Commit ·
481ef75
1
Parent(s): cefed97
Upload folder using huggingface_hub
Browse files- README.md +21 -0
- adapter_config.json +21 -0
- adapter_model.bin +3 -0
- special_tokens_map.json +24 -0
- tokenizer.json +0 -0
- tokenizer_config.json +32 -0
- train_args.json +123 -0
- training_args.bin +3 -0
- training_logs.json +3179 -0
README.md
ADDED
|
@@ -0,0 +1,21 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
library_name: peft
|
| 3 |
+
---
|
| 4 |
+
## Training procedure
|
| 5 |
+
|
| 6 |
+
|
| 7 |
+
The following `bitsandbytes` quantization config was used during training:
|
| 8 |
+
- quant_method: bitsandbytes
|
| 9 |
+
- load_in_8bit: False
|
| 10 |
+
- load_in_4bit: True
|
| 11 |
+
- llm_int8_threshold: 6.0
|
| 12 |
+
- llm_int8_skip_modules: None
|
| 13 |
+
- llm_int8_enable_fp32_cpu_offload: False
|
| 14 |
+
- llm_int8_has_fp16_weight: False
|
| 15 |
+
- bnb_4bit_quant_type: nf4
|
| 16 |
+
- bnb_4bit_use_double_quant: True
|
| 17 |
+
- bnb_4bit_compute_dtype: float16
|
| 18 |
+
### Framework versions
|
| 19 |
+
|
| 20 |
+
|
| 21 |
+
- PEFT 0.5.0
|
adapter_config.json
ADDED
|
@@ -0,0 +1,21 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"auto_mapping": null,
|
| 3 |
+
"base_model_name_or_path": "models/llama-13b",
|
| 4 |
+
"bias": "none",
|
| 5 |
+
"fan_in_fan_out": false,
|
| 6 |
+
"inference_mode": true,
|
| 7 |
+
"init_lora_weights": true,
|
| 8 |
+
"layers_pattern": null,
|
| 9 |
+
"layers_to_transform": null,
|
| 10 |
+
"lora_alpha": 16,
|
| 11 |
+
"lora_dropout": 0.1,
|
| 12 |
+
"modules_to_save": null,
|
| 13 |
+
"peft_type": "LORA",
|
| 14 |
+
"r": 64,
|
| 15 |
+
"revision": null,
|
| 16 |
+
"target_modules": [
|
| 17 |
+
"q_proj",
|
| 18 |
+
"v_proj"
|
| 19 |
+
],
|
| 20 |
+
"task_type": "SEQ_CLS"
|
| 21 |
+
}
|
adapter_model.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c09043e9efc5740ba9fc9403a9257b8d9feacc0fd7d73fe7e1a42f2386d65a9e
|
| 3 |
+
size 209811921
|
special_tokens_map.json
ADDED
|
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bos_token": {
|
| 3 |
+
"content": "<s>",
|
| 4 |
+
"lstrip": false,
|
| 5 |
+
"normalized": true,
|
| 6 |
+
"rstrip": false,
|
| 7 |
+
"single_word": false
|
| 8 |
+
},
|
| 9 |
+
"eos_token": {
|
| 10 |
+
"content": "</s>",
|
| 11 |
+
"lstrip": false,
|
| 12 |
+
"normalized": true,
|
| 13 |
+
"rstrip": false,
|
| 14 |
+
"single_word": false
|
| 15 |
+
},
|
| 16 |
+
"pad_token": "</s>",
|
| 17 |
+
"unk_token": {
|
| 18 |
+
"content": "<unk>",
|
| 19 |
+
"lstrip": false,
|
| 20 |
+
"normalized": true,
|
| 21 |
+
"rstrip": false,
|
| 22 |
+
"single_word": false
|
| 23 |
+
}
|
| 24 |
+
}
|
tokenizer.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
tokenizer_config.json
ADDED
|
@@ -0,0 +1,32 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bos_token": {
|
| 3 |
+
"__type": "AddedToken",
|
| 4 |
+
"content": "<s>",
|
| 5 |
+
"lstrip": false,
|
| 6 |
+
"normalized": true,
|
| 7 |
+
"rstrip": false,
|
| 8 |
+
"single_word": false
|
| 9 |
+
},
|
| 10 |
+
"clean_up_tokenization_spaces": false,
|
| 11 |
+
"eos_token": {
|
| 12 |
+
"__type": "AddedToken",
|
| 13 |
+
"content": "</s>",
|
| 14 |
+
"lstrip": false,
|
| 15 |
+
"normalized": true,
|
| 16 |
+
"rstrip": false,
|
| 17 |
+
"single_word": false
|
| 18 |
+
},
|
| 19 |
+
"model_max_length": 2048,
|
| 20 |
+
"pad_token": null,
|
| 21 |
+
"sp_model_kwargs": {},
|
| 22 |
+
"tokenizer_class": "LlamaTokenizer",
|
| 23 |
+
"unk_token": {
|
| 24 |
+
"__type": "AddedToken",
|
| 25 |
+
"content": "<unk>",
|
| 26 |
+
"lstrip": false,
|
| 27 |
+
"normalized": true,
|
| 28 |
+
"rstrip": false,
|
| 29 |
+
"single_word": false
|
| 30 |
+
},
|
| 31 |
+
"use_default_system_prompt": true
|
| 32 |
+
}
|
train_args.json
ADDED
|
@@ -0,0 +1,123 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"initial_model_dir": "models/classify_lora/llama-13b-code/checkpoint-150",
|
| 3 |
+
"distribution_id": "code",
|
| 4 |
+
"date_trained": "09/10/2023 23:19:58",
|
| 5 |
+
"output_dir": "models/classify_lora/llama-13b-code",
|
| 6 |
+
"overwrite_output_dir": false,
|
| 7 |
+
"do_train": false,
|
| 8 |
+
"do_eval": true,
|
| 9 |
+
"do_predict": false,
|
| 10 |
+
"evaluation_strategy": "steps",
|
| 11 |
+
"prediction_loss_only": false,
|
| 12 |
+
"per_device_train_batch_size": 16,
|
| 13 |
+
"per_device_eval_batch_size": 16,
|
| 14 |
+
"per_gpu_train_batch_size": null,
|
| 15 |
+
"per_gpu_eval_batch_size": null,
|
| 16 |
+
"gradient_accumulation_steps": 1,
|
| 17 |
+
"eval_accumulation_steps": null,
|
| 18 |
+
"eval_delay": 0,
|
| 19 |
+
"learning_rate": 0.0002,
|
| 20 |
+
"weight_decay": 0.0,
|
| 21 |
+
"adam_beta1": 0.9,
|
| 22 |
+
"adam_beta2": 0.999,
|
| 23 |
+
"adam_epsilon": 1e-08,
|
| 24 |
+
"max_grad_norm": 0.3,
|
| 25 |
+
"num_train_epochs": 6,
|
| 26 |
+
"max_steps": 300,
|
| 27 |
+
"lr_scheduler_type": "constant",
|
| 28 |
+
"warmup_ratio": 0.03,
|
| 29 |
+
"warmup_steps": 0,
|
| 30 |
+
"log_level": "passive",
|
| 31 |
+
"log_level_replica": "warning",
|
| 32 |
+
"log_on_each_node": true,
|
| 33 |
+
"logging_dir": "models/classify_lora/llama-13b-code/runs/Oct09_19-32-51_compute-permanent-node-990",
|
| 34 |
+
"logging_strategy": "steps",
|
| 35 |
+
"logging_first_step": false,
|
| 36 |
+
"logging_steps": 1,
|
| 37 |
+
"logging_nan_inf_filter": true,
|
| 38 |
+
"save_strategy": "steps",
|
| 39 |
+
"save_steps": 25,
|
| 40 |
+
"save_total_limit": 0,
|
| 41 |
+
"save_safetensors": false,
|
| 42 |
+
"save_on_each_node": false,
|
| 43 |
+
"no_cuda": false,
|
| 44 |
+
"use_cpu": false,
|
| 45 |
+
"use_mps_device": false,
|
| 46 |
+
"seed": 42,
|
| 47 |
+
"data_seed": null,
|
| 48 |
+
"jit_mode_eval": false,
|
| 49 |
+
"use_ipex": false,
|
| 50 |
+
"bf16": false,
|
| 51 |
+
"fp16": false,
|
| 52 |
+
"fp16_opt_level": "O1",
|
| 53 |
+
"half_precision_backend": "auto",
|
| 54 |
+
"bf16_full_eval": false,
|
| 55 |
+
"fp16_full_eval": false,
|
| 56 |
+
"tf32": null,
|
| 57 |
+
"local_rank": 0,
|
| 58 |
+
"ddp_backend": null,
|
| 59 |
+
"tpu_num_cores": null,
|
| 60 |
+
"tpu_metrics_debug": false,
|
| 61 |
+
"debug": [],
|
| 62 |
+
"dataloader_drop_last": false,
|
| 63 |
+
"eval_steps": 25,
|
| 64 |
+
"dataloader_num_workers": 0,
|
| 65 |
+
"past_index": -1,
|
| 66 |
+
"run_name": "train|models-classify_lora-llama-13b-code",
|
| 67 |
+
"disable_tqdm": false,
|
| 68 |
+
"remove_unused_columns": false,
|
| 69 |
+
"label_names": null,
|
| 70 |
+
"load_best_model_at_end": false,
|
| 71 |
+
"metric_for_best_model": "eval_code_score",
|
| 72 |
+
"greater_is_better": true,
|
| 73 |
+
"ignore_data_skip": false,
|
| 74 |
+
"sharded_ddp": [],
|
| 75 |
+
"fsdp": [],
|
| 76 |
+
"fsdp_min_num_params": 0,
|
| 77 |
+
"fsdp_config": {
|
| 78 |
+
"min_num_params": 0,
|
| 79 |
+
"xla": false,
|
| 80 |
+
"xla_fsdp_grad_ckpt": false
|
| 81 |
+
},
|
| 82 |
+
"fsdp_transformer_layer_cls_to_wrap": null,
|
| 83 |
+
"deepspeed": "configs/ds_zero_1.json",
|
| 84 |
+
"label_smoothing_factor": 0.0,
|
| 85 |
+
"optim": "paged_adamw_32bit",
|
| 86 |
+
"optim_args": null,
|
| 87 |
+
"adafactor": false,
|
| 88 |
+
"group_by_length": false,
|
| 89 |
+
"length_column_name": "length",
|
| 90 |
+
"report_to": [
|
| 91 |
+
"wandb"
|
| 92 |
+
],
|
| 93 |
+
"ddp_find_unused_parameters": false,
|
| 94 |
+
"ddp_bucket_cap_mb": null,
|
| 95 |
+
"ddp_broadcast_buffers": null,
|
| 96 |
+
"dataloader_pin_memory": true,
|
| 97 |
+
"skip_memory_metrics": true,
|
| 98 |
+
"use_legacy_prediction_loop": false,
|
| 99 |
+
"push_to_hub": false,
|
| 100 |
+
"resume_from_checkpoint": null,
|
| 101 |
+
"hub_model_id": null,
|
| 102 |
+
"hub_strategy": "every_save",
|
| 103 |
+
"hub_token": null,
|
| 104 |
+
"hub_private_repo": false,
|
| 105 |
+
"hub_always_push": false,
|
| 106 |
+
"gradient_checkpointing": false,
|
| 107 |
+
"include_inputs_for_metrics": false,
|
| 108 |
+
"fp16_backend": "auto",
|
| 109 |
+
"push_to_hub_model_id": null,
|
| 110 |
+
"push_to_hub_organization": null,
|
| 111 |
+
"push_to_hub_token": null,
|
| 112 |
+
"_n_gpu": 1,
|
| 113 |
+
"mp_parameters": "",
|
| 114 |
+
"auto_find_batch_size": false,
|
| 115 |
+
"full_determinism": false,
|
| 116 |
+
"torchdynamo": null,
|
| 117 |
+
"ray_scope": "last",
|
| 118 |
+
"ddp_timeout": 1800,
|
| 119 |
+
"torch_compile": false,
|
| 120 |
+
"torch_compile_backend": null,
|
| 121 |
+
"torch_compile_mode": null,
|
| 122 |
+
"dispatch_batches": null
|
| 123 |
+
}
|
training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6877989a44958189c82de55e0723e0899a7549f16dd9c54f2f83b941e2c6a514
|
| 3 |
+
size 5115
|
training_logs.json
ADDED
|
@@ -0,0 +1,3179 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"epoch": 0.02,
|
| 4 |
+
"learning_rate": 0.0002,
|
| 5 |
+
"loss": 0.7109,
|
| 6 |
+
"step": 1
|
| 7 |
+
},
|
| 8 |
+
{
|
| 9 |
+
"epoch": 0.04,
|
| 10 |
+
"learning_rate": 0.0002,
|
| 11 |
+
"loss": 0.7151,
|
| 12 |
+
"step": 2
|
| 13 |
+
},
|
| 14 |
+
{
|
| 15 |
+
"epoch": 0.05,
|
| 16 |
+
"learning_rate": 0.0002,
|
| 17 |
+
"loss": 0.7036,
|
| 18 |
+
"step": 3
|
| 19 |
+
},
|
| 20 |
+
{
|
| 21 |
+
"epoch": 0.07,
|
| 22 |
+
"learning_rate": 0.0002,
|
| 23 |
+
"loss": 0.6805,
|
| 24 |
+
"step": 4
