840
+ "rewards/chosen": -2.6745243072509766,
1841
+ "rewards/margins": 14.485051155090332,
1842
+ "rewards/rejected": -17.159576416015625,
1843
+ "step": 2600
1844
+ },
1845
+ {
1846
+ "epoch": 2.32,
1847
+ "learning_rate": 5.381984036488029e-06,
1848
+ "logits/chosen": -1.125,
1849
+ "logits/rejected": -0.58203125,
1850
+ "logps/chosen": -276.0,
1851
+ "logps/rejected": -234.0,
1852
+ "loss": 0.0016,
1853
+ "rewards/accuracies": 1.0,
1854
+ "rewards/chosen": -2.587552547454834,
1855
+ "rewards/margins": 13.87269115447998,
1856
+ "rewards/rejected": -16.46024513244629,
1857
+ "step": 2620
1858
+ },
1859
+ {
1860
+ "epoch": 2.34,
1861
+ "learning_rate": 5.324971493728621e-06,
1862
+ "logits/chosen": -1.109375,
1863
+ "logits/rejected": -0.53515625,
1864
+ "logps/chosen": -268.0,
1865
+ "logps/rejected": -241.0,
1866
+ "loss": 0.0027,
1867
+ "rewards/accuracies": 1.0,
1868
+ "rewards/chosen": -2.578281879425049,
1869
+ "rewards/margins": 14.362253189086914,
1870
+ "rewards/rejected": -16.940534591674805,
1871
+ "step": 2640
1872
+ },
1873
+ {
1874
+ "epoch": 2.36,
1875
+ "learning_rate": 5.267958950969214e-06,
1876
+ "logits/chosen": -1.0546875,
1877
+ "logits/rejected": -0.44140625,
1878
+ "logps/chosen": -282.0,
1879
+ "logps/rejected": -262.0,
1880
+ "loss": 0.0024,
1881
+ "rewards/accuracies": 1.0,
1882
+ "rewards/chosen": -3.1921732425689697,
1883
+ "rewards/margins": 15.826263427734375,
1884
+ "rewards/rejected": -19.018436431884766,
1885
+ "step": 2660
1886
+ },
1887
+ {
1888
+ "epoch": 2.38,
1889
+ "learning_rate": 5.210946408209807e-06,
1890
+ "logits/chosen": -1.078125,
1891
+ "logits/rejected": -0.455078125,
1892
+ "logps/chosen": -284.0,
1893
+ "logps/rejected": -266.0,
1894
+ "loss": 0.0031,
1895
+ "rewards/accuracies": 0.999218761920929,
1896
+ "rewards/chosen": -3.5577616691589355,
1897
+ "rewards/margins": 15.753667831420898,
1898
+ "rewards/rejected": -19.311429977416992,
1899
+ "step": 2680
1900
+ },
1901
+ {
1902
+ "epoch": 2.39,
1903
+ "learning_rate": 5.153933865450399e-06,
1904
+ "logits/chosen": -1.140625,
1905
+ "logits/rejected": -0.58984375,
1906
+ "logps/chosen": -280.0,
1907
+ "logps/rejected": -242.0,
1908
+ "loss": 0.0051,
1909
+ "rewards/accuracies": 0.999218761920929,
1910
+ "rewards/chosen": -2.8636865615844727,
1911
+ "rewards/margins": 14.327638626098633,
1912
+ "rewards/rejected": -17.191326141357422,
1913
+ "step": 2700
1914
+ },
1915
+ {
1916
+ "epoch": 2.41,
1917
+ "learning_rate": 5.096921322690992e-06,
1918
+ "logits/chosen": -1.0390625,
1919
+ "logits/rejected": -0.498046875,
1920
+ "logps/chosen": -274.0,
1921
+ "logps/rejected": -246.0,
1922
+ "loss": 0.0028,
1923
+ "rewards/accuracies": 1.0,
1924
+ "rewards/chosen": -2.9436042308807373,
1925
+ "rewards/margins": 14.590520858764648,
1926
+ "rewards/rejected": -17.53412437438965,
1927
+ "step": 2720
1928
+ },
1929
+ {
1930
+ "epoch": 2.43,
1931
+ "learning_rate": 5.039908779931585e-06,
1932
+ "logits/chosen": -1.03125,
1933
+ "logits/rejected": -0.4453125,
1934
+ "logps/chosen": -284.0,
1935
+ "logps/rejected": -260.0,
1936
+ "loss": 0.0021,
1937
+ "rewards/accuracies": 1.0,
1938
+ "rewards/chosen": -3.5330169200897217,
1939
+ "rewards/margins": 15.42271614074707,
1940
+ "rewards/rejected": -18.955734252929688,
1941
+ "step": 2740
1942
+ },
1943
+ {
1944
+ "epoch": 2.45,
1945
+ "learning_rate": 4.982896237172178e-06,
1946
+ "logits/chosen": -1.0390625,
1947
+ "logits/rejected": -0.4921875,
1948
+ "logps/chosen": -266.0,
1949
+ "logps/rejected": -251.0,
1950
+ "loss": 0.0156,
1951
+ "rewards/accuracies": 0.9937499761581421,
1952
+ "rewards/chosen": -3.311683177947998,
1953
+ "rewards/margins": 14.74211597442627,
1954
+ "rewards/rejected": -18.05379867553711,
1955
+ "step": 2760
1956
+ },
1957
+ {
1958
+ "epoch": 2.46,
1959
+ "learning_rate": 4.925883694412771e-06,
1960
+ "logits/chosen": -1.0234375,
1961
+ "logits/rejected": -0.490234375,
1962
+ "logps/chosen": -272.0,
1963
+ "logps/rejected": -258.0,
1964
+ "loss": 0.0058,
1965
+ "rewards/accuracies": 0.9976562261581421,
1966
+ "rewards/chosen": -3.4664859771728516,
1967
+ "rewards/margins": 15.337985038757324,
1968
+ "rewards/rejected": -18.804471969604492,
1969
+ "step": 2780
1970
+ },
1971
+ {
1972
+ "epoch": 2.48,
1973
+ "learning_rate": 4.868871151653364e-06,
1974
+ "logits/chosen": -1.140625,
1975
+ "logits/rejected": -0.63671875,
1976
+ "logps/chosen": -272.0,
1977
+ "logps/rejected": -252.0,
1978
+ "loss": 0.0041,
1979
+ "rewards/accuracies": 0.9984375238418579,
1980
+ "rewards/chosen": -3.0301337242126465,
1981
+ "rewards/margins": 15.127180099487305,
1982
+ "rewards/rejected": -18.15731430053711,
1983
+ "step": 2800
1984
+ },
1985
+ {
1986
+ "epoch": 2.5,
1987
+ "learning_rate": 4.811858608893957e-06,
1988
+ "logits/chosen": -1.1640625,
1989
+ "logits/rejected": -0.6875,
1990
+ "logps/chosen": -264.0,
1991
+ "logps/rejected": -238.0,
1992
+ "loss": 0.0112,
1993
+ "rewards/accuracies": 0.9976562261581421,
1994
+ "rewards/chosen": -2.7115814685821533,
1995
+ "rewards/margins": 14.367854118347168,
1996
+ "rewards/rejected": -17.079435348510742,
1997
+ "step": 2820
1998
+ },