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.09,
|
| 28 |
+
"learning_rate": 0.0002,
|
| 29 |
+
"loss": 0.6678,
|
| 30 |
+
"step": 5
|
| 31 |
+
},
|
| 32 |
+
{
|
| 33 |
+
"epoch": 0.11,
|
| 34 |
+
"learning_rate": 0.0002,
|
| 35 |
+
"loss": 0.6692,
|
| 36 |
+
"step": 6
|
| 37 |
+
},
|
| 38 |
+
{
|
| 39 |
+
"epoch": 0.12,
|
| 40 |
+
"learning_rate": 0.0002,
|
| 41 |
+
"loss": 0.6459,
|
| 42 |
+
"step": 7
|
| 43 |
+
},
|
| 44 |
+
{
|
| 45 |
+
"epoch": 0.14,
|
| 46 |
+
"learning_rate": 0.0002,
|
| 47 |
+
"loss": 0.6262,
|
| 48 |
+
"step": 8
|
| 49 |
+
},
|
| 50 |
+
{
|
| 51 |
+
"epoch": 0.16,
|
| 52 |
+
"learning_rate": 0.0002,
|
| 53 |
+
"loss": 0.6574,
|
| 54 |
+
"step": 9
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.18,
|
| 58 |
+
"learning_rate": 0.0002,
|
| 59 |
+
"loss": 0.6311,
|
| 60 |
+
"step": 10
|
| 61 |
+
},
|
| 62 |
+
{
|
| 63 |
+
"epoch": 0.19,
|
| 64 |
+
"learning_rate": 0.0002,
|
| 65 |
+
"loss": 0.5219,
|
| 66 |
+
"step": 11
|
| 67 |
+
},
|
| 68 |
+
{
|
| 69 |
+
"epoch": 0.21,
|
| 70 |
+
"learning_rate": 0.0002,
|
| 71 |
+
"loss": 0.5933,
|
| 72 |
+
"step": 12
|
| 73 |
+
},
|
| 74 |
+
{
|
| 75 |
+
"epoch": 0.23,
|
| 76 |
+
"learning_rate": 0.0002,
|
| 77 |
+
"loss": 0.5881,
|
| 78 |
+
"step": 13
|
| 79 |
+
},
|
| 80 |
+
{
|
| 81 |
+
"epoch": 0.25,
|
| 82 |
+
"learning_rate": 0.0002,
|
| 83 |
+
"loss": 0.6345,
|
| 84 |
+
"step": 14
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.26,
|
| 88 |
+
"learning_rate": 0.0002,
|
| 89 |
+
"loss": 0.559,
|
| 90 |
+
"step": 15
|
| 91 |
+
},
|
| 92 |
+
{
|
| 93 |
+
"epoch": 0.28,
|
| 94 |
+
"learning_rate": 0.0002,
|
| 95 |
+
"loss": 0.5686,
|
| 96 |
+
"step": 16
|
| 97 |
+
},
|
| 98 |
+
{
|
| 99 |
+
"epoch": 0.3,
|
| 100 |
+
"learning_rate": 0.0002,
|
| 101 |
+
"loss": 0.577,
|
| 102 |
+
"step": 17
|
| 103 |
+
},
|
| 104 |
+
{
|
| 105 |
+
"epoch": 0.32,
|
| 106 |
+
"learning_rate": 0.0002,
|
| 107 |
+
"loss": 0.4884,
|
| 108 |
+
"step": 18
|
| 109 |
+
},
|
| 110 |
+
{
|
| 111 |
+
"epoch": 0.33,
|
| 112 |
+
"learning_rate": 0.0002,
|
| 113 |
+
"loss": 0.4593,
|
| 114 |
+
"step": 19
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.35,
|
| 118 |
+
"learning_rate": 0.0002,
|
| 119 |
+
"loss": 0.4582,
|
| 120 |
+
"step": 20
|
| 121 |
+
},
|
| 122 |
+
{
|
| 123 |
+
"epoch": 0.37,
|
| 124 |
+
"learning_rate": 0.0002,
|
| 125 |
+
"loss": 0.428,
|
| 126 |
+
"step": 21
|
| 127 |
+
},
|
| 128 |
+
{
|
| 129 |
+
"epoch": 0.39,
|
| 130 |
+
"learning_rate": 0.0002,
|
| 131 |
+
"loss": 0.495,
|
| 132 |
+
"step": 22
|
| 133 |
+
},
|
| 134 |
+
{
|
| 135 |
+
"epoch": 0.4,
|
| 136 |
+
"learning_rate": 0.0002,
|
| 137 |
+
"loss": 0.4683,
|
| 138 |
+
"step": 23
|
| 139 |
+
},
|
| 140 |
+
{
|
| 141 |
+
"epoch": 0.42,
|
| 142 |
+
"learning_rate": 0.0002,
|
| 143 |
+
"loss": 0.3779,
|
| 144 |
+
"step": 24
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 0.44,
|
| 148 |
+
"learning_rate": 0.0002,
|
| 149 |
+
"loss": 0.5694,
|
| 150 |
+
"step": 25
|
| 151 |
+
},
|
| 152 |
+
{
|
| 153 |
+
"epoch": 0.44,
|
| 154 |
+
"eval_code_accuracy": 0.81,
|
| 155 |
+
"eval_code_average_probability": 0.6998826861381531,
|
| 156 |
+
"eval_code_brier_score": 0.13871249556541443,
|
| 157 |
+
"eval_code_loss": 0.392591655254364,
|
| 158 |
+
"eval_code_probabilities": [
|
| 159 |
+
0.9998756647109985,
|
| 160 |
+
0.9996539354324341,
|
| 161 |
+
0.9968274235725403,
|
| 162 |
+
0.867797315120697,
|
| 163 |
+
0.604396641254425,
|
| 164 |
+
0.8194839358329773,
|
| 165 |
+
0.9978196620941162,
|
| 166 |
+
0.9973512887954712,
|
| 167 |
+
0.8165198564529419,
|
| 168 |
+
0.929094672203064,
|
| 169 |
+
0.9696783423423767,
|
| 170 |
+
0.8749808073043823,
|
| 171 |
+
0.9949828386306763,
|
| 172 |
+
0.9988803267478943,
|
| 173 |
+
0.9974466562271118,
|
| 174 |
+
0.5661922693252563,
|
| 175 |
+
0.7031828761100769,
|
| 176 |
+
0.7817732095718384,
|
| 177 |
+
0.7952773571014404,
|
| 178 |
+
0.700941801071167,
|
| 179 |
+
0.5993526577949524,
|
| 180 |
+
0.4928004741668701,
|
| 181 |
+
0.7662289142608643,
|
| 182 |
+
0.7449039816856384,
|
| 183 |
+
0.26918813586235046,
|
| 184 |
+
0.23834775388240814,
|
| 185 |
+
0.21881021559238434,
|
| 186 |
+
0.5174687504768372,
|
| 187 |
+
0.48715656995773315,
|
| 188 |
+
0.5120030045509338,
|
| 189 |
+
0.3874703049659729,
|
| 190 |
+
0.46171045303344727,
|
| 191 |
+
0.6353215575218201,
|
| 192 |
+
0.9995631575584412,
|
| 193 |
+
0.9988592863082886,
|
| 194 |
+
0.9959748387336731,
|
| 195 |
+
0.9921092391014099,
|
| 196 |
+
0.9928039908409119,
|
| 197 |
+
0.9970149993896484,
|
| 198 |
+
0.5361520648002625,
|
| 199 |
+
0.6019002795219421,
|
| 200 |
+
0.4536624550819397,
|
| 201 |
+
0.6162904500961304,
|
| 202 |
+
0.5367547869682312,
|
| 203 |
+
0.620251476764679,
|
| 204 |
+
0.5513021349906921,
|
| 205 |
+
0.5809685587882996,
|
| 206 |
+
0.5933180451393127,
|
| 207 |
+
0.726946234703064,
|
| 208 |
+
0.6555966138839722,
|
| 209 |
+
0.999262273311615,
|
| 210 |
+
0.9996010661125183,
|
| 211 |
+
0.9998406171798706,
|
| 212 |
+
0.7900382280349731,
|
| 213 |
+
0.5323582291603088,
|
| 214 |
+
0.5186650156974792,
|
| 215 |
+
0.5038493275642395,
|
| 216 |
+
0.9872013330459595,
|
| 217 |
+
0.5808623433113098,
|
| 218 |
+
0.9900431632995605,
|
| 219 |
+
0.9054338335990906,
|
| 220 |
+
0.8925741910934448,
|
| 221 |
+
0.8963204026222229,
|
| 222 |
+
0.48921719193458557,
|
| 223 |
+
0.565790057182312,
|
| 224 |
+
0.544992208480835,
|
| 225 |
+
0.6663545370101929,
|
| 226 |
+
0.49584266543388367,
|
| 227 |
+
0.4396786093711853,
|
| 228 |
+
0.5050247311592102,
|
| 229 |
+
0.4423765242099762,
|
| 230 |
+
0.5252822637557983,
|
| 231 |
+
0.5083388686180115,
|
| 232 |
+
0.5204058885574341,
|
| 233 |
+
0.4985828101634979,
|
| 234 |
+
0.9953283071517944,
|
| 235 |
+
0.9802095890045166,
|
| 236 |
+
0.9042825698852539,
|
| 237 |
+
0.9987348914146423,
|
| 238 |
+
0.9832821488380432,
|
| 239 |
+
0.9921707510948181,
|
| 240 |
+
0.5249463319778442,
|
| 241 |
+
0.496512770652771,
|
| 242 |
+
0.529527485370636,
|
| 243 |
+
0.6597751975059509,
|
| 244 |
+
0.9941974878311157,
|
| 245 |
+
0.8405609130859375,
|
| 246 |
+
0.6745859384536743,
|
| 247 |
+
0.6913660168647766,
|
| 248 |
+
0.4706657826900482,
|
| 249 |
+
0.6228535771369934,
|
| 250 |
+
0.7330396175384521,
|
| 251 |
+
0.5534575581550598,
|
| 252 |
+
0.4912925362586975,
|
| 253 |
+
0.5196889042854309,
|
| 254 |
+
0.5061907768249512,
|
| 255 |
+
0.527052640914917,
|
| 256 |
+
0.46116262674331665,
|
| 257 |
+
0.40944358706474304,
|
| 258 |
+
0.4376073181629181
|
| 259 |
+
],
|
| 260 |
+
"eval_code_runtime": 105.7673,
|
| 261 |
+
"eval_code_samples_per_second": 0.945,
|
| 262 |
+
"eval_code_score": -0.13871249556541443,
|
| 263 |
+
"eval_code_steps_per_second": 0.038,
|
| 264 |
+
"step": 25
|
| 265 |
+
},
|
| 266 |
+
{
|
| 267 |
+
"epoch": 0.46,
|
| 268 |
+
"learning_rate": 0.0002,
|
| 269 |
+
"loss": 0.3947,
|
| 270 |
+
"step": 26
|
| 271 |
+
},
|
| 272 |
+
{
|
| 273 |
+
"epoch": 0.47,
|
| 274 |
+
"learning_rate": 0.0002,
|
| 275 |
+
"loss": 0.4236,
|
| 276 |
+
"step": 27
|
| 277 |
+
},
|
| 278 |
+
{
|
| 279 |
+
"epoch": 0.49,
|
| 280 |
+
"learning_rate": 0.0002,
|
| 281 |
+
"loss": 0.349,
|
| 282 |
+
"step": 28
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"epoch": 0.51,
|
| 286 |
+
"learning_rate": 0.0002,
|
| 287 |
+
"loss": 0.3414,
|
| 288 |
+
"step": 29
|
| 289 |
+
},
|
| 290 |
+
{
|
| 291 |
+
"epoch": 0.53,
|
| 292 |
+
"learning_rate": 0.0002,
|
| 293 |
+
"loss": 0.5019,
|
| 294 |
+
"step": 30
|
| 295 |
+
},
|
| 296 |
+
{
|
| 297 |
+
"epoch": 0.54,
|
| 298 |
+
"learning_rate": 0.0002,
|
| 299 |
+
"loss": 0.3768,
|
| 300 |
+
"step": 31
|
| 301 |
+
},
|
| 302 |
+
{
|
| 303 |
+
"epoch": 0.56,
|
| 304 |
+
"learning_rate": 0.0002,
|
| 305 |
+
"loss": 0.3902,
|
| 306 |
+
"step": 32
|
| 307 |
+
},
|
| 308 |
+
{
|
| 309 |
+
"epoch": 0.58,
|
| 310 |
+
"learning_rate": 0.0002,
|
| 311 |
+
"loss": 0.2838,
|
| 312 |
+
"step": 33
|
| 313 |
+
},
|
| 314 |
+
{
|
| 315 |
+
"epoch": 0.6,
|
| 316 |
+
"learning_rate": 0.0002,
|
| 317 |
+
"loss": 0.1972,
|
| 318 |
+
"step": 34
|
| 319 |
+
},
|
| 320 |
+
{
|
| 321 |
+
"epoch": 0.61,
|
| 322 |
+
"learning_rate": 0.0002,
|
| 323 |
+
"loss": 0.3796,
|
| 324 |
+
"step": 35
|
| 325 |
+
},
|
| 326 |
+
{
|
| 327 |
+
"epoch": 0.63,
|
| 328 |
+
"learning_rate": 0.0002,
|
| 329 |
+
"loss": 0.3741,
|
| 330 |
+
"step": 36
|
| 331 |
+
},
|
| 332 |
+
{
|
| 333 |
+
"epoch": 0.65,
|
| 334 |
+
"learning_rate": 0.0002,
|
| 335 |
+
"loss": 0.2624,
|
| 336 |
+
"step": 37
|
| 337 |
+
},
|
| 338 |
+
{
|
| 339 |
+
"epoch": 0.67,
|
| 340 |
+
"learning_rate": 0.0002,
|
| 341 |
+
"loss": 0.2505,
|
| 342 |
+
"step": 38
|
| 343 |
+
},
|
| 344 |
+
{
|
| 345 |
+
"epoch": 0.68,
|
| 346 |
+
"learning_rate": 0.0002,
|
| 347 |
+
"loss": 0.2687,
|
| 348 |
+
"step": 39
|
| 349 |
+
},
|
| 350 |
+
{
|
| 351 |
+
"epoch": 0.7,
|
| 352 |
+
"learning_rate": 0.0002,
|
| 353 |
+
"loss": 0.2856,
|
| 354 |
+
"step": 40
|
| 355 |
+
},
|
| 356 |
+
{
|
| 357 |
+
"epoch": 0.72,
|
| 358 |
+
"learning_rate": 0.0002,
|
| 359 |
+
"loss": 0.1638,
|
| 360 |
+
"step": 41
|
| 361 |
+
},
|
| 362 |
+
{
|
| 363 |
+
"epoch": 0.74,
|
| 364 |
+
"learning_rate": 0.0002,
|
| 365 |
+
"loss": 0.2595,
|
| 366 |
+
"step": 42
|
| 367 |
+
},
|
| 368 |
+
{
|
| 369 |
+
"epoch": 0.75,
|
| 370 |
+
"learning_rate": 0.0002,
|
| 371 |
+
"loss": 0.2886,
|
| 372 |
+
"step": 43
|
| 373 |
+
},
|
| 374 |
+
{
|
| 375 |
+
"epoch": 0.77,
|
| 376 |
+
"learning_rate": 0.0002,
|
| 377 |
+
"loss": 0.1604,
|
| 378 |
+
"step": 44
|
| 379 |
+
},
|
| 380 |
+
{
|
| 381 |
+
"epoch": 0.79,
|
| 382 |
+
"learning_rate": 0.0002,
|
| 383 |
+
"loss": 0.2069,
|
| 384 |
+
"step": 45
|
| 385 |
+
},
|
| 386 |
+
{
|
| 387 |
+
"epoch": 0.81,
|
| 388 |
+
"learning_rate": 0.0002,
|
| 389 |
+
"loss": 0.2285,
|
| 390 |
+
"step": 46
|
| 391 |
+
},
|
| 392 |
+
{
|
| 393 |
+
"epoch": 0.82,
|
| 394 |
+
"learning_rate": 0.0002,
|
| 395 |
+
"loss": 0.1975,
|
| 396 |
+
"step": 47
|
| 397 |
+
},
|
| 398 |
+
{
|
| 399 |
+
"epoch": 0.84,
|
| 400 |
+
"learning_rate": 0.0002,
|
| 401 |
+
"loss": 0.1726,
|
| 402 |
+
"step": 48
|
| 403 |
+
},
|
| 404 |
+
{
|
| 405 |
+
"epoch": 0.86,
|
| 406 |
+
"learning_rate": 0.0002,
|
| 407 |
+
"loss": 0.0881,
|
| 408 |
+
"step": 49
|
| 409 |
+
},
|
| 410 |
+
{
|
| 411 |
+
"epoch": 0.88,
|
| 412 |
+
"learning_rate": 0.0002,
|
| 413 |
+
"loss": 0.2712,
|
| 414 |
+
"step": 50
|
| 415 |
+
},
|
| 416 |
+
{
|
| 417 |
+
"epoch": 0.88,
|
| 418 |
+
"eval_code_accuracy": 0.93,
|
| 419 |
+
"eval_code_average_probability": 0.8494662642478943,
|
| 420 |
+
"eval_code_brier_score": 0.06509324163198471,
|
| 421 |
+
"eval_code_loss": 0.18686418235301971,
|
| 422 |
+
"eval_code_probabilities": [
|
| 423 |
+
0.9999998807907104,
|
| 424 |
+
0.9999940395355225,
|
| 425 |
+
0.9999998807907104,
|
| 426 |
+
0.999998927116394,
|
| 427 |
+
0.9994799494743347,
|
| 428 |
+
0.9999616146087646,
|
| 429 |
+
1.0,
|
| 430 |
+
0.9999617338180542,
|
| 431 |
+
0.997948944568634,
|
| 432 |
+
0.9999988079071045,
|
| 433 |
+
1.0,
|
| 434 |
+
0.9999945163726807,
|
| 435 |
+
0.9999972581863403,
|
| 436 |
+
0.9999998807907104,
|
| 437 |
+
0.9972482323646545,
|
| 438 |
+
0.8824266195297241,
|
| 439 |
+
0.5207855105400085,
|
| 440 |
+
0.9811423420906067,
|
| 441 |
+
0.9948011636734009,
|
| 442 |
+
0.9970309734344482,
|
| 443 |
+
0.9518866539001465,
|
| 444 |
+
0.9435040950775146,
|
| 445 |
+
0.9998762607574463,
|
| 446 |
+
0.9999566078186035,
|
| 447 |
+
0.5264937281608582,
|
| 448 |
+
0.3567848801612854,
|
| 449 |
+
0.2969517111778259,
|
| 450 |
+
0.5564099550247192,
|
| 451 |
+
0.5566191673278809,
|
| 452 |
+
0.6488668918609619,
|
| 453 |
+
0.2498033493757248,
|
| 454 |
+
0.8209301233291626,
|
| 455 |
+
0.9682015776634216,
|
| 456 |
+
0.999983549118042,
|
| 457 |
+
0.998921275138855,
|
| 458 |
+
0.9983420372009277,
|
| 459 |
+
0.9999994039535522,
|
| 460 |
+
1.0,
|
| 461 |
+
1.0,
|
| 462 |
+
0.9042081236839294,
|
| 463 |
+
0.9448812007904053,
|
| 464 |
+
0.8440216779708862,
|
| 465 |
+
0.9969866871833801,
|
| 466 |
+
0.935389518737793,
|
| 467 |
+
0.9999938011169434,
|
| 468 |
+
0.8614246249198914,
|
| 469 |
+
0.9949424862861633,
|
| 470 |
+
0.9876769781112671,
|
| 471 |
+
0.99991774559021,
|
| 472 |
+
0.9980294108390808,
|
| 473 |
+
0.9999949932098389,
|
| 474 |
+
1.0,
|
| 475 |
+
1.0,
|
| 476 |
+
0.9999998807907104,
|
| 477 |
+
0.6444177627563477,
|
| 478 |
+
0.5885342359542847,
|
| 479 |
+
0.6353741884231567,
|
| 480 |
+
0.9999998807907104,
|
| 481 |
+
0.6010631322860718,
|
| 482 |
+
0.999997615814209,
|
| 483 |
+
0.9806910753250122,
|
| 484 |
+
0.9655673503875732,
|
| 485 |
+
0.9658817648887634,
|
| 486 |
+
0.829046368598938,
|
| 487 |
+
0.9939888715744019,
|
| 488 |
+
0.8807443380355835,
|
| 489 |
+
0.9380287528038025,
|
| 490 |
+
0.7914523482322693,
|
| 491 |
+
0.3484261929988861,
|
| 492 |
+
0.6201522350311279,
|
| 493 |
+
0.34111207723617554,
|
| 494 |
+
0.7237049341201782,
|
| 495 |
+
0.6027003526687622,
|
| 496 |
+
0.8865858912467957,
|
| 497 |
+
0.632958710193634,
|
| 498 |
+
1.0,
|
| 499 |
+
0.9925962090492249,
|
| 500 |
+
0.9999537467956543,
|
| 501 |
+
1.0,
|
| 502 |
+
0.999997615814209,
|
| 503 |
+
0.9999998807907104,
|
| 504 |
+
0.6945015788078308,
|
| 505 |
+
0.5074270963668823,
|
| 506 |
+
0.7571401000022888,
|
| 507 |
+
0.9955568909645081,
|
| 508 |
+
0.9997404217720032,
|
| 509 |
+
0.9955176711082458,
|
| 510 |
+
0.7693279385566711,
|
| 511 |
+
0.8310465812683105,
|
| 512 |
+
0.5401557683944702,
|
| 513 |
+
0.9749163389205933,
|
| 514 |
+
0.963936448097229,
|
| 515 |
+
0.9640229940414429,
|
| 516 |
+
0.5689054131507874,
|
| 517 |
+
0.8011016249656677,
|
| 518 |
+
0.6685754060745239,
|
| 519 |
+
0.6061564087867737,
|
| 520 |
+
0.44276222586631775,
|
| 521 |
+
0.4091780185699463,
|
| 522 |
+
0.7819089889526367
|
| 523 |
+
],
|
| 524 |
+
"eval_code_runtime": 105.8373,
|
| 525 |
+
"eval_code_samples_per_second": 0.945,
|
| 526 |
+
"eval_code_score": -0.06509324163198471,
|
| 527 |
+
"eval_code_steps_per_second": 0.038,
|
| 528 |
+
"step": 50
|
| 529 |
+
},
|
| 530 |
+
{
|
| 531 |
+
"epoch": 0.89,
|
| 532 |
+
"learning_rate": 0.0002,
|
| 533 |
+
"loss": 0.181,
|
| 534 |
+
"step": 51
|
| 535 |
+
},
|
| 536 |
+
{
|
| 537 |
+
"epoch": 0.91,
|
| 538 |
+
"learning_rate": 0.0002,
|
| 539 |
+
"loss": 0.2069,
|
| 540 |
+
"step": 52
|
| 541 |
+
},
|
| 542 |
+
{
|
| 543 |
+
"epoch": 0.93,
|
| 544 |
+
"learning_rate": 0.0002,
|
| 545 |
+
"loss": 0.2289,
|
| 546 |
+
"step": 53
|
| 547 |
+
},
|
| 548 |
+
{
|
| 549 |
+
"epoch": 0.95,
|
| 550 |
+
"learning_rate": 0.0002,
|
| 551 |
+
"loss": 0.2987,
|
| 552 |
+
"step": 54
|
| 553 |
+
},
|
| 554 |
+
{
|
| 555 |
+
"epoch": 0.96,
|
| 556 |
+
"learning_rate": 0.0002,
|
| 557 |
+
"loss": 0.3474,
|
| 558 |
+
"step": 55
|
| 559 |
+
},
|
| 560 |
+
{
|
| 561 |
+
"epoch": 0.98,
|
| 562 |
+
"learning_rate": 0.0002,
|
| 563 |
+
"loss": 0.284,
|
| 564 |
+
"step": 56
|
| 565 |
+
},
|
| 566 |
+
{
|
| 567 |
+
"epoch": 1.0,
|
| 568 |
+
"learning_rate": 0.0002,
|
| 569 |
+
"loss": 0.7535,
|
| 570 |
+
"step": 57
|
| 571 |
+
},
|
| 572 |
+
{
|
| 573 |
+
"epoch": 1.02,
|
| 574 |
+
"learning_rate": 0.0002,
|
| 575 |
+
"loss": 0.1286,
|
| 576 |
+
"step": 58
|
| 577 |
+
},
|
| 578 |
+
{
|
| 579 |
+
"epoch": 1.04,
|
| 580 |
+
"learning_rate": 0.0002,
|
| 581 |
+
"loss": 0.0816,
|
| 582 |
+
"step": 59
|
| 583 |
+
},
|
| 584 |
+
{
|
| 585 |
+
"epoch": 1.05,
|
| 586 |
+
"learning_rate": 0.0002,
|
| 587 |
+
"loss": 0.0607,
|
| 588 |
+
"step": 60
|
| 589 |
+
},
|
| 590 |
+
{
|
| 591 |
+
"epoch": 1.07,
|
| 592 |
+
"learning_rate": 0.0002,
|
| 593 |
+
"loss": 0.0772,
|
| 594 |
+
"step": 61
|
| 595 |
+
},
|
| 596 |
+
{
|
| 597 |
+
"epoch": 1.09,
|
| 598 |
+
"learning_rate": 0.0002,
|
| 599 |
+
"loss": 0.1129,
|
| 600 |
+
"step": 62
|
| 601 |
+
},
|
| 602 |
+
{
|
| 603 |
+
"epoch": 1.11,
|
| 604 |
+
"learning_rate": 0.0002,
|
| 605 |
+
"loss": 0.0509,
|
| 606 |
+
"step": 63
|
| 607 |
+
},
|
| 608 |
+
{
|
| 609 |
+
"epoch": 1.12,
|
| 610 |
+
"learning_rate": 0.0002,
|
| 611 |
+
"loss": 0.1928,
|
| 612 |
+
"step": 64
|
| 613 |
+
},
|
| 614 |
+
{
|
| 615 |
+
"epoch": 1.14,
|
| 616 |
+
"learning_rate": 0.0002,
|
| 617 |
+
"loss": 0.0945,
|
| 618 |
+
"step": 65
|
| 619 |
+
},
|
| 620 |
+
{
|
| 621 |
+
"epoch": 1.16,
|
| 622 |
+
"learning_rate": 0.0002,
|
| 623 |
+
"loss": 0.1268,
|
| 624 |
+
"step": 66
|
| 625 |
+
},
|
| 626 |
+
{
|
| 627 |
+
"epoch": 1.18,
|
| 628 |
+
"learning_rate": 0.0002,
|
| 629 |
+
"loss": 0.1186,
|
| 630 |
+
"step": 67
|
| 631 |
+
},
|
| 632 |
+
{
|
| 633 |
+
"epoch": 1.19,
|
| 634 |
+
"learning_rate": 0.0002,
|
| 635 |
+
"loss": 0.1123,
|
| 636 |
+
"step": 68
|
| 637 |
+
},
|
| 638 |
+
{
|
| 639 |
+
"epoch": 1.21,
|
| 640 |
+
"learning_rate": 0.0002,
|
| 641 |
+
"loss": 0.1512,
|
| 642 |
+
"step": 69
|
| 643 |
+
},
|
| 644 |
+
{
|
| 645 |
+
"epoch": 1.23,
|
| 646 |
+
"learning_rate": 0.0002,
|
| 647 |
+
"loss": 0.1303,
|
| 648 |
+
"step": 70
|
| 649 |
+
},
|
| 650 |
+
{
|
| 651 |
+
"epoch": 1.25,
|
| 652 |
+
"learning_rate": 0.0002,
|
| 653 |
+
"loss": 0.0441,
|
| 654 |
+
"step": 71
|
| 655 |
+
},
|
| 656 |
+
{
|
| 657 |
+
"epoch": 1.26,
|
| 658 |
+
"learning_rate": 0.0002,
|
| 659 |
+
"loss": 0.1391,
|
| 660 |
+
"step": 72
|
| 661 |
+
},
|
| 662 |
+
{
|
| 663 |
+
"epoch": 1.28,
|
| 664 |
+
"learning_rate": 0.0002,
|
| 665 |
+
"loss": 0.0685,
|
| 666 |
+
"step": 73
|
| 667 |
+
},
|
| 668 |
+
{
|
| 669 |
+
"epoch": 1.3,
|
| 670 |
+
"learning_rate": 0.0002,
|
| 671 |
+
"loss": 0.0762,
|
| 672 |
+
"step": 74
|
| 673 |
+
},
|
| 674 |
+
{
|
| 675 |
+
"epoch": 1.32,
|
| 676 |
+
"learning_rate": 0.0002,
|
| 677 |
+
"loss": 0.1011,
|
| 678 |
+
"step": 75
|
| 679 |
+
},
|
| 680 |
+
{
|
| 681 |
+
"epoch": 1.32,
|
| 682 |
+
"eval_code_accuracy": 0.91,
|
| 683 |
+
"eval_code_average_probability": 0.889110267162323,
|
| 684 |
+
"eval_code_brier_score": 0.06802475452423096,
|
| 685 |
+
"eval_code_loss": 0.21632270514965057,
|
| 686 |
+
"eval_code_probabilities": [
|
| 687 |
+
0.9983546137809753,
|
| 688 |
+
0.8964507579803467,
|
| 689 |
+
0.9998843669891357,
|
| 690 |
+
1.0,
|
| 691 |
+
0.9999983310699463,
|
| 692 |
+
0.9999896287918091,
|
| 693 |
+
1.0,
|
| 694 |
+
1.0,
|
| 695 |
+
0.9999992847442627,
|
| 696 |
+
1.0,
|
| 697 |
+
1.0,
|
| 698 |
+
1.0,
|
| 699 |
+
1.0,
|
| 700 |
+
1.0,
|
| 701 |
+
1.0,
|
| 702 |