1999
+ {
2000
+ "epoch": 2.52,
2001
+ "learning_rate": 4.75484606613455e-06,
2002
+ "logits/chosen": -1.203125,
2003
+ "logits/rejected": -0.7265625,
2004
+ "logps/chosen": -260.0,
2005
+ "logps/rejected": -238.0,
2006
+ "loss": 0.0034,
2007
+ "rewards/accuracies": 0.9984375238418579,
2008
+ "rewards/chosen": -2.354907274246216,
2009
+ "rewards/margins": 14.6588773727417,
2010
+ "rewards/rejected": -17.013784408569336,
2011
+ "step": 2840
2012
+ },
2013
+ {
2014
+ "epoch": 2.54,
2015
+ "learning_rate": 4.697833523375142e-06,
2016
+ "logits/chosen": -1.1796875,
2017
+ "logits/rejected": -0.69921875,
2018
+ "logps/chosen": -270.0,
2019
+ "logps/rejected": -246.0,
2020
+ "loss": 0.0047,
2021
+ "rewards/accuracies": 0.999218761920929,
2022
+ "rewards/chosen": -2.840439796447754,
2023
+ "rewards/margins": 14.848834037780762,
2024
+ "rewards/rejected": -17.68927574157715,
2025
+ "step": 2860
2026
+ },
2027
+ {
2028
+ "epoch": 2.55,
2029
+ "learning_rate": 4.640820980615735e-06,
2030
+ "logits/chosen": -1.1015625,
2031
+ "logits/rejected": -0.60546875,
2032
+ "logps/chosen": -274.0,
2033
+ "logps/rejected": -262.0,
2034
+ "loss": 0.0029,
2035
+ "rewards/accuracies": 0.999218761920929,
2036
+ "rewards/chosen": -3.4766571521759033,
2037
+ "rewards/margins": 15.726774215698242,
2038
+ "rewards/rejected": -19.203433990478516,
2039
+ "step": 2880
2040
+ },
2041
+ {
2042
+ "epoch": 2.57,
2043
+ "learning_rate": 4.583808437856328e-06,
2044
+ "logits/chosen": -1.125,
2045
+ "logits/rejected": -0.58984375,
2046
+ "logps/chosen": -280.0,
2047
+ "logps/rejected": -262.0,
2048
+ "loss": 0.0043,
2049
+ "rewards/accuracies": 0.9984375238418579,
2050
+ "rewards/chosen": -3.400480270385742,
2051
+ "rewards/margins": 15.654681205749512,
2052
+ "rewards/rejected": -19.05516242980957,
2053
+ "step": 2900
2054
+ },
2055
+ {
2056
+ "epoch": 2.59,
2057
+ "learning_rate": 4.526795895096921e-06,
2058
+ "logits/chosen": -1.0546875,
2059
+ "logits/rejected": -0.5625,
2060
+ "logps/chosen": -282.0,
2061
+ "logps/rejected": -253.0,
2062
+ "loss": 0.003,
2063
+ "rewards/accuracies": 1.0,
2064
+ "rewards/chosen": -3.4995570182800293,
2065
+ "rewards/margins": 14.704421997070312,
2066
+ "rewards/rejected": -18.2039794921875,
2067
+ "step": 2920
2068
+ },
2069
+ {
2070
+ "epoch": 2.61,
2071
+ "learning_rate": 4.469783352337514e-06,
2072
+ "logits/chosen": -1.046875,
2073
+ "logits/rejected": -0.5234375,
2074
+ "logps/chosen": -276.0,
2075
+ "logps/rejected": -252.0,
2076
+ "loss": 0.0025,
2077
+ "rewards/accuracies": 1.0,
2078
+ "rewards/chosen": -3.058525323867798,
2079
+ "rewards/margins": 15.031721115112305,
2080
+ "rewards/rejected": -18.090246200561523,
2081
+ "step": 2940
2082
+ },
2083
+ {
2084
+ "epoch": 2.62,
2085
+ "learning_rate": 4.412770809578107e-06,
2086
+ "logits/chosen": -1.03125,
2087
+ "logits/rejected": -0.46875,
2088
+ "logps/chosen": -272.0,
2089
+ "logps/rejected": -254.0,
2090
+ "loss": 0.0035,
2091
+ "rewards/accuracies": 0.999218761920929,
2092
+ "rewards/chosen": -2.96889066696167,
2093
+ "rewards/margins": 15.424901962280273,
2094
+ "rewards/rejected": -18.393795013427734,
2095
+ "step": 2960
2096
+ },
2097
+ {
2098
+ "epoch": 2.64,
2099
+ "learning_rate": 4.3557582668187004e-06,
2100
+ "logits/chosen": -1.0,
2101
+ "logits/rejected": -0.47265625,
2102
+ "logps/chosen": -268.0,
2103
+ "logps/rejected": -254.0,
2104
+ "loss": 0.0104,
2105
+ "rewards/accuracies": 0.996874988079071,
2106
+ "rewards/chosen": -3.0877561569213867,
2107
+ "rewards/margins": 15.393818855285645,
2108
+ "rewards/rejected": -18.4815731048584,
2109
+ "step": 2980
2110
+ },
2111
+ {
2112
+ "epoch": 2.66,
2113
+ "learning_rate": 4.2987457240592934e-06,
2114
+ "logits/chosen": -0.99609375,
2115
+ "logits/rejected": -0.490234375,
2116
+ "logps/chosen": -266.0,
2117
+ "logps/rejected": -251.0,
2118
+ "loss": 0.0097,
2119
+ "rewards/accuracies": 0.999218761920929,
2120
+ "rewards/chosen": -2.89310884475708,
2121
+ "rewards/margins": 15.272462844848633,
2122
+ "rewards/rejected": -18.165569305419922,
2123
+ "step": 3000
2124
+ },
2125
+ {
2126
+ "epoch": 2.68,
2127
+ "learning_rate": 4.2417331812998865e-06,
2128
+ "logits/chosen": -1.0234375,
2129
+ "logits/rejected": -0.45703125,
2130
+ "logps/chosen": -270.0,
2131
+ "logps/rejected": -253.0,
2132
+ "loss": 0.0044,
2133
+ "rewards/accuracies": 1.0,
2134
+ "rewards/chosen": -2.787309169769287,
2135
+ "rewards/margins": 15.575048446655273,
2136
+ "rewards/rejected": -18.362356185913086,
2137
+ "step": 3020
2138
+ },
2139
+ {
2140
+ "epoch": 2.7,
2141
+ "learning_rate": 4.1847206385404795e-06,
2142
+ "logits/chosen": -1.0078125,
2143
+ "logits/rejected": -0.458984375,
2144
+ "logps/chosen": -268.0,
2145
+ "logps/rejected": -244.0,
2146
+ "loss": 0.0127,
2147
+ "rewards/accuracies": 0.99609375,
2148
+ "rewards/chosen": -2.9177849292755127,
2149
+ "rewards/margins": 14.414236068725586,
2150
+ "rewards/rejected": -17.33202362060547,
2151
+ "step": 3040
2152
+ },
2153
+ {
2154
+ "epoch": 2.71,
2155
+ "learning_rate": 4.1277080957810725e-06,
2156
+ "logits/chosen": -1.0234375,
2157
+ "logits/rejected": -0.462890625,
2158