+
0.9493870139122009,
|
| 703 |
+
0.22386565804481506,
|
| 704 |
+
0.9954581260681152,
|
| 705 |
+
0.9999986886978149,
|
| 706 |
+
1.0,
|
| 707 |
+
0.9999638795852661,
|
| 708 |
+
0.9999364614486694,
|
| 709 |
+
1.0,
|
| 710 |
+
1.0,
|
| 711 |
+
0.14285829663276672,
|
| 712 |
+
0.05121216922998428,
|
| 713 |
+
0.06197505444288254,
|
| 714 |
+
0.8036693334579468,
|
| 715 |
+
0.784625768661499,
|
| 716 |
+
0.9341561198234558,
|
| 717 |
+
0.2160544991493225,
|
| 718 |
+
0.9977596998214722,
|
| 719 |
+
0.9990569949150085,
|
| 720 |
+
0.9999992847442627,
|
| 721 |
+
0.5561187863349915,
|
| 722 |
+
0.9838144183158875,
|
| 723 |
+
1.0,
|
| 724 |
+
1.0,
|
| 725 |
+
1.0,
|
| 726 |
+
0.9999961853027344,
|
| 727 |
+
0.999188244342804,
|
| 728 |
+
0.9992071986198425,
|
| 729 |
+
0.9999998807907104,
|
| 730 |
+
0.9616034030914307,
|
| 731 |
+
0.9999998807907104,
|
| 732 |
+
0.9486005902290344,
|
| 733 |
+
0.9999855756759644,
|
| 734 |
+
0.9963693618774414,
|
| 735 |
+
0.9999858140945435,
|
| 736 |
+
0.9972230195999146,
|
| 737 |
+
1.0,
|
| 738 |
+
1.0,
|
| 739 |
+
1.0,
|
| 740 |
+
1.0,
|
| 741 |
+
0.7309120893478394,
|
| 742 |
+
0.8254848122596741,
|
| 743 |
+
0.7734106183052063,
|
| 744 |
+
1.0,
|
| 745 |
+
0.5664382576942444,
|
| 746 |
+
1.0,
|
| 747 |
+
0.9999998807907104,
|
| 748 |
+
0.9999990463256836,
|
| 749 |
+
0.9999992847442627,
|
| 750 |
+
0.9987679123878479,
|
| 751 |
+
0.9999841451644897,
|
| 752 |
+
0.9796060919761658,
|
| 753 |
+
0.9999995231628418,
|
| 754 |
+
0.4220101237297058,
|
| 755 |
+
0.517703652381897,
|
| 756 |
+
0.8652507066726685,
|
| 757 |
+
0.007634077221155167,
|
| 758 |
+
0.9999922513961792,
|
| 759 |
+
0.9957844614982605,
|
| 760 |
+
0.9999879598617554,
|
| 761 |
+
0.9737687706947327,
|
| 762 |
+
1.0,
|
| 763 |
+
0.9995587468147278,
|
| 764 |
+
1.0,
|
| 765 |
+
1.0,
|
| 766 |
+
1.0,
|
| 767 |
+
1.0,
|
| 768 |
+
0.9666233658790588,
|
| 769 |
+
0.6209402084350586,
|
| 770 |
+
0.978766143321991,
|
| 771 |
+
0.9995095729827881,
|
| 772 |
+
0.9826872944831848,
|
| 773 |
+
1.0,
|
| 774 |
+
0.9902674555778503,
|
| 775 |
+
0.9968993663787842,
|
| 776 |
+
0.8134205937385559,
|
| 777 |
+
0.9999436140060425,
|
| 778 |
+
0.9293919205665588,
|
| 779 |
+
0.9999128580093384,
|
| 780 |
+
0.8822279572486877,
|
| 781 |
+
0.9978179931640625,
|
| 782 |
+
0.9373266100883484,
|
| 783 |
+
0.7644922733306885,
|
| 784 |
+
0.43501412868499756,
|
| 785 |
+
0.4749135673046112,
|
| 786 |
+
0.9878302216529846
|
| 787 |
+
],
|
| 788 |
+
"eval_code_runtime": 105.7625,
|
| 789 |
+
"eval_code_samples_per_second": 0.946,
|
| 790 |
+
"eval_code_score": -0.06802475452423096,
|
| 791 |
+
"eval_code_steps_per_second": 0.038,
|
| 792 |
+
"step": 75
|
| 793 |
+
},
|
| 794 |
+
{
|
| 795 |
+
"epoch": 1.33,
|
| 796 |
+
"learning_rate": 0.0002,
|
| 797 |
+
"loss": 0.0665,
|
| 798 |
+
"step": 76
|
| 799 |
+
},
|
| 800 |
+
{
|
| 801 |
+
"epoch": 1.35,
|
| 802 |
+
"learning_rate": 0.0002,
|
| 803 |
+
"loss": 0.0808,
|
| 804 |
+
"step": 77
|
| 805 |
+
},
|
| 806 |
+
{
|
| 807 |
+
"epoch": 1.37,
|
| 808 |
+
"learning_rate": 0.0002,
|
| 809 |
+
"loss": 0.0637,
|
| 810 |
+
"step": 78
|
| 811 |
+
},
|
| 812 |
+
{
|
| 813 |
+
"epoch": 1.39,
|
| 814 |
+
"learning_rate": 0.0002,
|
| 815 |
+
"loss": 0.0323,
|
| 816 |
+
"step": 79
|
| 817 |
+
},
|
| 818 |
+
{
|
| 819 |
+
"epoch": 1.4,
|
| 820 |
+
"learning_rate": 0.0002,
|
| 821 |
+
"loss": 0.0441,
|
| 822 |
+
"step": 80
|
| 823 |
+
},
|
| 824 |
+
{
|
| 825 |
+
"epoch": 1.42,
|
| 826 |
+
"learning_rate": 0.0002,
|
| 827 |
+
"loss": 0.1698,
|
| 828 |
+
"step": 81
|
| 829 |
+
},
|
| 830 |
+
{
|
| 831 |
+
"epoch": 1.44,
|
| 832 |
+
"learning_rate": 0.0002,
|
| 833 |
+
"loss": 0.156,
|
| 834 |
+
"step": 82
|
| 835 |
+
},
|
| 836 |
+
{
|
| 837 |
+
"epoch": 1.46,
|
| 838 |
+
"learning_rate": 0.0002,
|
| 839 |
+
"loss": 0.3788,
|
| 840 |
+
"step": 83
|
| 841 |
+
},
|
| 842 |
+
{
|
| 843 |
+
"epoch": 1.47,
|
| 844 |
+
"learning_rate": 0.0002,
|
| 845 |
+
"loss": 0.1222,
|
| 846 |
+
"step": 84
|
| 847 |
+
},
|
| 848 |
+
{
|
| 849 |
+
"epoch": 1.49,
|
| 850 |
+
"learning_rate": 0.0002,
|
| 851 |
+
"loss": 0.1639,
|
| 852 |
+
"step": 85
|
| 853 |
+
},
|
| 854 |
+
{
|
| 855 |
+
"epoch": 1.51,
|
| 856 |
+
"learning_rate": 0.0002,
|
| 857 |
+
"loss": 0.0538,
|
| 858 |
+
"step": 86
|
| 859 |
+
},
|
| 860 |
+
{
|
| 861 |
+
"epoch": 1.53,
|
| 862 |
+
"learning_rate": 0.0002,
|
| 863 |
+
"loss": 0.0328,
|
| 864 |
+
"step": 87
|
| 865 |
+
},
|
| 866 |
+
{
|
| 867 |
+
"epoch": 1.54,
|
| 868 |
+
"learning_rate": 0.0002,
|
| 869 |
+
"loss": 0.1745,
|
| 870 |
+
"step": 88
|
| 871 |
+
},
|
| 872 |
+
{
|
| 873 |
+
"epoch": 1.56,
|
| 874 |
+
"learning_rate": 0.0002,
|
| 875 |
+
"loss": 0.0874,
|
| 876 |
+
"step": 89
|
| 877 |
+
},
|
| 878 |
+
{
|
| 879 |
+
"epoch": 1.58,
|
| 880 |
+
"learning_rate": 0.0002,
|
| 881 |
+
"loss": 0.1177,
|
| 882 |
+
"step": 90
|
| 883 |
+
},
|
| 884 |
+
{
|
| 885 |
+
"epoch": 1.6,
|
| 886 |
+
"learning_rate": 0.0002,
|
| 887 |
+
"loss": 0.1185,
|
| 888 |
+
"step": 91
|
| 889 |
+
},
|
| 890 |
+
{
|
| 891 |
+
"epoch": 1.61,
|
| 892 |
+
"learning_rate": 0.0002,
|
| 893 |
+
"loss": 0.0543,
|
| 894 |
+
"step": 92
|
| 895 |
+
},
|
| 896 |
+
{
|
| 897 |
+
"epoch": 1.63,
|
| 898 |
+
"learning_rate": 0.0002,
|
| 899 |
+
"loss": 0.0392,
|
| 900 |
+
"step": 93
|
| 901 |
+
},
|
| 902 |
+
{
|
| 903 |
+
"epoch": 1.65,
|
| 904 |
+
"learning_rate": 0.0002,
|
| 905 |
+
"loss": 0.2006,
|
| 906 |
+
"step": 94
|
| 907 |
+
},
|
| 908 |
+
{
|
| 909 |
+
"epoch": 1.67,
|
| 910 |
+
"learning_rate": 0.0002,
|
| 911 |
+
"loss": 0.1249,
|
| 912 |
+
"step": 95
|
| 913 |
+
},
|
| 914 |
+
{
|
| 915 |
+
"epoch": 1.68,
|
| 916 |
+
"learning_rate": 0.0002,
|
| 917 |
+
"loss": 0.0341,
|
| 918 |
+
"step": 96
|
| 919 |
+
},
|
| 920 |
+
{
|
| 921 |
+
"epoch": 1.7,
|
| 922 |
+
"learning_rate": 0.0002,
|
| 923 |
+
"loss": 0.0666,
|
| 924 |
+
"step": 97
|
| 925 |
+
},
|
| 926 |
+
{
|
| 927 |
+
"epoch": 1.72,
|
| 928 |
+
"learning_rate": 0.0002,
|
| 929 |
+
"loss": 0.1614,
|
| 930 |
+
"step": 98
|
| 931 |
+
},
|
| 932 |
+
{
|
| 933 |
+
"epoch": 1.74,
|
| 934 |
+
"learning_rate": 0.0002,
|
| 935 |
+
"loss": 0.0831,
|
| 936 |
+
"step": 99
|
| 937 |
+
},
|
| 938 |
+
{
|
| 939 |
+
"epoch": 1.75,
|
| 940 |
+
"learning_rate": 0.0002,
|
| 941 |
+
"loss": 0.0273,
|
| 942 |
+
"step": 100
|
| 943 |
+
},
|
| 944 |
+
{
|
| 945 |
+
"epoch": 1.75,
|
| 946 |
+
"eval_code_accuracy": 0.96,
|
| 947 |
+
"eval_code_average_probability": 0.9304368495941162,
|
| 948 |
+
"eval_code_brier_score": 0.03531455993652344,
|
| 949 |
+
"eval_code_loss": 0.10408665984869003,
|
| 950 |
+
"eval_code_probabilities": [
|
| 951 |
+
1.0,
|
| 952 |
+
1.0,
|
| 953 |
+
1.0,
|
| 954 |
+
1.0,
|
| 955 |
+
0.9999995231628418,
|
| 956 |
+
1.0,
|
| 957 |
+
1.0,
|
| 958 |
+
0.9999160766601562,
|
| 959 |
+
0.9988158941268921,
|
| 960 |
+
0.9999899864196777,
|
| 961 |
+
0.9999986886978149,
|
| 962 |
+
0.9999879598617554,
|
| 963 |
+
0.9999833106994629,
|
| 964 |
+
0.9999370574951172,
|
| 965 |
+
0.9992890357971191,
|
| 966 |
+
0.9659052491188049,
|
| 967 |
+
0.6254108548164368,
|
| 968 |
+
0.997655987739563,
|
| 969 |
+
0.9994457364082336,
|
| 970 |
+
0.999336302280426,
|
| 971 |
+
0.9953096508979797,
|
| 972 |
+
0.9922773838043213,
|
| 973 |
+
0.9999829530715942,
|
| 974 |
+
0.9999885559082031,
|
| 975 |
+
0.9871056079864502,
|
| 976 |
+
0.9849441647529602,
|
| 977 |
+
0.9815734028816223,
|
| 978 |
+
0.7201115489006042,
|
| 979 |
+
0.6514120101928711,
|
| 980 |
+
0.8476458191871643,
|
| 981 |
+
0.1292794942855835,
|
| 982 |
+
0.9985151886940002,
|
| 983 |
+
0.9995579123497009,
|
| 984 |
+
1.0,
|
| 985 |
+
0.9999856948852539,
|
| 986 |
+
0.9999977350234985,
|
| 987 |
+
0.9999998807907104,
|
| 988 |
+
1.0,
|
| 989 |
+
1.0,
|
| 990 |
+
0.9999991655349731,
|
| 991 |
+
0.9998373985290527,
|
| 992 |
+
0.999966025352478,
|
| 993 |
+
1.0,
|
| 994 |
+
0.9999432563781738,
|
| 995 |
+
1.0,
|
| 996 |
+
0.997358500957489,
|
| 997 |
+
1.0,
|
| 998 |
+
0.9999942779541016,
|
| 999 |
+
0.9999959468841553,
|
| 1000 |
+
0.9998447895050049,
|
| 1001 |
+
1.0,
|
| 1002 |
+
1.0,
|
| 1003 |
+
1.0,
|
| 1004 |
+
1.0,
|
| 1005 |
+
0.9860179424285889,
|
| 1006 |
+
0.8810455203056335,
|
| 1007 |
+
0.8414822816848755,
|
| 1008 |
+
1.0,
|
| 1009 |
+
0.9645673632621765,
|
| 1010 |
+
1.0,
|
| 1011 |
+
1.0,
|
| 1012 |
+
1.0,
|
| 1013 |
+
1.0,
|
| 1014 |
+
0.9987167119979858,
|
| 1015 |
+
0.9995377063751221,
|
| 1016 |
+
0.969241201877594,
|
| 1017 |
+
0.9999886751174927,
|
| 1018 |
+
0.2950672507286072,
|
| 1019 |
+
0.44002464413642883,
|
| 1020 |
+
0.8467398881912231,
|
| 1021 |
+
0.06261960417032242,
|
| 1022 |
+
0.9991006851196289,
|
| 1023 |
+
0.9985753297805786,
|
| 1024 |
+
0.9999589920043945,
|
| 1025 |
+
0.9881368279457092,
|
| 1026 |
+
1.0,
|
| 1027 |
+
0.9875013828277588,
|
| 1028 |
+
0.9999996423721313,
|
| 1029 |
+
1.0,
|
| 1030 |
+
0.9999927282333374,
|
| 1031 |
+
0.9999997615814209,
|
| 1032 |
+
0.9641613960266113,
|
| 1033 |
+
0.5904451012611389,
|
| 1034 |
+
0.991640567779541,
|
| 1035 |
+
0.9999064207077026,
|
| 1036 |
+
0.996203601360321,
|
| 1037 |
+
0.999998927116394,
|
| 1038 |
+
0.9644280672073364,