+ "logps/chosen": -276.0,
2159
+ "logps/rejected": -247.0,
2160
+ "loss": 0.0024,
2161
+ "rewards/accuracies": 1.0,
2162
+ "rewards/chosen": -2.717505931854248,
2163
+ "rewards/margins": 14.872530937194824,
2164
+ "rewards/rejected": -17.590036392211914,
2165
+ "step": 3060
2166
+ },
2167
+ {
2168
+ "epoch": 2.73,
2169
+ "learning_rate": 4.0706955530216655e-06,
2170
+ "logits/chosen": -0.9921875,
2171
+ "logits/rejected": -0.4140625,
2172
+ "logps/chosen": -274.0,
2173
+ "logps/rejected": -248.0,
2174
+ "loss": 0.0038,
2175
+ "rewards/accuracies": 0.9984375238418579,
2176
+ "rewards/chosen": -2.782890796661377,
2177
+ "rewards/margins": 15.000226020812988,
2178
+ "rewards/rejected": -17.78311538696289,
2179
+ "step": 3080
2180
+ },
2181
+ {
2182
+ "epoch": 2.75,
2183
+ "learning_rate": 4.0136830102622585e-06,
2184
+ "logits/chosen": -0.9296875,
2185
+ "logits/rejected": -0.357421875,
2186
+ "logps/chosen": -278.0,
2187
+ "logps/rejected": -262.0,
2188
+ "loss": 0.0033,
2189
+ "rewards/accuracies": 0.9984375238418579,
2190
+ "rewards/chosen": -3.5591626167297363,
2191
+ "rewards/margins": 15.839828491210938,
2192
+ "rewards/rejected": -19.398990631103516,
2193
+ "step": 3100
2194
+ },
2195
+ {
2196
+ "epoch": 2.77,
2197
+ "learning_rate": 3.956670467502851e-06,
2198
+ "logits/chosen": -0.98046875,
2199
+ "logits/rejected": -0.384765625,
2200
+ "logps/chosen": -280.0,
2201
+ "logps/rejected": -266.0,
2202
+ "loss": 0.0036,
2203
+ "rewards/accuracies": 0.999218761920929,
2204
+ "rewards/chosen": -3.5741126537323,
2205
+ "rewards/margins": 15.891105651855469,
2206
+ "rewards/rejected": -19.465219497680664,
2207
+ "step": 3120
2208
+ },
2209
+ {
2210
+ "epoch": 2.78,
2211
+ "learning_rate": 3.899657924743444e-06,
2212
+ "logits/chosen": -1.046875,
2213
+ "logits/rejected": -0.462890625,
2214
+ "logps/chosen": -274.0,
2215
+ "logps/rejected": -260.0,
2216
+ "loss": 0.0023,
2217
+ "rewards/accuracies": 1.0,
2218
+ "rewards/chosen": -3.375121593475342,
2219
+ "rewards/margins": 15.77069091796875,
2220
+ "rewards/rejected": -19.145811080932617,
2221
+ "step": 3140
2222
+ },
2223
+ {
2224
+ "epoch": 2.8,
2225
+ "learning_rate": 3.842645381984037e-06,
2226
+ "logits/chosen": -1.0390625,
2227
+ "logits/rejected": -0.474609375,
2228
+ "logps/chosen": -276.0,
2229
+ "logps/rejected": -258.0,
2230
+ "loss": 0.0031,
2231
+ "rewards/accuracies": 0.999218761920929,
2232
+ "rewards/chosen": -3.2375900745391846,
2233
+ "rewards/margins": 15.49463176727295,
2234
+ "rewards/rejected": -18.732221603393555,
2235
+ "step": 3160
2236
+ },
2237
+ {
2238
+ "epoch": 2.82,
2239
+ "learning_rate": 3.7856328392246298e-06,
2240
+ "logits/chosen": -1.046875,
2241
+ "logits/rejected": -0.46484375,
2242
+ "logps/chosen": -274.0,
2243
+ "logps/rejected": -247.0,
2244
+ "loss": 0.0109,
2245
+ "rewards/accuracies": 0.9976562261581421,
2246
+ "rewards/chosen": -2.936633348464966,
2247
+ "rewards/margins": 15.029220581054688,
2248
+ "rewards/rejected": -17.96585464477539,
2249
+ "step": 3180
2250
+ },
2251
+ {
2252
+ "epoch": 2.84,
2253
+ "learning_rate": 3.7286202964652224e-06,
2254
+ "logits/chosen": -0.984375,
2255
+ "logits/rejected": -0.44921875,
2256
+ "logps/chosen": -288.0,
2257
+ "logps/rejected": -262.0,
2258
+ "loss": 0.0026,
2259
+ "rewards/accuracies": 0.999218761920929,
2260
+ "rewards/chosen": -3.5291266441345215,
2261
+ "rewards/margins": 15.453519821166992,
2262
+ "rewards/rejected": -18.982646942138672,
2263
+ "step": 3200
2264
+ },
2265
+ {
2266
+ "epoch": 2.85,
2267
+ "learning_rate": 3.6716077537058154e-06,
2268
+ "logits/chosen": -0.921875,
2269
+ "logits/rejected": -0.375,
2270
+ "logps/chosen": -278.0,
2271
+ "logps/rejected": -270.0,
2272
+ "loss": 0.0024,
2273
+ "rewards/accuracies": 1.0,
2274
+ "rewards/chosen": -4.0102667808532715,
2275
+ "rewards/margins": 16.090139389038086,
2276
+ "rewards/rejected": -20.100406646728516,
2277
+ "step": 3220
2278
+ },
2279
+ {
2280
+ "epoch": 2.87,
2281
+ "learning_rate": 3.6145952109464084e-06,
2282
+ "logits/chosen": -1.015625,
2283
+ "logits/rejected": -0.419921875,
2284
+ "logps/chosen": -284.0,
2285
+ "logps/rejected": -264.0,
2286
+ "loss": 0.0039,
2287
+ "rewards/accuracies": 0.999218761920929,
2288
+ "rewards/chosen": -3.3913803100585938,
2289
+ "rewards/margins": 15.949442863464355,
2290
+ "rewards/rejected": -19.340822219848633,
2291
+ "step": 3240
2292
+ },
2293
+ {
2294
+ "epoch": 2.89,
2295
+ "learning_rate": 3.5575826681870014e-06,
2296
+ "logits/chosen": -0.9921875,
2297
+ "logits/rejected": -0.3671875,
2298
+ "logps/chosen": -274.0,
2299
+ "logps/rejected": -264.0,
2300
+ "loss": 0.0062,
2301
+ "rewards/accuracies": 0.999218761920929,
2302
+ "rewards/chosen": -3.4078826904296875,
2303
+ "rewards/margins": 16.163570404052734,
2304
+ "rewards/rejected": -19.571453094482422,
2305
+ "step": 3260
2306
+ },
2307
+ {
2308
+ "epoch": 2.91,
2309
+ "learning_rate": 3.500570125427594e-06,
2310
+ "logits/chosen": -0.91796875,
2311
+ "logits/rejected": -0.318359375,
2312
+ "logps/chosen": -280.0,
2313
+ "logps/rejected": -270.0,
2314
+ "loss": 0.0049,
2315
+ "rewards/accuracies": 0.999218761920929,