|
| 1039 |
+
0.9944401979446411,
|
| 1040 |
+
0.8003358840942383,
|
| 1041 |
+
0.9963403940200806,
|
| 1042 |
+
0.9804515242576599,
|
| 1043 |
+
0.989639163017273,
|
| 1044 |
+
0.852015495300293,
|
| 1045 |
+
0.9918712377548218,
|
| 1046 |
+
0.9663655757904053,
|
| 1047 |
+
0.8847030997276306,
|
| 1048 |
+
0.5973413586616516,
|
| 1049 |
+
0.5435407757759094,
|
| 1050 |
+
0.811240017414093
|
| 1051 |
+
],
|
| 1052 |
+
"eval_code_runtime": 105.7613,
|
| 1053 |
+
"eval_code_samples_per_second": 0.946,
|
| 1054 |
+
"eval_code_score": -0.03531455993652344,
|
| 1055 |
+
"eval_code_steps_per_second": 0.038,
|
| 1056 |
+
"step": 100
|
| 1057 |
+
},
|
| 1058 |
+
{
|
| 1059 |
+
"epoch": 1.77,
|
| 1060 |
+
"learning_rate": 0.0002,
|
| 1061 |
+
"loss": 0.0341,
|
| 1062 |
+
"step": 101
|
| 1063 |
+
},
|
| 1064 |
+
{
|
| 1065 |
+
"epoch": 1.79,
|
| 1066 |
+
"learning_rate": 0.0002,
|
| 1067 |
+
"loss": 0.0952,
|
| 1068 |
+
"step": 102
|
| 1069 |
+
},
|
| 1070 |
+
{
|
| 1071 |
+
"epoch": 1.81,
|
| 1072 |
+
"learning_rate": 0.0002,
|
| 1073 |
+
"loss": 0.0358,
|
| 1074 |
+
"step": 103
|
| 1075 |
+
},
|
| 1076 |
+
{
|
| 1077 |
+
"epoch": 1.82,
|
| 1078 |
+
"learning_rate": 0.0002,
|
| 1079 |
+
"loss": 0.1143,
|
| 1080 |
+
"step": 104
|
| 1081 |
+
},
|
| 1082 |
+
{
|
| 1083 |
+
"epoch": 1.84,
|
| 1084 |
+
"learning_rate": 0.0002,
|
| 1085 |
+
"loss": 0.07,
|
| 1086 |
+
"step": 105
|
| 1087 |
+
},
|
| 1088 |
+
{
|
| 1089 |
+
"epoch": 1.86,
|
| 1090 |
+
"learning_rate": 0.0002,
|
| 1091 |
+
"loss": 0.1232,
|
| 1092 |
+
"step": 106
|
| 1093 |
+
},
|
| 1094 |
+
{
|
| 1095 |
+
"epoch": 1.88,
|
| 1096 |
+
"learning_rate": 0.0002,
|
| 1097 |
+
"loss": 0.065,
|
| 1098 |
+
"step": 107
|
| 1099 |
+
},
|
| 1100 |
+
{
|
| 1101 |
+
"epoch": 1.89,
|
| 1102 |
+
"learning_rate": 0.0002,
|
| 1103 |
+
"loss": 0.0167,
|
| 1104 |
+
"step": 108
|
| 1105 |
+
},
|
| 1106 |
+
{
|
| 1107 |
+
"epoch": 1.91,
|
| 1108 |
+
"learning_rate": 0.0002,
|
| 1109 |
+
"loss": 0.0486,
|
| 1110 |
+
"step": 109
|
| 1111 |
+
},
|
| 1112 |
+
{
|
| 1113 |
+
"epoch": 1.93,
|
| 1114 |
+
"learning_rate": 0.0002,
|
| 1115 |
+
"loss": 0.011,
|
| 1116 |
+
"step": 110
|
| 1117 |
+
},
|
| 1118 |
+
{
|
| 1119 |
+
"epoch": 1.95,
|
| 1120 |
+
"learning_rate": 0.0002,
|
| 1121 |
+
"loss": 0.0425,
|
| 1122 |
+
"step": 111
|
| 1123 |
+
},
|
| 1124 |
+
{
|
| 1125 |
+
"epoch": 1.96,
|
| 1126 |
+
"learning_rate": 0.0002,
|
| 1127 |
+
"loss": 0.1082,
|
| 1128 |
+
"step": 112
|
| 1129 |
+
},
|
| 1130 |
+
{
|
| 1131 |
+
"epoch": 1.98,
|
| 1132 |
+
"learning_rate": 0.0002,
|
| 1133 |
+
"loss": 0.0196,
|
| 1134 |
+
"step": 113
|
| 1135 |
+
},
|
| 1136 |
+
{
|
| 1137 |
+
"epoch": 2.0,
|
| 1138 |
+
"learning_rate": 0.0002,
|
| 1139 |
+
"loss": 0.0152,
|
| 1140 |
+
"step": 114
|
| 1141 |
+
},
|
| 1142 |
+
{
|
| 1143 |
+
"epoch": 2.02,
|
| 1144 |
+
"learning_rate": 0.0002,
|
| 1145 |
+
"loss": 0.012,
|
| 1146 |
+
"step": 115
|
| 1147 |
+
},
|
| 1148 |
+
{
|
| 1149 |
+
"epoch": 2.04,
|
| 1150 |
+
"learning_rate": 0.0002,
|
| 1151 |
+
"loss": 0.0186,
|
| 1152 |
+
"step": 116
|
| 1153 |
+
},
|
| 1154 |
+
{
|
| 1155 |
+
"epoch": 2.05,
|
| 1156 |
+
"learning_rate": 0.0002,
|
| 1157 |
+
"loss": 0.0186,
|
| 1158 |
+
"step": 117
|
| 1159 |
+
},
|
| 1160 |
+
{
|
| 1161 |
+
"epoch": 2.07,
|
| 1162 |
+
"learning_rate": 0.0002,
|
| 1163 |
+
"loss": 0.0238,
|
| 1164 |
+
"step": 118
|
| 1165 |
+
},
|
| 1166 |
+
{
|
| 1167 |
+
"epoch": 2.09,
|
| 1168 |
+
"learning_rate": 0.0002,
|
| 1169 |
+
"loss": 0.0097,
|
| 1170 |
+
"step": 119
|
| 1171 |
+
},
|
| 1172 |
+
{
|
| 1173 |
+
"epoch": 2.11,
|
| 1174 |
+
"learning_rate": 0.0002,
|
| 1175 |
+
"loss": 0.0007,
|
| 1176 |
+
"step": 120
|
| 1177 |
+
},
|
| 1178 |
+
{
|
| 1179 |
+
"epoch": 2.12,
|
| 1180 |
+
"learning_rate": 0.0002,
|
| 1181 |
+
"loss": 0.0409,
|
| 1182 |
+
"step": 121
|
| 1183 |
+
},
|
| 1184 |
+
{
|
| 1185 |
+
"epoch": 2.14,
|
| 1186 |
+
"learning_rate": 0.0002,
|
| 1187 |
+
"loss": 0.0238,
|
| 1188 |
+
"step": 122
|
| 1189 |
+
},
|
| 1190 |
+
{
|
| 1191 |
+
"epoch": 2.16,
|
| 1192 |
+
"learning_rate": 0.0002,
|
| 1193 |
+
"loss": 0.0045,
|
| 1194 |
+
"step": 123
|
| 1195 |
+
},
|
| 1196 |
+
{
|
| 1197 |
+
"epoch": 2.18,
|
| 1198 |
+
"learning_rate": 0.0002,
|
| 1199 |
+
"loss": 0.0008,
|
| 1200 |
+
"step": 124
|
| 1201 |
+
},
|
| 1202 |
+
{
|
| 1203 |
+
"epoch": 2.19,
|
| 1204 |
+
"learning_rate": 0.0002,
|
| 1205 |
+
"loss": 0.0307,
|
| 1206 |
+
"step": 125
|
| 1207 |
+
},
|
| 1208 |
+
{
|
| 1209 |
+
"epoch": 2.19,
|
| 1210 |
+
"eval_code_accuracy": 0.96,
|
| 1211 |
+
"eval_code_average_probability": 0.9499993324279785,
|
| 1212 |
+
"eval_code_brier_score": 0.028195565566420555,
|
| 1213 |
+
"eval_code_loss": 0.11333052814006805,
|
| 1214 |
+
"eval_code_probabilities": [
|
| 1215 |
+
1.0,
|
| 1216 |
+
1.0,
|
| 1217 |
+
1.0,
|
| 1218 |
+
1.0,
|
| 1219 |
+
1.0,
|
| 1220 |
+
1.0,
|
| 1221 |
+
1.0,
|
| 1222 |
+
1.0,
|
| 1223 |
+
0.9999997615814209,
|
| 1224 |
+
1.0,
|
| 1225 |
+
1.0,
|
| 1226 |
+
0.9999997615814209,
|
| 1227 |
+
0.9999997615814209,
|
| 1228 |
+
0.9999985694885254,
|
| 1229 |
+
0.9999561309814453,
|
| 1230 |
+
0.8770335912704468,
|
| 1231 |
+
0.4641159176826477,
|
| 1232 |
+
0.9271451234817505,
|
| 1233 |
+
1.0,
|
| 1234 |
+
1.0,
|
| 1235 |
+
0.9999996423721313,
|
| 1236 |
+
0.9998699426651001,
|
| 1237 |
+
0.9999997615814209,
|
| 1238 |
+
0.9999998807907104,
|
| 1239 |
+
0.9999991655349731,
|
| 1240 |
+
0.9999380111694336,
|
| 1241 |
+
0.9999819993972778,
|
| 1242 |
+
0.7533411979675293,
|
| 1243 |
+
0.7846699357032776,
|
| 1244 |
+
0.89438796043396,
|
| 1245 |
+
0.015444471500813961,
|
| 1246 |
+
0.999987006187439,
|
| 1247 |
+
1.0,
|
| 1248 |
+
1.0,
|
| 1249 |
+
0.9969304203987122,
|
| 1250 |
+
0.9999985694885254,
|
| 1251 |
+
0.9999998807907104,
|
| 1252 |
+
1.0,
|
| 1253 |
+
1.0,
|
| 1254 |
+
1.0,
|
| 1255 |
+
0.9999716281890869,
|
| 1256 |
+
0.9999990463256836,
|
| 1257 |
+
1.0,
|
| 1258 |
+
1.0,
|
| 1259 |
+
1.0,
|
| 1260 |
+
0.9998655319213867,
|
| 1261 |
+
1.0,
|
| 1262 |
+
0.9999997615814209,
|
| 1263 |
+
0.9999998807907104,
|
| 1264 |
+
0.9999963045120239,
|
| 1265 |
+
1.0,
|
| 1266 |
+
1.0,
|
| 1267 |
+
1.0,
|
| 1268 |
+
1.0,
|
| 1269 |
+
0.9978950023651123,
|
| 1270 |
+
0.7412748336791992,
|
| 1271 |
+
0.8329392671585083,
|
| 1272 |
+
1.0,
|
| 1273 |
+
0.999669075012207,
|
| 1274 |
+
1.0,
|
| 1275 |
+
1.0,
|
| 1276 |
+
1.0,
|
| 1277 |
+
1.0,
|
| 1278 |
+
0.9999936819076538,
|
| 1279 |
+
0.9999263286590576,
|
| 1280 |
+
0.9996693134307861,
|
| 1281 |
+
1.0,
|
| 1282 |
+
0.9178900122642517,
|
| 1283 |
+
0.8888350129127502,
|
| 1284 |
+
0.9699975848197937,
|
| 1285 |
+
0.014796692878007889,
|
| 1286 |
+
0.9991298317909241,
|
| 1287 |
+
0.9968131184577942,
|
| 1288 |
+
0.9999946355819702,
|
| 1289 |
+
0.9282417297363281,
|
| 1290 |
+
1.0,
|
| 1291 |
+
0.9971838593482971,
|
| 1292 |
+
1.0,
|
| 1293 |
+
1.0,
|
| 1294 |
+
0.9999544620513916,
|
| 1295 |
+
0.9999992847442627,
|
| 1296 |
+
0.9987694621086121,
|
| 1297 |
+
0.8008559346199036,
|
| 1298 |
+
0.9995920062065125,
|
| 1299 |
+
1.0,
|
| 1300 |
+
0.9999116659164429,
|
| 1301 |
+
1.0,
|
| 1302 |
+
0.9975823163986206,
|
| 1303 |
+
0.9999178647994995,
|
| 1304 |
+
0.9812232255935669,
|
| 1305 |
+
0.9962044358253479,
|
| 1306 |
+
0.9992068409919739,
|
| 1307 |
+
0.9994266033172607,
|
| 1308 |
+
0.9672200679779053,
|
| 1309 |
+
0.998022198677063,
|
| 1310 |
+
0.9971696734428406,
|
| 1311 |
+
0.9978540539741516,
|
| 1312 |
+
0.9273316860198975,
|
| 1313 |
+
0.8601265549659729,
|
| 1314 |
+
0.4826805889606476
|
| 1315 |
+
],
|
| 1316 |
+
"eval_code_runtime": 105.781,
|
| 1317 |
+
"eval_code_samples_per_second": 0.945,
|
| 1318 |
+
"eval_code_score": -0.028195565566420555,
|
| 1319 |
+
"eval_code_steps_per_second": 0.038,
|
| 1320 |
+
"step": 125
|
| 1321 |
+
},
|
| 1322 |
+
{
|
| 1323 |
+
"epoch": 2.21,
|
| 1324 |
+
"learning_rate": 0.0002,
|
| 1325 |
+
"loss": 0.0048,
|
| 1326 |
+
"step": 126
|
| 1327 |
+
},
|
| 1328 |
+
{
|
| 1329 |
+
"epoch": 2.23,
|
| 1330 |
+
"learning_rate": 0.0002,
|
| 1331 |
+
"loss": 0.0093,
|
| 1332 |
+
"step": 127
|
| 1333 |
+
},
|
| 1334 |
+
{
|
| 1335 |
+
"epoch": 2.25,
|
| 1336 |
+
"learning_rate": 0.0002,
|
| 1337 |
+
"loss": 0.0056,
|
| 1338 |
+
"step": 128
|
| 1339 |
+
},
|
| 1340 |
+
{
|
| 1341 |
+
"epoch": 2.26,
|
| 1342 |
+
"learning_rate": 0.0002,
|
| 1343 |
+
"loss": 0.0123,
|
| 1344 |
+
"step": 129
|
| 1345 |
+
},
|
| 1346 |
+
{
|
| 1347 |
+
"epoch": 2.28,
|
| 1348 |
+
"learning_rate": 0.0002,
|
| 1349 |
+
"loss": 0.0825,
|
| 1350 |
+
"step": 130
|
| 1351 |
+
},
|
| 1352 |
+
{
|
| 1353 |
+
"epoch": 2.3,
|
| 1354 |
+
"learning_rate": 0.0002,
|
| 1355 |
+
"loss": 0.0199,
|
| 1356 |
+
"step": 131
|
| 1357 |
+
},
|
| 1358 |
+
{
|
| 1359 |
+
"epoch": 2.32,
|
| 1360 |
+
"learning_rate": 0.0002,
|
| 1361 |
+
"loss": 0.025,
|
| 1362 |
+
"step": 132
|
| 1363 |
+
},
|
| 1364 |
+
{
|
| 1365 |
+
"epoch": 2.33,
|
| 1366 |
+
"learning_rate": 0.0002,
|
| 1367 |
+
"loss": 0.0477,
|
| 1368 |
+
"step": 133
|
| 1369 |
+
},
|
| 1370 |
+
{
|
| 1371 |
+
"epoch": 2.35,
|
| 1372 |
+
"learning_rate": 0.0002,
|
| 1373 |
+
"loss": 0.0403,
|
| 1374 |
+
"step": 134
|
| 1375 |
+
},
|
| 1376 |
+
{
|
| 1377 |
+
"epoch": 2.37,
|
| 1378 |
+
"learning_rate": 0.0002,
|
| 1379 |
+
"loss": 0.0209,
|
| 1380 |
+
"step": 135
|
| 1381 |
+
},
|
| 1382 |
+
{
|
| 1383 |
+
"epoch": 2.39,
|
| 1384 |
+
"learning_rate": 0.0002,
|
| 1385 |
+
"loss": 0.0004,
|
| 1386 |
+
"step": 136
|
| 1387 |
+
},
|
| 1388 |
+
{
|
| 1389 |
+
"epoch": 2.4,
|
| 1390 |
+
"learning_rate": 0.0002,
|
| 1391 |
+
"loss": 0.0003,
|
| 1392 |
+
"step": 137
|
| 1393 |
+
},
|
| 1394 |
+
{
|
| 1395 |
+
"epoch": 2.42,
|
| 1396 |
+
"learning_rate": 0.0002,
|
| 1397 |
+
"loss": 0.024,
|
| 1398 |
+
"step": 138
|
| 1399 |
+
},
|
| 1400 |
+
{
|
| 1401 |
+
"epoch": 2.44,
|
| 1402 |
+
"learning_rate": 0.0002,
|
| 1403 |
+
"loss": 0.0215,
|
| 1404 |
+
"step": 139
|
| 1405 |
+
},
|
| 1406 |
+
{
|
| 1407 |
+
"epoch": 2.46,
|
| 1408 |
+
"learning_rate": 0.0002,
|
| 1409 |
+
"loss": 0.0321,
|
| 1410 |
+
"step": 140
|
| 1411 |
+
},
|
| 1412 |
+
{
|
| 1413 |
+
"epoch": 2.47,
|
| 1414 |
+
"learning_rate": 0.0002,
|
| 1415 |
+
"loss": 0.0067,
|
| 1416 |
+
"step": 141
|
| 1417 |
+
},
|
| 1418 |
+
{
|
| 1419 |
+
"epoch": 2.49,
|
| 1420 |
+
"learning_rate": 0.0002,
|
| 1421 |
+
"loss": 0.0144,
|
| 1422 |
+
"step": 142
|
| 1423 |
+
},
|
| 1424 |
+
{
|
| 1425 |
+
"epoch": 2.51,
|
| 1426 |
+
"learning_rate": 0.0002,
|
| 1427 |
+
"loss": 0.0017,
|
| 1428 |
+
"step": 143
|
| 1429 |
+
},
|
| 1430 |
+
{
|
| 1431 |
+
"epoch": 2.53,
|
| 1432 |
+
"learning_rate": 0.0002,
|
| 1433 |
+
"loss": 0.025,
|
| 1434 |
+
"step": 144
|
| 1435 |
+
},
|
| 1436 |
+
{
|
| 1437 |
+
"epoch": 2.54,
|
| 1438 |
+
"learning_rate": 0.0002,
|
| 1439 |
+
"loss": 0.0605,
|
| 1440 |
+
"step": 145
|
| 1441 |
+
},
|
| 1442 |
+
{
|
| 1443 |
+
"epoch": 2.56,
|
| 1444 |
+
"learning_rate": 0.0002,
|
| 1445 |
+
"loss": 0.0154,
|
| 1446 |
+
"step": 146
|
| 1447 |
+
},
|
| 1448 |
+
{
|
| 1449 |
+
"epoch": 2.58,
|
| 1450 |
+
"learning_rate": 0.0002,
|
| 1451 |
+
"loss": 0.002,
|
| 1452 |
+
"step": 147
|
| 1453 |
+
},
|
| 1454 |
+
{
|
| 1455 |
+
"epoch": 2.6,
|
| 1456 |
+
"learning_rate": 0.0002,
|
| 1457 |
+
"loss": 0.0012,
|
| 1458 |
+
"step": 148
|
| 1459 |
+
},
|
| 1460 |
+
{
|
| 1461 |
+
"epoch": 2.61,
|
| 1462 |
+
"learning_rate": 0.0002,
|
| 1463 |
+
"loss": 0.0066,
|
| 1464 |
+
"step": 149
|
| 1465 |
+
},
|
| 1466 |
+
{
|
| 1467 |
+
"epoch": 2.63,
|
| 1468 |
+
"learning_rate": 0.0002,
|
| 1469 |
+
"loss": 0.0065,
|
| 1470 |
+
"step": 150
|
| 1471 |
+
},
|
| 1472 |
+
{
|
| 1473 |
+
"epoch": 2.63,
|
| 1474 |
+
"eval_code_accuracy": 0.95,
|
| 1475 |
+
"eval_code_average_probability": 0.9435591101646423,
|
| 1476 |
+
"eval_code_brier_score": 0.03846995532512665,
|
| 1477 |
+
"eval_code_loss": 0.17103993892669678,
|
| 1478 |
+
"eval_code_probabilities": [
|
| 1479 |
+
1.0,
|
| 1480 |
+
1.0,
|
| 1481 |
+
1.0,
|
| 1482 |
+
1.0,
|
| 1483 |
+
1.0,
|
| 1484 |
+
1.0,
|
| 1485 |
+
1.0,
|
| 1486 |
+
1.0,
|
| 1487 |
+
1.0,
|
| 1488 |
+
1.0,
|
| 1489 |
+
1.0,
|
| 1490 |
+
1.0,
|
| 1491 |
+
1.0,
|
| 1492 |
+
1.0,
|
| 1493 |
+
0.9999970197677612,
|
| 1494 |
+
0.9055734872817993,
|
| 1495 |
+
0.42537397146224976,
|
| 1496 |
+
0.9499343633651733,
|
| 1497 |
+
1.0,
|
| 1498 |
+
1.0,
|
| 1499 |
+
1.0,
|
| 1500 |
+
0.9999983310699463,
|
| 1501 |
+
1.0,
|
| 1502 |
+
1.0,
|
| 1503 |
+
1.0,
|
| 1504 |
+
0.9999964237213135,
|
| 1505 |
+
0.9999994039535522,
|
| 1506 |
+
0.8147205710411072,
|
| 1507 |
+
0.8424564599990845,
|
| 1508 |
+
0.9474385976791382,
|
| 1509 |
+
0.003861561883240938,
|
| 1510 |
+
1.0,
|
| 1511 |
+
1.0,
|
| 1512 |
+
1.0,
|
| 1513 |
+
0.035812485963106155,
|
| 1514 |
+
0.9962039589881897,
|
| 1515 |
+
1.0,
|
| 1516 |
+
1.0,
|
| 1517 |
+
1.0,
|
| 1518 |
+
1.0,
|
| 1519 |
+
0.9999990463256836,
|
| 1520 |
+
0.9999998807907104,
|
| 1521 |
+
1.0,
|
| 1522 |
+
1.0,
|
| 1523 |
+
1.0,
|
| 1524 |
+
0.9999754428863525,
|
| 1525 |
+
1.0,
|
| 1526 |
+
1.0,
|
| 1527 |
+
1.0,
|
| 1528 |
+
0.9999997615814209,
|
| 1529 |
+
1.0,
|
| 1530 |
+
1.0,
|
| 1531 |
+
1.0,
|
| 1532 |
+
1.0,
|
| 1533 |
+
0.9926579594612122,
|
| 1534 |
+
0.7481365203857422,
|
| 1535 |
+
0.9441766142845154,
|
| 1536 |
+
1.0,
|
| 1537 |
+
0.9999790191650391,
|
| 1538 |
+
1.0,
|
| 1539 |
+
1.0,
|
| 1540 |
+
1.0,
|
| 1541 |
+
1.0,
|
| 1542 |
+
0.9999998807907104,
|
| 1543 |
+
0.9999998807907104,
|
| 1544 |
+
0.9856025576591492,
|
| 1545 |
+
1.0,
|
| 1546 |
+
0.9969327449798584,
|
| 1547 |
+
0.9738338589668274,
|
| 1548 |
+
0.9861930012702942,
|
| 1549 |
+
0.007046875543892384,
|
| 1550 |
+
0.9993067979812622,
|
| 1551 |
+
0.9975935816764832,
|
| 1552 |
+
0.9999728202819824,
|
| 1553 |
+
0.3661898970603943,
|
| 1554 |
+
1.0,
|
| 1555 |
+
0.9999417066574097,
|
| 1556 |
+
1.0,
|
| 1557 |
+
1.0,
|
| 1558 |
+
0.9999967813491821,
|
| 1559 |
+
1.0,
|
| 1560 |
+
0.9991926550865173,
|
| 1561 |
+
0.8271266222000122,
|
| 1562 |
+
0.9992584586143494,
|
| 1563 |
+
1.0,
|
| 1564 |
+
0.9992876648902893,
|
| 1565 |
+
1.0,
|
| 1566 |
+
0.9999542236328125,
|
| 1567 |
+
0.9999967813491821,
|
| 1568 |
+
0.9966945648193359,
|
| 1569 |
+
0.9999995231628418,
|
| 1570 |
+
0.9955873489379883,
|
| 1571 |
+
0.9999998807907104,
|
| 1572 |
+
0.8881388306617737,
|
| 1573 |
+
0.9999102354049683,
|
| 1574 |
+
0.9994561076164246,
|
| 1575 |
+
0.9989873766899109,
|
| 1576 |
+
0.9133449792861938,
|
| 1577 |
+
0.9107068181037903,
|
| 1578 |
+
0.9093677401542664
|
| 1579 |
+
],
|
| 1580 |
+
"eval_code_runtime": 105.8053,
|
| 1581 |
+
"eval_code_samples_per_second": 0.945,
|
| 1582 |
+
"eval_code_score": -0.03846995532512665,
|
| 1583 |
+
"eval_code_steps_per_second": 0.038,
|
| 1584 |
+
"step": 150
|
| 1585 |
+
},
|
| 1586 |
+
{
|
| 1587 |
+
"loss": 0.0046,
|
| 1588 |
+
"learning_rate": 0.0002,
|
| 1589 |
+
"epoch": 2.65,
|
| 1590 |
+
"step": 151
|
| 1591 |
+
},
|
| 1592 |
+
{
|
| 1593 |
+
"loss": 0.0011,
|
| 1594 |
+
"learning_rate": 0.0002,
|
| 1595 |
+
"epoch": 2.67,
|
| 1596 |
+
"step": 152
|
| 1597 |
+
},
|
| 1598 |
+
{
|
| 1599 |
+
"loss": 0.0004,
|
| 1600 |
+
"learning_rate": 0.0002,
|
| 1601 |
+
"epoch": 2.68,
|
| 1602 |
+
"step": 153
|
| 1603 |
+
},
|
| 1604 |
+
{
|
| 1605 |
+
"loss": 0.0009,
|
| 1606 |
+
"learning_rate": 0.0002,
|
| 1607 |
+
"epoch": 2.7,
|
| 1608 |
+
"step": 154
|
| 1609 |
+
},
|
| 1610 |
+
{
|
| 1611 |
+
"loss": 0.0167,
|
| 1612 |
+
"learning_rate": 0.0002,
|
| 1613 |
+
"epoch": 2.72,
|
| 1614 |
+
"step": 155
|
| 1615 |
+
},
|
| 1616 |
+
{
|
| 1617 |
+
"loss": 0.0033,
|
| 1618 |
+
"learning_rate": 0.0002,
|
| 1619 |
+
"epoch": 2.74,
|
| 1620 |
+
"step": 156
|
| 1621 |
+
},
|
| 1622 |
+
{
|
| 1623 |
+
"loss": 0.0004,
|
| 1624 |
+
"learning_rate": 0.0002,
|
| 1625 |
+
"epoch": 2.75,
|
| 1626 |
+
"step": 157
|
| 1627 |
+
},
|
| 1628 |
+
{
|
| 1629 |
+
"loss": 0.0013,
|
| 1630 |
+
"learning_rate": 0.0002,
|
| 1631 |
+
"epoch": 2.77,
|
| 1632 |
+
"step": 158
|
| 1633 |
+
},
|
| 1634 |
+
{
|
| 1635 |
+
"loss": 0.0003,
|
| 1636 |
+
"learning_rate": 0.0002,
|
| 1637 |
+
"epoch": 2.79,
|
| 1638 |
+
"step": 159
|
| 1639 |
+
},
|
| 1640 |
+
{
|
| 1641 |
+
"loss": 0.0005,
|
| 1642 |
+
"learning_rate": 0.0002,
|
| 1643 |
+
"epoch": 2.81,
|
| 1644 |
+
"step": 160
|
| 1645 |
+
},
|
| 1646 |
+
{
|
| 1647 |
+
"loss": 0.002,
|
| 1648 |
+
"learning_rate": 0.0002,
|
| 1649 |
+
"epoch": 2.82,
|
| 1650 |
+
"step": 161
|
| 1651 |
+
},
|
| 1652 |
+
{
|
| 1653 |
+
"loss": 0.0001,
|
| 1654 |
+
"learning_rate": 0.0002,
|
| 1655 |
+
"epoch": 2.84,
|
| 1656 |
+
"step": 162
|
| 1657 |
+
},
|
| 1658 |
+
{
|
| 1659 |
+
"loss": 0.0052,
|
| 1660 |
+
"learning_rate": 0.0002,
|
| 1661 |
+
"epoch": 2.86,
|
| 1662 |
+
"step": 163
|
| 1663 |
+
},
|
| 1664 |
+
{
|
| 1665 |
+
"loss": 0.0003,
|
| 1666 |
+
"learning_rate": 0.0002,
|
| 1667 |
+
"epoch": 2.88,
|
| 1668 |
+
"step": 164
|
| 1669 |
+
},
|
| 1670 |
+
{
|
| 1671 |
+
"loss": 0.0,
|
| 1672 |
+
"learning_rate": 0.0002,
|
| 1673 |
+
"epoch": 2.89,
|
| 1674 |
+
"step": 165
|
| 1675 |
+
},
|
| 1676 |
+
{
|
| 1677 |
+
"loss": 0.0043,
|
| 1678 |
+
"learning_rate": 0.0002,
|
| 1679 |
+
"epoch": 2.91,
|
| 1680 |
+
"step": 166
|
| 1681 |
+
},
|
| 1682 |
+
{
|
| 1683 |
+
"loss": 0.0,
|
| 1684 |
+
"learning_rate": 0.0002,
|
| 1685 |
+
"epoch": 2.93,
|
| 1686 |
+
"step": 167
|
| 1687 |
+
},
|
| 1688 |
+
{
|
| 1689 |
+
"loss": 0.0011,
|
| 1690 |
+
"learning_rate": 0.0002,
|
| 1691 |
+
"epoch": 2.95,
|
| 1692 |
+
"step": 168
|
| 1693 |
+
},
|
| 1694 |
+
{
|
| 1695 |
+
"loss": 0.0024,
|
| 1696 |
+
"learning_rate": 0.0002,
|
| 1697 |
+
"epoch": 2.96,
|
| 1698 |
+
"step": 169
|
| 1699 |
+
},
|
| 1700 |
+
{
|
| 1701 |
+
"loss": 0.0,
|
| 1702 |
+
"learning_rate": 0.0002,
|
| 1703 |
+
"epoch": 2.98,
|
| 1704 |
+
"step": 170
|
| 1705 |
+
},
|
| 1706 |
+
{
|
| 1707 |
+
"loss": 0.0009,
|
| 1708 |
+
"learning_rate": 0.0002,
|
| 1709 |
+
"epoch": 3.0,
|
| 1710 |
+
"step": 171
|
| 1711 |
+
},
|
| 1712 |
+
{
|
| 1713 |
+
"loss": 0.0008,
|
| 1714 |
+
"learning_rate": 0.0002,
|
| 1715 |
+
"epoch": 3.02,
|
| 1716 |
+
"step": 172
|
| 1717 |
+
},
|
| 1718 |
+
{
|
| 1719 |
+
"loss": 0.0007,
|
| 1720 |
+
"learning_rate": 0.0002,
|
| 1721 |
+
"epoch": 3.04,
|
| 1722 |
+
"step": 173
|
| 1723 |
+
},
|
| 1724 |
+
{
|
| 1725 |
+
"loss": 0.0001,
|
| 1726 |
+