2316
+ "rewards/chosen": -3.7832558155059814,
2317
+ "rewards/margins": 16.340469360351562,
2318
+ "rewards/rejected": -20.12372398376465,
2319
+ "step": 3280
2320
+ },
2321
+ {
2322
+ "epoch": 2.93,
2323
+ "learning_rate": 3.443557582668187e-06,
2324
+ "logits/chosen": -0.99609375,
2325
+ "logits/rejected": -0.404296875,
2326
+ "logps/chosen": -272.0,
2327
+ "logps/rejected": -264.0,
2328
+ "loss": 0.0009,
2329
+ "rewards/accuracies": 1.0,
2330
+ "rewards/chosen": -3.330648422241211,
2331
+ "rewards/margins": 15.921122550964355,
2332
+ "rewards/rejected": -19.25177001953125,
2333
+ "step": 3300
2334
+ },
2335
+ {
2336
+ "epoch": 2.94,
2337
+ "learning_rate": 3.38654503990878e-06,
2338
+ "logits/chosen": -0.97265625,
2339
+ "logits/rejected": -0.34375,
2340
+ "logps/chosen": -292.0,
2341
+ "logps/rejected": -270.0,
2342
+ "loss": 0.0027,
2343
+ "rewards/accuracies": 0.9984375238418579,
2344
+ "rewards/chosen": -3.5318961143493652,
2345
+ "rewards/margins": 16.236053466796875,
2346
+ "rewards/rejected": -19.767948150634766,
2347
+ "step": 3320
2348
+ },
2349
+ {
2350
+ "epoch": 2.96,
2351
+ "learning_rate": 3.329532497149373e-06,
2352
+ "logits/chosen": -1.015625,
2353
+ "logits/rejected": -0.396484375,
2354
+ "logps/chosen": -284.0,
2355
+ "logps/rejected": -264.0,
2356
+ "loss": 0.0035,
2357
+ "rewards/accuracies": 0.999218761920929,
2358
+ "rewards/chosen": -3.347155809402466,
2359
+ "rewards/margins": 16.07892608642578,
2360
+ "rewards/rejected": -19.426082611083984,
2361
+ "step": 3340
2362
+ },
2363
+ {
2364
+ "epoch": 2.98,
2365
+ "learning_rate": 3.272519954389966e-06,
2366
+ "logits/chosen": -0.99609375,
2367
+ "logits/rejected": -0.359375,
2368
+ "logps/chosen": -284.0,
2369
+ "logps/rejected": -268.0,
2370
+ "loss": 0.0029,
2371
+ "rewards/accuracies": 0.999218761920929,
2372
+ "rewards/chosen": -3.578090190887451,
2373
+ "rewards/margins": 16.344289779663086,
2374
+ "rewards/rejected": -19.922382354736328,
2375
+ "step": 3360
2376
+ },
2377
+ {
2378
+ "epoch": 3.0,
2379
+ "learning_rate": 3.215507411630559e-06,
2380
+ "logits/chosen": -0.94921875,
2381
+ "logits/rejected": -0.349609375,
2382
+ "logps/chosen": -276.0,
2383
+ "logps/rejected": -274.0,
2384
+ "loss": 0.0065,
2385
+ "rewards/accuracies": 0.999218761920929,
2386
+ "rewards/chosen": -3.7228190898895264,
2387
+ "rewards/margins": 16.553964614868164,
2388
+ "rewards/rejected": -20.276782989501953,
2389
+ "step": 3380
2390
+ },
2391
+ {
2392
+ "epoch": 3.01,
2393
+ "learning_rate": 3.158494868871152e-06,
2394
+ "logits/chosen": -0.99609375,
2395
+ "logits/rejected": -0.375,
2396
+ "logps/chosen": -270.0,
2397
+ "logps/rejected": -264.0,
2398
+ "loss": 0.0023,
2399
+ "rewards/accuracies": 1.0,
2400
+ "rewards/chosen": -3.2026283740997314,
2401
+ "rewards/margins": 16.301708221435547,
2402
+ "rewards/rejected": -19.504337310791016,
2403
+ "step": 3400
2404
+ },
2405
+ {
2406
+ "epoch": 3.03,
2407
+ "learning_rate": 3.1014823261117447e-06,
2408
+ "logits/chosen": -0.96484375,
2409
+ "logits/rejected": -0.318359375,
2410
+ "logps/chosen": -286.0,
2411
+ "logps/rejected": -280.0,
2412
+ "loss": 0.0006,
2413
+ "rewards/accuracies": 1.0,
2414
+ "rewards/chosen": -3.813472270965576,
2415
+ "rewards/margins": 17.21391487121582,
2416
+ "rewards/rejected": -21.027387619018555,
2417
+ "step": 3420
2418
+ },
2419
+ {
2420
+ "epoch": 3.05,
2421
+ "learning_rate": 3.0444697833523378e-06,
2422
+ "logits/chosen": -0.97265625,
2423
+ "logits/rejected": -0.29296875,
2424
+ "logps/chosen": -278.0,
2425
+ "logps/rejected": -286.0,
2426
+ "loss": 0.002,
2427
+ "rewards/accuracies": 0.9984375238418579,
2428
+ "rewards/chosen": -3.973036527633667,
2429
+ "rewards/margins": 17.74779510498047,
2430
+ "rewards/rejected": -21.7208309173584,
2431
+ "step": 3440
2432
+ },
2433
+ {
2434
+ "epoch": 3.07,
2435
+ "learning_rate": 2.9874572405929308e-06,
2436
+ "logits/chosen": -0.96875,
2437
+ "logits/rejected": -0.29296875,
2438
+ "logps/chosen": -276.0,
2439
+ "logps/rejected": -282.0,
2440
+ "loss": 0.0006,
2441
+ "rewards/accuracies": 1.0,
2442
+ "rewards/chosen": -4.1294169425964355,
2443
+ "rewards/margins": 17.12251853942871,
2444
+ "rewards/rejected": -21.251937866210938,
2445
+ "step": 3460
2446
+ },
2447
+ {
2448
+ "epoch": 3.09,
2449
+ "learning_rate": 2.930444697833524e-06,
2450
+ "logits/chosen": -1.03125,
2451
+ "logits/rejected": -0.34375,
2452
+ "logps/chosen": -290.0,
2453
+ "logps/rejected": -294.0,
2454
+ "loss": 0.0003,
2455
+ "rewards/accuracies": 1.0,
2456
+ "rewards/chosen": -3.9173831939697266,
2457
+ "rewards/margins": 18.43880844116211,
2458
+ "rewards/rejected": -22.356189727783203,
2459
+ "step": 3480
2460
+ },
2461
+ {
2462
+ "epoch": 3.1,
2463
+ "learning_rate": 2.873432155074117e-06,
2464
+ "logits/chosen": -1.0625,
2465
+ "logits/rejected": -0.361328125,
2466
+ "logps/chosen": -282.0,
2467
+ "logps/rejected": -286.0,
2468
+ "loss": 0.0007,
2469
+ "rewards/accuracies": 1.0,
2470
+ "rewards/chosen": -3.7644851207733154,
2471
+ "rewards/margins": 17.8516845703125,
2472
+ "rewards/rejected": -21.61617088317871,
2473
+ "step": 3500
2474
+ },
2475
+ {
2476