"learning_rate": 0.0002,
|
| 1727 |
+
"epoch": 3.05,
|
| 1728 |
+
"step": 174
|
| 1729 |
+
},
|
| 1730 |
+
{
|
| 1731 |
+
"loss": 0.0182,
|
| 1732 |
+
"learning_rate": 0.0002,
|
| 1733 |
+
"epoch": 3.07,
|
| 1734 |
+
"step": 175
|
| 1735 |
+
},
|
| 1736 |
+
{
|
| 1737 |
+
"eval_code_loss": 0.24447500705718994,
|
| 1738 |
+
"eval_code_score": -0.04580119252204895,
|
| 1739 |
+
"eval_code_brier_score": 0.04580119252204895,
|
| 1740 |
+
"eval_code_average_probability": 0.9384365677833557,
|
| 1741 |
+
"eval_code_accuracy": 0.95,
|
| 1742 |
+
"eval_code_probabilities": [
|
| 1743 |
+
1.0,
|
| 1744 |
+
0.9999996423721313,
|
| 1745 |
+
1.0,
|
| 1746 |
+
1.0,
|
| 1747 |
+
1.0,
|
| 1748 |
+
1.0,
|
| 1749 |
+
1.0,
|
| 1750 |
+
1.0,
|
| 1751 |
+
1.0,
|
| 1752 |
+
1.0,
|
| 1753 |
+
1.0,
|
| 1754 |
+
1.0,
|
| 1755 |
+
1.0,
|
| 1756 |
+
1.0,
|
| 1757 |
+
0.9999998807907104,
|
| 1758 |
+
0.9705566167831421,
|
| 1759 |
+
0.28825750946998596,
|
| 1760 |
+
0.990113377571106,
|
| 1761 |
+
1.0,
|
| 1762 |
+
1.0,
|
| 1763 |
+
1.0,
|
| 1764 |
+
1.0,
|
| 1765 |
+
1.0,
|
| 1766 |
+
1.0,
|
| 1767 |
+
1.0,
|
| 1768 |
+
1.0,
|
| 1769 |
+
1.0,
|
| 1770 |
+
0.8121438026428223,
|
| 1771 |
+
0.8320725560188293,
|
| 1772 |
+
0.9545630216598511,
|
| 1773 |
+
0.022090904414653778,
|
| 1774 |
+
1.0,
|
| 1775 |
+
1.0,
|
| 1776 |
+
1.0,
|
| 1777 |
+
0.0003882711462210864,
|
| 1778 |
+
0.9997542500495911,
|
| 1779 |
+
1.0,
|
| 1780 |
+
1.0,
|
| 1781 |
+
1.0,
|
| 1782 |
+
1.0,
|
| 1783 |
+
0.9999994039535522,
|
| 1784 |
+
0.9999996423721313,
|
| 1785 |
+
1.0,
|
| 1786 |
+
1.0,
|
| 1787 |
+
1.0,
|
| 1788 |
+
0.9999974966049194,
|
| 1789 |
+
1.0,
|
| 1790 |
+
1.0,
|
| 1791 |
+
1.0,
|
| 1792 |
+
1.0,
|
| 1793 |
+
1.0,
|
| 1794 |
+
1.0,
|
| 1795 |
+
1.0,
|
| 1796 |
+
1.0,
|
| 1797 |
+
0.9805604219436646,
|
| 1798 |
+
0.656412661075592,
|
| 1799 |
+
0.9261404871940613,
|
| 1800 |
+
1.0,
|
| 1801 |
+
0.9999086856842041,
|
| 1802 |
+
1.0,
|
| 1803 |
+
1.0,
|
| 1804 |
+
1.0,
|
| 1805 |
+
1.0,
|
| 1806 |
+
0.9999998807907104,
|
| 1807 |
+
1.0,
|
| 1808 |
+
0.9739339351654053,
|
| 1809 |
+
1.0,
|
| 1810 |
+
0.9995070695877075,
|
| 1811 |
+
0.9991480112075806,
|
| 1812 |
+
0.9978482723236084,
|
| 1813 |
+
0.0005162784364074469,
|
| 1814 |
+
0.999909520149231,
|
| 1815 |
+
1.0,
|
| 1816 |
+
1.0,
|
| 1817 |
+
0.09138258546590805,
|
| 1818 |
+
1.0,
|
| 1819 |
+
0.9999922513961792,
|
| 1820 |
+
1.0,
|
| 1821 |
+
1.0,
|
| 1822 |
+
1.0,
|
| 1823 |
+
1.0,
|
| 1824 |
+
0.9993377327919006,
|
| 1825 |
+
0.829042375087738,
|
| 1826 |
+
0.993126630783081,
|
| 1827 |
+
1.0,
|
| 1828 |
+
0.9995529055595398,
|
| 1829 |
+
1.0,
|
| 1830 |
+
0.9999717473983765,
|
| 1831 |
+
0.9999934434890747,
|
| 1832 |
+
0.9994831085205078,
|
| 1833 |
+
1.0,
|
| 1834 |
+
0.9715592861175537,
|
| 1835 |
+
1.0,
|
| 1836 |
+
0.7726230025291443,
|
| 1837 |
+
0.9999821186065674,
|
| 1838 |
+
0.9998100399971008,
|
| 1839 |
+
0.9989909529685974,
|
| 1840 |
+
0.8984997272491455,
|
| 1841 |
+
0.8994256854057312,
|
| 1842 |
+
0.9870588183403015
|
| 1843 |
+
],
|
| 1844 |
+
"eval_code_runtime": 104.8369,
|
| 1845 |
+
"eval_code_samples_per_second": 0.954,
|
| 1846 |
+
"eval_code_steps_per_second": 0.038,
|
| 1847 |
+
"epoch": 3.07,
|
| 1848 |
+
"step": 175
|
| 1849 |
+
},
|
| 1850 |
+
{
|
| 1851 |
+
"loss": 0.0027,
|
| 1852 |
+
"learning_rate": 0.0002,
|
| 1853 |
+
"epoch": 3.09,
|
| 1854 |
+
"step": 176
|
| 1855 |
+
},
|
| 1856 |
+
{
|
| 1857 |
+
"loss": 0.0001,
|
| 1858 |
+
"learning_rate": 0.0002,
|
| 1859 |
+
"epoch": 3.11,
|
| 1860 |
+
"step": 177
|
| 1861 |
+
},
|
| 1862 |
+
{
|
| 1863 |
+
"loss": 0.0005,
|
| 1864 |
+
"learning_rate": 0.0002,
|
| 1865 |
+
"epoch": 3.12,
|
| 1866 |
+
"step": 178
|
| 1867 |
+
},
|
| 1868 |
+
{
|
| 1869 |
+
"loss": 0.0002,
|
| 1870 |
+
"learning_rate": 0.0002,
|
| 1871 |
+
"epoch": 3.14,
|
| 1872 |
+
"step": 179
|
| 1873 |
+
},
|
| 1874 |
+
{
|
| 1875 |
+
"loss": 0.0001,
|
| 1876 |
+
"learning_rate": 0.0002,
|
| 1877 |
+
"epoch": 3.16,
|
| 1878 |
+
"step": 180
|
| 1879 |
+
},
|
| 1880 |
+
{
|
| 1881 |
+
"loss": 0.0,
|
| 1882 |
+
"learning_rate": 0.0002,
|
| 1883 |
+
"epoch": 3.18,
|
| 1884 |
+
"step": 181
|
| 1885 |
+
},
|
| 1886 |
+
{
|
| 1887 |
+
"loss": 0.0005,
|
| 1888 |
+
"learning_rate": 0.0002,
|
| 1889 |
+
"epoch": 3.19,
|
| 1890 |
+
"step": 182
|
| 1891 |
+
},
|
| 1892 |
+
{
|
| 1893 |
+
"loss": 0.001,
|
| 1894 |
+
"learning_rate": 0.0002,
|
| 1895 |
+
"epoch": 3.21,
|
| 1896 |
+
"step": 183
|
| 1897 |
+
},
|
| 1898 |
+
{
|
| 1899 |
+
"loss": 0.0,
|
| 1900 |
+
"learning_rate": 0.0002,
|
| 1901 |
+
"epoch": 3.23,
|
| 1902 |
+
"step": 184
|
| 1903 |
+
},
|
| 1904 |
+
{
|
| 1905 |
+
"loss": 0.0004,
|
| 1906 |
+
"learning_rate": 0.0002,
|
| 1907 |
+
"epoch": 3.25,
|
| 1908 |
+
"step": 185
|
| 1909 |
+
},
|
| 1910 |
+
{
|
| 1911 |
+
"loss": 0.0008,
|
| 1912 |
+
"learning_rate": 0.0002,
|
| 1913 |
+
"epoch": 3.26,
|
| 1914 |
+
"step": 186
|
| 1915 |
+
},
|
| 1916 |
+
{
|
| 1917 |
+
"loss": 0.0315,
|
| 1918 |
+
"learning_rate": 0.0002,
|
| 1919 |
+
"epoch": 3.28,
|
| 1920 |
+
"step": 187
|
| 1921 |
+
},
|
| 1922 |
+
{
|
| 1923 |
+
"loss": 0.0148,
|
| 1924 |
+
"learning_rate": 0.0002,
|
| 1925 |
+
"epoch": 3.3,
|
| 1926 |
+
"step": 188
|
| 1927 |
+
},
|
| 1928 |
+
{
|
| 1929 |
+
"loss": 0.0,
|
| 1930 |
+
"learning_rate": 0.0002,
|
| 1931 |
+
"epoch": 3.32,
|
| 1932 |
+
"step": 189
|
| 1933 |
+
},
|
| 1934 |
+
{
|
| 1935 |
+
"loss": 0.0016,
|
| 1936 |
+
"learning_rate": 0.0002,
|
| 1937 |
+
"epoch": 3.33,
|
| 1938 |
+
"step": 190
|
| 1939 |
+
},
|
| 1940 |
+
{
|
| 1941 |
+
"loss": 0.0053,
|
| 1942 |
+
"learning_rate": 0.0002,
|
| 1943 |
+
"epoch": 3.35,
|
| 1944 |
+
"step": 191
|
| 1945 |
+
},
|
| 1946 |
+
{
|
| 1947 |
+
"loss": 0.0001,
|
| 1948 |
+
"learning_rate": 0.0002,
|
| 1949 |
+
"epoch": 3.37,
|
| 1950 |
+
"step": 192
|
| 1951 |
+
},
|
| 1952 |
+
{
|
| 1953 |
+
"loss": 0.0009,
|
| 1954 |
+
"learning_rate": 0.0002,
|
| 1955 |
+
"epoch": 3.39,
|
| 1956 |
+
"step": 193
|
| 1957 |
+
},
|
| 1958 |
+
{
|
| 1959 |
+
"loss": 0.1843,
|
| 1960 |
+
"learning_rate": 0.0002,
|
| 1961 |
+
"epoch": 3.4,
|
| 1962 |
+
"step": 194
|
| 1963 |
+
},
|
| 1964 |
+
{
|
| 1965 |
+
"loss": 0.0001,
|
| 1966 |
+
"learning_rate": 0.0002,
|
| 1967 |
+
"epoch": 3.42,
|
| 1968 |
+
"step": 195
|
| 1969 |
+
},
|
| 1970 |
+
{
|
| 1971 |
+
"loss": 0.012,
|
| 1972 |
+
"learning_rate": 0.0002,
|
| 1973 |
+
"epoch": 3.44,
|
| 1974 |
+
"step": 196
|
| 1975 |
+
},
|
| 1976 |
+
{
|
| 1977 |
+
"loss": 0.0031,
|
| 1978 |
+
"learning_rate": 0.0002,
|
| 1979 |
+
"epoch": 3.46,
|
| 1980 |
+
"step": 197
|
| 1981 |
+
},
|
| 1982 |
+
{
|
| 1983 |
+
"loss": 0.0003,
|
| 1984 |
+
"learning_rate": 0.0002,
|
| 1985 |
+
"epoch": 3.47,
|
| 1986 |
+
"step": 198
|
| 1987 |
+
},
|
| 1988 |
+
{
|
| 1989 |
+
"loss": 0.0062,
|
| 1990 |
+
"learning_rate": 0.0002,
|
| 1991 |
+
"epoch": 3.49,
|
| 1992 |
+
"step": 199
|
| 1993 |
+
},
|
| 1994 |
+
{
|
| 1995 |
+
"loss": 0.0011,
|
| 1996 |
+
"learning_rate": 0.0002,
|
| 1997 |
+
"epoch": 3.51,
|
| 1998 |
+
"step": 200
|
| 1999 |
+
},
|
| 2000 |
+
{
|
| 2001 |
+
"eval_code_loss": 0.20703525841236115,
|
| 2002 |
+
"eval_code_score": -0.036074623465538025,
|
| 2003 |
+
"eval_code_brier_score": 0.036074623465538025,
|
| 2004 |
+
"eval_code_average_probability": 0.9416353702545166,
|
| 2005 |
+
"eval_code_accuracy": 0.96,
|
| 2006 |
+
"eval_code_probabilities": [
|
| 2007 |
+
1.0,
|
| 2008 |
+
1.0,
|
| 2009 |
+
1.0,
|
| 2010 |
+
1.0,
|
| 2011 |
+
1.0,
|
| 2012 |
+
1.0,
|
| 2013 |
+
1.0,
|
| 2014 |
+
1.0,
|
| 2015 |
+
1.0,
|
| 2016 |
+
1.0,
|
| 2017 |
+
1.0,
|
| 2018 |
+
1.0,
|
| 2019 |
+
1.0,
|
| 2020 |
+
0.9999991655349731,
|
| 2021 |
+
0.999995231628418,
|
| 2022 |
+
0.9498335719108582,
|
| 2023 |
+
0.30828458070755005,
|
| 2024 |
+
0.987034261226654,
|
| 2025 |
+
0.999970555305481,
|
| 2026 |
+
0.9999929666519165,
|
| 2027 |
+
0.9999411106109619,
|
| 2028 |
+
0.999946117401123,
|
| 2029 |
+
0.9999997615814209,
|
| 2030 |
+
1.0,
|
| 2031 |
+
1.0,
|
| 2032 |
+
1.0,
|
| 2033 |
+
1.0,
|
| 2034 |
+
0.717918872833252,
|
| 2035 |
+
0.6784900426864624,
|
| 2036 |
+
0.8770557641983032,
|
| 2037 |
+
7.296191597561119e-06,
|
| 2038 |
+
0.9999998807907104,
|
| 2039 |
+
0.9999997615814209,
|
| 2040 |
+
1.0,
|
| 2041 |
+
0.9999603033065796,
|
| 2042 |
+
1.0,
|
| 2043 |
+
1.0,
|
| 2044 |
+
1.0,
|
| 2045 |
+
1.0,
|
| 2046 |
+
1.0,
|
| 2047 |
+
0.9999985694885254,
|
| 2048 |
+
0.9999916553497314,
|
| 2049 |
+
1.0,
|
| 2050 |
+
1.0,
|
| 2051 |
+
1.0,
|
| 2052 |
+
0.9999998807907104,
|
| 2053 |
+
1.0,
|
| 2054 |
+
1.0,
|
| 2055 |
+
0.9999997615814209,
|
| 2056 |
+
0.9999793767929077,
|
| 2057 |
+
1.0,
|
| 2058 |
+
1.0,
|
| 2059 |
+
1.0,
|
| 2060 |
+
1.0,
|
| 2061 |
+
0.9963011741638184,
|
| 2062 |
+
0.5810772180557251,
|
| 2063 |
+
0.9937513470649719,
|
| 2064 |
+
1.0,
|
| 2065 |
+
0.9999946355819702,
|
| 2066 |
+
1.0,
|
| 2067 |
+
1.0,
|
| 2068 |
+
1.0,
|
| 2069 |
+
1.0,
|
| 2070 |
+
0.9999998807907104,
|
| 2071 |
+
0.9999996423721313,
|
| 2072 |
+
0.999351441860199,
|
| 2073 |
+
1.0,
|
| 2074 |
+
0.9979580640792847,
|
| 2075 |
+
0.9855725169181824,
|
| 2076 |
+
0.9849231243133545,
|
| 2077 |
+