+ "epoch": 3.12,
2477
+ "learning_rate": 2.81641961231471e-06,
2478
+ "logits/chosen": -0.9921875,
2479
+ "logits/rejected": -0.32421875,
2480
+ "logps/chosen": -286.0,
2481
+ "logps/rejected": -282.0,
2482
+ "loss": 0.0013,
2483
+ "rewards/accuracies": 1.0,
2484
+ "rewards/chosen": -3.8479247093200684,
2485
+ "rewards/margins": 17.440242767333984,
2486
+ "rewards/rejected": -21.28816795349121,
2487
+ "step": 3520
2488
+ },
2489
+ {
2490
+ "epoch": 3.14,
2491
+ "learning_rate": 2.759407069555302e-06,
2492
+ "logits/chosen": -1.0390625,
2493
+ "logits/rejected": -0.404296875,
2494
+ "logps/chosen": -274.0,
2495
+ "logps/rejected": -282.0,
2496
+ "loss": 0.0009,
2497
+ "rewards/accuracies": 1.0,
2498
+ "rewards/chosen": -3.8636913299560547,
2499
+ "rewards/margins": 17.501983642578125,
2500
+ "rewards/rejected": -21.365676879882812,
2501
+ "step": 3540
2502
+ },
2503
+ {
2504
+ "epoch": 3.16,
2505
+ "learning_rate": 2.702394526795895e-06,
2506
+ "logits/chosen": -1.0859375,
2507
+ "logits/rejected": -0.4375,
2508
+ "logps/chosen": -280.0,
2509
+ "logps/rejected": -280.0,
2510
+ "loss": 0.0005,
2511
+ "rewards/accuracies": 1.0,
2512
+ "rewards/chosen": -3.7761924266815186,
2513
+ "rewards/margins": 17.286312103271484,
2514
+ "rewards/rejected": -21.062503814697266,
2515
+ "step": 3560
2516
+ },
2517
+ {
2518
+ "epoch": 3.17,
2519
+ "learning_rate": 2.645381984036488e-06,
2520
+ "logits/chosen": -1.0546875,
2521
+ "logits/rejected": -0.408203125,
2522
+ "logps/chosen": -280.0,
2523
+ "logps/rejected": -280.0,
2524
+ "loss": 0.0005,
2525
+ "rewards/accuracies": 1.0,
2526
+ "rewards/chosen": -3.7179946899414062,
2527
+ "rewards/margins": 17.326318740844727,
2528
+ "rewards/rejected": -21.0443115234375,
2529
+ "step": 3580
2530
+ },
2531
+ {
2532
+ "epoch": 3.19,
2533
+ "learning_rate": 2.588369441277081e-06,
2534
+ "logits/chosen": -1.1015625,
2535
+ "logits/rejected": -0.453125,
2536
+ "logps/chosen": -284.0,
2537
+ "logps/rejected": -278.0,
2538
+ "loss": 0.0009,
2539
+ "rewards/accuracies": 1.0,
2540
+ "rewards/chosen": -3.4939072132110596,
2541
+ "rewards/margins": 17.304866790771484,
2542
+ "rewards/rejected": -20.79877281188965,
2543
+ "step": 3600
2544
+ },
2545
+ {
2546
+ "epoch": 3.21,
2547
+ "learning_rate": 2.531356898517674e-06,
2548
+ "logits/chosen": -1.078125,
2549
+ "logits/rejected": -0.41015625,
2550
+ "logps/chosen": -278.0,
2551
+ "logps/rejected": -284.0,
2552
+ "loss": 0.0003,
2553
+ "rewards/accuracies": 1.0,
2554
+ "rewards/chosen": -3.338261842727661,
2555
+ "rewards/margins": 17.951480865478516,
2556
+ "rewards/rejected": -21.289743423461914,
2557
+ "step": 3620
2558
+ },
2559
+ {
2560
+ "epoch": 3.23,
2561
+ "learning_rate": 2.474344355758267e-06,
2562
+ "logits/chosen": -1.0234375,
2563
+ "logits/rejected": -0.3671875,
2564
+ "logps/chosen": -284.0,
2565
+ "logps/rejected": -290.0,
2566
+ "loss": 0.0004,
2567
+ "rewards/accuracies": 1.0,
2568
+ "rewards/chosen": -4.018970012664795,
2569
+ "rewards/margins": 17.885696411132812,
2570
+ "rewards/rejected": -21.904666900634766,
2571
+ "step": 3640
2572
+ },
2573
+ {
2574
+ "epoch": 3.25,
2575
+ "learning_rate": 2.4173318129988597e-06,
2576
+ "logits/chosen": -1.09375,
2577
+ "logits/rejected": -0.37890625,
2578
+ "logps/chosen": -278.0,
2579
+ "logps/rejected": -284.0,
2580
+ "loss": 0.0012,
2581
+ "rewards/accuracies": 0.999218761920929,
2582
+ "rewards/chosen": -3.494691848754883,
2583
+ "rewards/margins": 17.86090850830078,
2584
+ "rewards/rejected": -21.355600357055664,
2585
+ "step": 3660
2586
+ },
2587
+ {
2588
+ "epoch": 3.26,
2589
+ "learning_rate": 2.3603192702394527e-06,
2590
+ "logits/chosen": -1.0546875,
2591
+ "logits/rejected": -0.34375,
2592
+ "logps/chosen": -288.0,
2593
+ "logps/rejected": -288.0,
2594
+ "loss": 0.0004,
2595
+ "rewards/accuracies": 1.0,
2596
+ "rewards/chosen": -3.877596378326416,
2597
+ "rewards/margins": 17.996723175048828,
2598
+ "rewards/rejected": -21.874319076538086,
2599
+ "step": 3680
2600
+ },
2601
+ {
2602
+ "epoch": 3.28,
2603
+ "learning_rate": 2.3033067274800457e-06,
2604
+ "logits/chosen": -1.0859375,
2605
+ "logits/rejected": -0.419921875,
2606
+ "logps/chosen": -270.0,
2607
+ "logps/rejected": -282.0,
2608
+ "loss": 0.0004,
2609
+ "rewards/accuracies": 1.0,
2610
+ "rewards/chosen": -3.740429401397705,
2611
+ "rewards/margins": 17.418020248413086,
2612
+ "rewards/rejected": -21.158451080322266,
2613
+ "step": 3700
2614
+ },
2615
+ {
2616
+ "epoch": 3.3,
2617
+ "learning_rate": 2.2462941847206388e-06,
2618
+ "logits/chosen": -1.0625,
2619
+ "logits/rejected": -0.38671875,
2620
+ "logps/chosen": -280.0,
2621
+ "logps/rejected": -298.0,
2622
+ "loss": 0.0005,
2623
+ "rewards/accuracies": 1.0,
2624
+ "rewards/chosen": -3.9601192474365234,
2625
+ "rewards/margins": 18.503660202026367,
2626
+ "rewards/rejected": -22.46377944946289,
2627
+ "step": 3720
2628
+ },
2629
+ {
2630
+ "epoch": 3.32,
2631
+ "learning_rate": 2.1892816419612318e-06,
2632
+ "logits/chosen": -1.0859375,
2633
+ "logits/rejected": -0.3828125,
2634
+ "logps/chosen": -280.0,
2635
+ "logps/rejected": -290.0,
2636
+ "loss": 0.0002,
2637
+ "rewards/accuracies": 1.0,
2638
+ "rewards/chosen": -3.7733867168426514,
2639
+ "rewards/margins": 18.210254669189453,
2640
+ "rewards/rejected": -21.983644485473633,
2641
+ "step": 3740
2642
+ },
2643
+ {
2644
+ "epoch": 3.33,
2645
+ "learning_rate": 2.132269099201825e-06,
2646
+ "logits/chosen": -1.1171875,
2647
+ "logits/rejected": -0.4375,
2648
+ "logps/chosen": -280.0,
2649
+ "logps/rejected": -286.0,
2650
+ "loss": 0.0015,
2651
+ "rewards/accuracies": 0.999218761920929,
2652
+ "rewards/chosen": -3.738826036453247,
2653
+ "rewards/margins": 17.777729034423828,
2654
+ "rewards/rejected": -21.51655387878418,
2655
+ "step": 3760
2656
+ },
2657
+ {
2658
+ "epoch": 3.35,
2659
+ "learning_rate": 2.075256556442418e-06,
2660
+ "logits/chosen": -1.109375,
2661
+ "logits/rejected": -0.4140625,
2662
+ "logps/chosen": -282.0,
2663
+ "logps/rejected": -284.0,
2664
+ "loss": 0.0004,
2665
+ "rewards/accuracies": 1.0,
2666
+ "rewards/chosen": -3.708034038543701,
2667
+ "rewards/margins": 17.657176971435547,
2668
+ "rewards/rejected": -21.365211486816406,
2669
+ "step": 3780
2670
+ },
2671
+ {
2672
+ "epoch": 3.37,
2673
+ "learning_rate": 2.0182440136830104e-06,
2674
+ "logits/chosen": -1.03125,
2675
+ "logits/rejected": -0.3359375,
2676
+ "logps/chosen": -284.0,
2677
+ "logps/rejected": -292.0,
2678
+ "loss": 0.0003,
2679
+ "rewards/accuracies": 1.0,
2680
+ "rewards/chosen": -4.185555934906006,
2681
+ "rewards/margins": 18.033720016479492,
2682
+ "rewards/rejected": -22.219276428222656,
2683
+ "step": 3800
2684
+ },
2685
+ {
2686
+ "epoch": 3.39,
2687
+ "learning_rate": 1.9612314709236034e-06,
2688
+ "logits/chosen": -1.0625,
2689
+ "logits/rejected": -0.361328125,
2690
+ "logps/chosen": -276.0,
2691
+ "logps/rejected": -288.0,
2692
+ "loss": 0.0006,
2693
+ "rewards/accuracies": 1.0,
2694
+ "rewards/chosen": -3.9414119720458984,
2695
+ "rewards/margins": 18.10892677307129,
2696
+ "rewards/rejected": -22.050336837768555,
2697
+ "step": 3820
2698
+ },
2699
+ {
2700
+ "epoch": 3.4,
2701
+ "learning_rate": 1.9042189281641962e-06,
2702
+ "logits/chosen": -1.0703125,
2703
+ "logits/rejected": -0.37109375,
2704
+ "logps/chosen": -284.0,
2705
+ "logps/rejected": -292.0,
2706
+ "loss": 0.0011,
2707
+ "rewards/accuracies": 0.999218761920929,
2708
+ "rewards/chosen": -3.8034446239471436,
2709
+ "rewards/margins": 18.498294830322266,
2710
+ "rewards/rejected": -22.301740646362305,
2711
+ "step": 3840
2712
+ },
2713
+ {
2714
+ "epoch": 3.42,
2715
+ "learning_rate": 1.8472063854047893e-06,
2716
+ "logits/chosen": -1.0859375,
2717
+ "logits/rejected": -0.39453125,
2718
+ "logps/chosen": -286.0,
2719
+ "logps/rejected": -292.0,
2720
+ "loss": 0.0002,
2721
+ "rewards/accuracies": 1.0,
2722
+ "rewards/chosen": -3.825915813446045,
2723
+ "rewards/margins": 18.246068954467773,
2724
+ "rewards/rejected": -22.07198715209961,
2725
+ "step": 3860
2726
+ },
2727
+ {
2728
+ "epoch": 3.44,
2729
+ "learning_rate": 1.790193842645382e-06,
2730
+ "logits/chosen": -1.0703125,
2731
+ "logits/rejected": -0.369140625,
2732
+ "logps/chosen": -284.0,
2733
+ "logps/rejected": -292.0,
2734
+ "loss": 0.0006,
2735
+ "rewards/accuracies": 1.0,
2736
+ "rewards/chosen": -3.7209835052490234,
2737
+ "rewards/margins": 18.366897583007812,
2738
+ "rewards/rejected": -22.087881088256836,
2739
+ "step": 3880
2740
+ },
2741
+ {
2742
+ "epoch": 3.46,
2743
+ "learning_rate": 1.733181299885975e-06,
2744
+ "logits/chosen": -1.078125,
2745
+ "logits/rejected": -0.392578125,
2746
+ "logps/chosen": -276.0,
2747
+ "logps/rejected": -286.0,
2748
+ "loss": 0.002,
2749
+ "rewards/accuracies": 0.999218761920929,
2750
+ "rewards/chosen": -3.494877576828003,
2751
+ "rewards/margins": 18.009967803955078,
2752
+ "rewards/rejected": -21.504846572875977,
2753
+ "step": 3900
2754
+ },
2755
+ {
2756
+ "epoch": 3.48,
2757
+ "learning_rate": 1.6761687571265681e-06,
2758
+ "logits/chosen": -1.0703125,
2759
+ "logits/rejected": -0.349609375,
2760
+ "logps/chosen": -282.0,
2761
+ "logps/rejected": -290.0,
2762
+ "loss": 0.0005,
2763
+ "rewards/accuracies": 1.0,
2764
+ "rewards/chosen": -3.914118528366089,
2765
+ "rewards/margins": 18.189693450927734,
2766
+ "rewards/rejected": -22.10381317138672,
2767
+ "step": 3920
2768
+ },
2769
+ {
2770
+ "epoch": 3.49,
2771
+ "learning_rate": 1.6191562143671611e-06,
2772
+ "logits/chosen": -1.109375,
2773
+ "logits/rejected": -0.37890625,
2774
+ "logps/chosen": -284.0,
2775
+ "logps/rejected": -292.0,
2776
+ "loss": 0.0004,
2777
+ "rewards/accuracies": 1.0,
2778
+ "rewards/chosen": -3.6217734813690186,
2779
+ "rewards/margins": 18.476469039916992,
2780
+ "rewards/rejected": -22.098241806030273,
2781
+ "step": 3940
2782
+ },
2783
+ {
2784
+ "epoch": 3.51,
2785
+ "learning_rate": 1.5621436716077537e-06,
2786
+ "logits/chosen": -1.09375,
2787
+ "logits/rejected": -0.392578125,
2788
+ "logps/chosen": -292.0,
2789
+ "logps/rejected": -290.0,
2790
+ "loss": 0.0002,
2791
+ "rewards/accuracies": 1.0,
2792
+ "rewards/chosen": -3.6812424659729004,
2793
+ "rewards/margins": 18.31012725830078,
2794
+ "rewards/rejected": -21.991369247436523,
2795
+ "step": 3960
2796
+ },
2797
+ {
2798
+ "epoch": 3.53,
2799
+ "learning_rate": 1.5051311288483467e-06,
2800
+ "logits/chosen": -1.0703125,
2801
+ "logits/rejected": -0.35546875,
2802
+ "logps/chosen": -290.0,
2803
+ "logps/rejected": -292.0,
2804
+ "loss": 0.0019,
2805
+ "rewards/accuracies": 0.999218761920929,
2806
+ "rewards/chosen": -3.9753355979919434,
2807
+ "rewards/margins": 18.298744201660156,
2808
+ "rewards/rejected": -22.27408218383789,
2809
+ "step": 3980
2810
+ },
2811
+ {
2812
+ "epoch": 3.55,
2813
+ "learning_rate": 1.4481185860889396e-06,
2814
+ "logits/chosen": -1.09375,
2815
+ "logits/rejected": -0.384765625,
2816
+ "logps/chosen": -280.0,
2817
+ "logps/rejected": -284.0,
2818
+ "loss": 0.0003,
2819
+ "rewards/accuracies": 1.0,
2820
+ "rewards/chosen": -3.4559783935546875,
2821
+ "rewards/margins": 18.017427444458008,
2822
+ "rewards/rejected": -21.473405838012695,
2823
+ "step": 4000
2824
+ },
2825
+ {
2826
+ "epoch": 3.56,
2827
+ "learning_rate": 1.3911060433295326e-06,
2828
+ "logits/chosen": -1.0859375,
2829
+ "logits/rejected": -0.37109375,
2830
+ "logps/chosen": -276.0,
2831
+ "logps/rejected": -286.0,
2832
+ "loss": 0.0002,
2833
+ "rewards/accuracies": 1.0,
2834
+ "rewards/chosen": -3.4795451164245605,
2835
+ "rewards/margins": 18.1534423828125,
2836
+ "rewards/rejected": -21.63298797607422,
2837
+ "step": 4020
2838
+ },
2839
+ {
2840
+ "epoch": 3.58,
2841
+ "learning_rate": 1.3340935005701256e-06,
2842
+ "logits/chosen": -1.0625,
2843
+ "logits/rejected": -0.345703125,
2844
+ "logps/chosen": -274.0,
2845
+ "logps/rejected": -286.0,
2846
+ "loss": 0.0009,
2847
+ "rewards/accuracies": 0.999218761920929,
2848
+ "rewards/chosen": -3.496410369873047,
2849
+ "rewards/margins": 18.135766983032227,
2850
+ "rewards/rejected": -21.63217544555664,
2851
+ "step": 4040
2852
+ },
2853
+ {
2854
+ "epoch": 3.6,
2855
+ "learning_rate": 1.2770809578107184e-06,
2856
+ "logits/chosen": -1.0546875,
2857
+ "logits/rejected": -0.3515625,
2858
+ "logps/chosen": -278.0,
2859
+ "logps/rejected": -288.0,
2860
+ "loss": 0.001,
2861
+ "rewards/accuracies": 0.999218761920929,
2862
+ "rewards/chosen": -3.894739866256714,
2863
+ "rewards/margins": 17.927284240722656,
2864
+ "rewards/rejected": -21.822025299072266,
2865
+ "step": 4060
2866
+ },
2867
+ {
2868
+ "epoch": 3.62,
2869
+ "learning_rate": 1.2200684150513114e-06,
2870
+ "logits/chosen": -1.0625,
2871
+ "logits/rejected": -0.3359375,
2872
+ "logps/chosen": -292.0,
2873
+ "logps/rejected": -292.0,
2874
+ "loss": 0.0042,
2875
+ "rewards/accuracies": 0.9984375238418579,
2876
+ "rewards/chosen": -3.9973883628845215,
2877
+ "rewards/margins": 18.207622528076172,
2878
+ "rewards/rejected": -22.205013275146484,
2879
+ "step": 4080
2880
+ },
2881
+ {
2882
+ "epoch": 3.64,
2883
+ "learning_rate": 1.1630558722919042e-06,
2884
+ "logits/chosen": -1.0703125,
2885
+ "logits/rejected": -0.37109375,
2886
+ "logps/chosen": -276.0,
2887
+ "logps/rejected": -284.0,
2888
+ "loss": 0.0034,
2889
+ "rewards/accuracies": 0.999218761920929,
2890
+ "rewards/chosen": -3.779940128326416,
2891
+ "rewards/margins": 17.809322357177734,
2892
+ "rewards/rejected": -21.589258193969727,
2893
+ "step": 4100
2894
+ },
2895
+ {
2896
+ "epoch": 3.65,
2897
+ "learning_rate": 1.1060433295324972e-06,
2898
+ "logits/chosen": -1.0859375,
2899
+ "logits/rejected": -0.369140625,
2900
+ "logps/chosen": -286.0,
2901
+ "logps/rejected": -286.0,
2902
+ "loss": 0.0001,
2903
+ "rewards/accuracies": 1.0,
2904
+ "rewards/chosen": -3.665940761566162,
2905
+ "rewards/margins": 18.151546478271484,
2906
+ "rewards/rejected": -21.817487716674805,
2907
+ "step": 4120
2908
+ },
2909
+ {
2910
+ "epoch": 3.67,
2911
+ "learning_rate": 1.0490307867730903e-06,
2912
+ "logits/chosen": -1.0546875,
2913
+ "logits/rejected": -0.337890625,
2914
+ "logps/chosen": -288.0,
2915
+ "logps/rejected": -290.0,
2916
+ "loss": 0.0006,
2917
+ "rewards/accuracies": 1.0,
2918
+ "rewards/chosen": -3.9725680351257324,
2919
+ "rewards/margins": 18.010974884033203,
2920
+ "rewards/rejected": -21.983543395996094,
2921
+ "step": 4140
2922
+ },
2923
+ {
2924
+ "epoch": 3.69,
2925
+ "learning_rate": 9.92018244013683e-07,
2926
+ "logits/chosen": -1.0859375,
2927
+ "logits/rejected": -0.36328125,
2928
+ "logps/chosen": -294.0,
2929
+ "logps/rejected": -286.0,
2930
+ "loss": 0.0002,
2931
+ "rewards/accuracies": 1.0,
2932
+ "rewards/chosen": -3.550889253616333,
2933
+ "rewards/margins": 18.127422332763672,
2934
+ "rewards/rejected": -21.67831039428711,
2935
+ "step": 4160
2936
+ },
2937
+ {
2938
+ "epoch": 3.71,
2939
+ "learning_rate": 9.35005701254276e-07,
2940
+ "logits/chosen": -1.0234375,
2941
+ "logits/rejected": -0.30859375,
2942
+ "logps/chosen": -268.0,
2943
+ "logps/rejected": -282.0,
2944
+ "loss": 0.0015,
2945
+ "rewards/accuracies": 0.999218761920929,
2946
+ "rewards/chosen": -3.5743343830108643,
2947
+ "rewards/margins": 17.971210479736328,
2948
+ "rewards/rejected": -21.545543670654297,
2949
+ "step": 4180
2950
+ },
2951
+ {
2952
+ "epoch": 3.72,
2953
+ "learning_rate": 8.779931584948689e-07,
2954
+ "logits/chosen": -1.0546875,
2955
+ "logits/rejected": -0.3515625,
2956
+ "logps/chosen": -276.0,
2957
+ "logps/rejected": -292.0,
2958
+ "loss": 0.0018,
2959
+ "rewards/accuracies": 0.999218761920929,
2960
+ "rewards/chosen": -3.8023428916931152,
2961
+ "rewards/margins": 18.363157272338867,
2962
+ "rewards/rejected": -22.16550064086914,
2963
+ "step": 4200
2964
+ },
2965
+ {
2966
+ "epoch": 3.74,
2967
+ "learning_rate": 8.209806157354619e-07,
2968
+ "logits/chosen": -1.015625,
2969
+ "logits/rejected": -0.353515625,
2970
+ "logps/chosen": -274.0,
2971
+ "logps/rejected": -296.0,
2972
+ "loss": 0.0007,
2973
+ "rewards/accuracies": 1.0,
2974
+ "rewards/chosen": -4.0028395652771,
2975
+ "rewards/margins": 18.4693603515625,
2976
+ "rewards/rejected": -22.47220230102539,
2977
+ "step": 4220
2978
+ },
2979
+ {
2980
+ "epoch": 3.76,
2981
+ "learning_rate": 7.639680729760548e-07,
2982
+ "logits/chosen": -1.0546875,
2983
+ "logits/rejected": -0.337890625,
2984
+ "logps/chosen": -286.0,
2985
+ "logps/rejected": -288.0,
2986
+ "loss": 0.0008,
2987
+ "rewards/accuracies": 1.0,
2988
+ "rewards/chosen": -3.7589478492736816,
2989
+ "rewards/margins": 18.027183532714844,
2990
+ "rewards/rejected": -21.786130905151367,
2991
+ "step": 4240
2992
+ },
2993
+ {
2994
+ "epoch": 3.78,
2995
+ "learning_rate": 7.069555302166476e-07,
2996
+ "logits/chosen": -1.0625,
2997
+ "logits/rejected": -0.373046875,
2998
+ "logps/chosen": -274.0,
2999
+ "logps/rejected": -290.0,
3000
+ "loss": 0.0007,
3001
+ "rewards/accuracies": 1.0,
3002
+ "rewards/chosen": -3.6975696086883545,
3003
+ "rewards/margins": 18.218469619750977,
3004
+ "rewards/rejected": -21.916040420532227,
3005
+ "step": 4260
3006
+ },
3007
+ {
3008
+ "epoch": 3.79,
3009
+ "learning_rate": 6.499429874572407e-07,
3010
+ "logits/chosen": -1.046875,
3011
+ "logits/rejected": -0.34375,
3012
+ "logps/chosen": -282.0,
3013
+ "logps/rejected": -290.0,
3014
+ "loss": 0.0013,
3015
+ "rewards/accuracies": 1.0,
3016
+ "rewards/chosen": -3.9434094429016113,
3017
+ "rewards/margins": 18.002216339111328,
3018
+ "rewards/rejected": -21.94562339782715,
3019
+ "step": 4280
3020
+ },
3021
+ {
3022
+ "epoch": 3.81,
3023
+ "learning_rate": 5.929304446978336e-07,
3024
+ "logits/chosen": -1.046875,
3025
+ "logits/rejected": -0.330078125,
3026
+ "logps/chosen": -278.0,
3027
+ "logps/rejected": -288.0,
3028
+ "loss": 0.0001,
3029
+ "rewards/accuracies": 1.0,
3030
+ "rewards/chosen": -3.605856418609619,
3031
+ "rewards/margins": 18.189517974853516,
3032
+ "rewards/rejected": -21.79537582397461,
3033
+ "step": 4300
3034
+ },
3035
+ {
3036
+ "epoch": 3.83,
3037
+ "learning_rate": 5.359179019384265e-07,
3038
+ "logits/chosen": -1.0625,
3039
+ "logits/rejected": -0.34375,
3040
+ "logps/chosen": -284.0,
3041
+ "logps/rejected": -298.0,
3042
+ "loss": 0.0004,
3043
+ "rewards/accuracies": 1.0,
3044
+ "rewards/chosen": -3.889907121658325,
3045
+ "rewards/margins": 18.669677734375,
3046
+ "rewards/rejected": -22.55958366394043,
3047
+ "step": 4320
3048
+ },
3049
+ {
3050
+ "epoch": 3.85,
3051
+ "learning_rate": 4.789053591790194e-07,
3052
+ "logits/chosen": -1.0546875,
3053
+ "logits/rejected": -0.33984375,
3054
+ "logps/chosen": -282.0,
3055
+ "logps/rejected": -294.0,
3056
+ "loss": 0.0002,
3057
+ "rewards/accuracies": 1.0,
3058
+ "rewards/chosen": -3.840752124786377,
3059
+ "rewards/margins": 18.512104034423828,
3060
+ "rewards/rejected": -22.352855682373047,
3061
+ "step": 4340
3062
+ },
3063
+ {
3064
+ "epoch": 3.87,
3065
+ "learning_rate": 4.218928164196123e-07,
3066
+ "logits/chosen": -1.0390625,
3067
+ "logits/rejected": -0.341796875,
3068
+ "logps/chosen": -280.0,
3069
+ "logps/rejected": -294.0,
3070
+ "loss": 0.0015,
3071
+ "rewards/accuracies": 0.999218761920929,
3072
+ "rewards/chosen": -4.100388526916504,
3073
+ "rewards/margins": 18.11039161682129,
3074
+ "rewards/rejected": -22.21078109741211,
3075
+ "step": 4360
3076
+ },
3077
+ {
3078
+ "epoch": 3.88,
3079
+ "learning_rate": 3.648802736602053e-07,
3080
+ "logits/chosen": -1.0234375,
3081
+ "logits/rejected": -0.341796875,
3082
+ "logps/chosen": -268.0,
3083
+ "logps/rejected": -286.0,
3084
+ "loss": 0.0002,
3085
+ "rewards/accuracies": 1.0,
3086
+ "rewards/chosen": -3.7678654193878174,
3087
+ "rewards/margins": 17.92337417602539,
3088
+ "rewards/rejected": -21.691240310668945,
3089
+ "step": 4380
3090
+ },
3091
+ {
3092
+ "epoch": 3.9,
3093
+ "learning_rate": 3.078677309007982e-07,
3094
+ "logits/chosen": -1.0390625,
3095
+ "logits/rejected": -0.33984375,
3096
+ "logps/chosen": -282.0,
3097
+ "logps/rejected": -294.0,
3098
+ "loss": 0.0004,
3099
+ "rewards/accuracies": 1.0,
3100
+ "rewards/chosen": -3.8660149574279785,
3101
+ "rewards/margins": 18.40440559387207,
3102
+ "rewards/rejected": -22.27042007446289,
3103
+ "step": 4400
3104
+ }
3105
+ ],
3106
+ "logging_steps": 20,
3107
+ "max_steps": 4508,
3108
+ "num_input_tokens_seen": 0,
3109
+ "num_train_epochs": 4,
3110
+ "save_steps": 200,
3111
+ "total_flos": 0.0,
3112
+ "train_batch_size": 8,
3113
+ "trial_name": null,
3114
+ "trial_params": null
3115
+ }
training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5e238a7355d73b3944d7775d498bb670befae66eb4fe099090b48fae73d0c43c
3
+ size 4728