0.017488110810518265,
|
| 2078 |
+
0.999018669128418,
|
| 2079 |
+
0.9788364768028259,
|
| 2080 |
+
0.9989467263221741,
|
| 2081 |
+
0.29412388801574707,
|
| 2082 |
+
1.0,
|
| 2083 |
+
0.9881665110588074,
|
| 2084 |
+
1.0,
|
| 2085 |
+
1.0,
|
| 2086 |
+
0.9999973773956299,
|
| 2087 |
+
1.0,
|
| 2088 |
+
0.9999986886978149,
|
| 2089 |
+
0.8509193658828735,
|
| 2090 |
+
0.9998080134391785,
|
| 2091 |
+
1.0,
|
| 2092 |
+
0.9999957084655762,
|
| 2093 |
+
1.0,
|
| 2094 |
+
0.9992415904998779,
|
| 2095 |
+
0.9998071789741516,
|
| 2096 |
+
0.9979315996170044,
|
| 2097 |
+
0.9996376037597656,
|
| 2098 |
+
0.9590405821800232,
|
| 2099 |
+
0.9996711015701294,
|
| 2100 |
+
0.5990511775016785,
|
| 2101 |
+
0.9998351335525513,
|
| 2102 |
+
0.9997956156730652,
|
| 2103 |
+
0.9722345471382141,
|
| 2104 |
+
0.7871147990226746,
|
| 2105 |
+
0.8062353730201721,
|
| 2106 |
+
0.8593048453330994
|
| 2107 |
+
],
|
| 2108 |
+
"eval_code_runtime": 104.8009,
|
| 2109 |
+
"eval_code_samples_per_second": 0.954,
|
| 2110 |
+
"eval_code_steps_per_second": 0.038,
|
| 2111 |
+
"epoch": 3.51,
|
| 2112 |
+
"step": 200
|
| 2113 |
+
},
|
| 2114 |
+
{
|
| 2115 |
+
"loss": 0.0071,
|
| 2116 |
+
"learning_rate": 0.0002,
|
| 2117 |
+
"epoch": 3.53,
|
| 2118 |
+
"step": 201
|
| 2119 |
+
},
|
| 2120 |
+
{
|
| 2121 |
+
"loss": 0.0311,
|
| 2122 |
+
"learning_rate": 0.0002,
|
| 2123 |
+
"epoch": 3.54,
|
| 2124 |
+
"step": 202
|
| 2125 |
+
},
|
| 2126 |
+
{
|
| 2127 |
+
"loss": 0.0009,
|
| 2128 |
+
"learning_rate": 0.0002,
|
| 2129 |
+
"epoch": 3.56,
|
| 2130 |
+
"step": 203
|
| 2131 |
+
},
|
| 2132 |
+
{
|
| 2133 |
+
"loss": 0.0003,
|
| 2134 |
+
"learning_rate": 0.0002,
|
| 2135 |
+
"epoch": 3.58,
|
| 2136 |
+
"step": 204
|
| 2137 |
+
},
|
| 2138 |
+
{
|
| 2139 |
+
"loss": 0.0004,
|
| 2140 |
+
"learning_rate": 0.0002,
|
| 2141 |
+
"epoch": 3.6,
|
| 2142 |
+
"step": 205
|
| 2143 |
+
},
|
| 2144 |
+
{
|
| 2145 |
+
"loss": 0.0018,
|
| 2146 |
+
"learning_rate": 0.0002,
|
| 2147 |
+
"epoch": 3.61,
|
| 2148 |
+
"step": 206
|
| 2149 |
+
},
|
| 2150 |
+
{
|
| 2151 |
+
"loss": 0.0065,
|
| 2152 |
+
"learning_rate": 0.0002,
|
| 2153 |
+
"epoch": 3.63,
|
| 2154 |
+
"step": 207
|
| 2155 |
+
},
|
| 2156 |
+
{
|
| 2157 |
+
"loss": 0.001,
|
| 2158 |
+
"learning_rate": 0.0002,
|
| 2159 |
+
"epoch": 3.65,
|
| 2160 |
+
"step": 208
|
| 2161 |
+
},
|
| 2162 |
+
{
|
| 2163 |
+
"loss": 0.0,
|
| 2164 |
+
"learning_rate": 0.0002,
|
| 2165 |
+
"epoch": 3.67,
|
| 2166 |
+
"step": 209
|
| 2167 |
+
},
|
| 2168 |
+
{
|
| 2169 |
+
"loss": 0.0302,
|
| 2170 |
+
"learning_rate": 0.0002,
|
| 2171 |
+
"epoch": 3.68,
|
| 2172 |
+
"step": 210
|
| 2173 |
+
},
|
| 2174 |
+
{
|
| 2175 |
+
"loss": 0.0662,
|
| 2176 |
+
"learning_rate": 0.0002,
|
| 2177 |
+
"epoch": 3.7,
|
| 2178 |
+
"step": 211
|
| 2179 |
+
},
|
| 2180 |
+
{
|
| 2181 |
+
"loss": 0.0071,
|
| 2182 |
+
"learning_rate": 0.0002,
|
| 2183 |
+
"epoch": 3.72,
|
| 2184 |
+
"step": 212
|
| 2185 |
+
},
|
| 2186 |
+
{
|
| 2187 |
+
"loss": 0.0001,
|
| 2188 |
+
"learning_rate": 0.0002,
|
| 2189 |
+
"epoch": 3.74,
|
| 2190 |
+
"step": 213
|
| 2191 |
+
},
|
| 2192 |
+
{
|
| 2193 |
+
"loss": 0.0527,
|
| 2194 |
+
"learning_rate": 0.0002,
|
| 2195 |
+
"epoch": 3.75,
|
| 2196 |
+
"step": 214
|
| 2197 |
+
},
|
| 2198 |
+
{
|
| 2199 |
+
"loss": 0.0069,
|
| 2200 |
+
"learning_rate": 0.0002,
|
| 2201 |
+
"epoch": 3.77,
|
| 2202 |
+
"step": 215
|
| 2203 |
+
},
|
| 2204 |
+
{
|
| 2205 |
+
"loss": 0.0002,
|
| 2206 |
+
"learning_rate": 0.0002,
|
| 2207 |
+
"epoch": 3.79,
|
| 2208 |
+
"step": 216
|
| 2209 |
+
},
|
| 2210 |
+
{
|
| 2211 |
+
"loss": 0.0056,
|
| 2212 |
+
"learning_rate": 0.0002,
|
| 2213 |
+
"epoch": 3.81,
|
| 2214 |
+
"step": 217
|
| 2215 |
+
},
|
| 2216 |
+
{
|
| 2217 |
+
"loss": 0.0205,
|
| 2218 |
+
"learning_rate": 0.0002,
|
| 2219 |
+
"epoch": 3.82,
|
| 2220 |
+
"step": 218
|
| 2221 |
+
},
|
| 2222 |
+
{
|
| 2223 |
+
"loss": 0.0,
|
| 2224 |
+
"learning_rate": 0.0002,
|
| 2225 |
+
"epoch": 3.84,
|
| 2226 |
+
"step": 219
|
| 2227 |
+
},
|
| 2228 |
+
{
|
| 2229 |
+
"loss": 0.0048,
|
| 2230 |
+
"learning_rate": 0.0002,
|
| 2231 |
+
"epoch": 3.86,
|
| 2232 |
+
"step": 220
|
| 2233 |
+
},
|
| 2234 |
+
{
|
| 2235 |
+
"loss": 0.0307,
|
| 2236 |
+
"learning_rate": 0.0002,
|
| 2237 |
+
"epoch": 3.88,
|
| 2238 |
+
"step": 221
|
| 2239 |
+
},
|
| 2240 |
+
{
|
| 2241 |
+
"loss": 0.0092,
|
| 2242 |
+
"learning_rate": 0.0002,
|
| 2243 |
+
"epoch": 3.89,
|
| 2244 |
+
"step": 222
|
| 2245 |
+
},
|
| 2246 |
+
{
|
| 2247 |
+
"loss": 0.0425,
|
| 2248 |
+
"learning_rate": 0.0002,
|
| 2249 |
+
"epoch": 3.91,
|
| 2250 |
+
"step": 223
|
| 2251 |
+
},
|
| 2252 |
+
{
|
| 2253 |
+
"loss": 0.0005,
|
| 2254 |
+
"learning_rate": 0.0002,
|
| 2255 |
+
"epoch": 3.93,
|
| 2256 |
+
"step": 224
|
| 2257 |
+
},
|
| 2258 |
+
{
|
| 2259 |
+
"loss": 0.0021,
|
| 2260 |
+
"learning_rate": 0.0002,
|
| 2261 |
+
"epoch": 3.95,
|
| 2262 |
+
"step": 225
|
| 2263 |
+
},
|
| 2264 |
+
{
|
| 2265 |
+
"eval_code_loss": 0.1442316770553589,
|
| 2266 |
+
"eval_code_score": -0.03884093835949898,
|
| 2267 |
+
"eval_code_brier_score": 0.03884093835949898,
|
| 2268 |
+
"eval_code_average_probability": 0.9314249157905579,
|
| 2269 |
+
"eval_code_accuracy": 0.95,
|
| 2270 |
+
"eval_code_probabilities": [
|
| 2271 |
+
1.0,
|
| 2272 |
+
1.0,
|
| 2273 |
+
1.0,
|
| 2274 |
+
1.0,
|
| 2275 |
+
1.0,
|
| 2276 |
+
1.0,
|
| 2277 |
+
1.0,
|
| 2278 |
+
1.0,
|
| 2279 |
+
1.0,
|
| 2280 |
+
1.0,
|
| 2281 |
+
1.0,
|
| 2282 |
+
1.0,
|
| 2283 |
+
1.0,
|
| 2284 |
+
0.9999998807907104,
|
| 2285 |
+
0.9999983310699463,
|
| 2286 |
+
0.9166925549507141,
|
| 2287 |
+
0.46314355731010437,
|
| 2288 |
+
0.9695256948471069,
|
| 2289 |
+
0.9999979734420776,
|
| 2290 |
+
0.9999997615814209,
|
| 2291 |
+
0.9999912977218628,
|
| 2292 |
+
0.9998043179512024,
|
| 2293 |
+
0.9999998807907104,
|
| 2294 |
+
1.0,
|
| 2295 |
+
0.999934196472168,
|
| 2296 |
+
0.9998651742935181,
|
| 2297 |
+
0.9999654293060303,
|
| 2298 |
+
0.7587816119194031,
|
| 2299 |
+
0.8365660309791565,
|
| 2300 |
+
0.8947399258613586,
|
| 2301 |
+
0.007870269939303398,
|
| 2302 |
+
0.9999408721923828,
|
| 2303 |
+
0.9807198643684387,
|
| 2304 |
+
1.0,
|
| 2305 |
+
1.0,
|
| 2306 |
+
1.0,
|
| 2307 |
+
1.0,
|
| 2308 |
+
1.0,
|
| 2309 |
+
1.0,
|
| 2310 |
+
1.0,
|
| 2311 |
+
1.0,
|
| 2312 |
+
1.0,
|
| 2313 |
+
1.0,
|
| 2314 |
+
1.0,
|
| 2315 |
+
1.0,
|
| 2316 |
+
0.9999988079071045,
|
| 2317 |
+
1.0,
|
| 2318 |
+
1.0,
|
| 2319 |
+
1.0,
|
| 2320 |
+
0.9999998807907104,
|
| 2321 |
+
1.0,
|
| 2322 |
+
1.0,
|
| 2323 |
+
1.0,
|
| 2324 |
+
1.0,
|
| 2325 |
+
0.9998663663864136,
|
| 2326 |
+
0.9656995534896851,
|
| 2327 |
+
0.935349702835083,
|
| 2328 |
+
1.0,
|
| 2329 |
+
0.9999972581863403,
|
| 2330 |
+
1.0,
|
| 2331 |
+
1.0,
|
| 2332 |
+
1.0,
|
| 2333 |
+
0.9999998807907104,
|
| 2334 |
+
0.9999998807907104,
|
| 2335 |
+
0.9999998807907104,
|
| 2336 |
+
0.9994795918464661,
|
| 2337 |
+
0.9999734163284302,
|
| 2338 |
+
0.4865659773349762,
|
| 2339 |
+
0.5832042694091797,
|
| 2340 |
+
0.9857308268547058,
|
| 2341 |
+
0.015312162227928638,
|
| 2342 |
+
0.999519944190979,
|
| 2343 |
+
0.7279074788093567,
|
| 2344 |
+
0.9087716937065125,
|
| 2345 |
+
0.6908769011497498,
|
| 2346 |
+
1.0,
|
| 2347 |
+
0.9999642372131348,
|
| 2348 |
+
1.0,
|
| 2349 |
+
1.0,
|
| 2350 |
+
0.9999998807907104,
|
| 2351 |
+
1.0,
|
| 2352 |
+
1.0,
|
| 2353 |
+
0.8065006136894226,
|
| 2354 |
+
0.9999997615814209,
|
| 2355 |
+
1.0,
|
| 2356 |
+
0.999992847442627,
|
| 2357 |
+
1.0,
|
| 2358 |
+
0.998808741569519,
|
| 2359 |
+
0.9994845390319824,
|
| 2360 |
+
0.9905843734741211,
|
| 2361 |
+
0.9999678134918213,
|
| 2362 |
+
0.9825097322463989,
|
| 2363 |
+
0.9999821186065674,
|
| 2364 |
+
0.316112756729126,
|
| 2365 |
+
0.9999034404754639,
|
| 2366 |
+
0.9957683086395264,
|
| 2367 |
+
0.8763726353645325,
|
| 2368 |
+
0.557384192943573,
|
| 2369 |
+
0.5571630001068115,
|
| 2370 |
+
0.936220109462738
|
| 2371 |
+
],
|
| 2372 |
+
"eval_code_runtime": 104.8083,
|
| 2373 |
+
"eval_code_samples_per_second": 0.954,
|
| 2374 |
+
"eval_code_steps_per_second": 0.038,
|
| 2375 |
+
"epoch": 3.95,
|
| 2376 |
+
"step": 225
|
| 2377 |
+
},
|
| 2378 |
+
{
|
| 2379 |
+
"loss": 0.0037,
|
| 2380 |
+
"learning_rate": 0.0002,
|
| 2381 |
+
"epoch": 3.96,
|
| 2382 |
+
"step": 226
|
| 2383 |
+
},
|
| 2384 |
+
{
|
| 2385 |
+
"loss": 0.0061,
|
| 2386 |
+
"learning_rate": 0.0002,
|
| 2387 |
+
"epoch": 3.98,
|
| 2388 |
+
"step": 227
|
| 2389 |
+
},
|
| 2390 |
+
{
|
| 2391 |
+
"loss": 0.0211,
|
| 2392 |
+
"learning_rate": 0.0002,
|
| 2393 |
+
"epoch": 4.0,
|
| 2394 |
+
"step": 228
|
| 2395 |
+
},
|
| 2396 |
+
{
|
| 2397 |
+
"loss": 0.0028,
|
| 2398 |
+
"learning_rate": 0.0002,
|
| 2399 |
+
"epoch": 4.02,
|
| 2400 |
+
"step": 229
|
| 2401 |
+
},
|
| 2402 |
+
{
|
| 2403 |
+
"loss": 0.0009,
|
| 2404 |
+
"learning_rate": 0.0002,
|
| 2405 |
+
"epoch": 4.04,
|
| 2406 |
+
"step": 230
|
| 2407 |
+
},
|
| 2408 |
+
{
|
| 2409 |
+
"loss": 0.0146,
|
| 2410 |
+
"learning_rate": 0.0002,
|
| 2411 |
+
"epoch": 4.05,
|
| 2412 |
+
"step": 231
|
| 2413 |
+
},
|
| 2414 |
+
{
|
| 2415 |
+
"loss": 0.0155,
|
| 2416 |
+
"learning_rate": 0.0002,
|
| 2417 |
+
"epoch": 4.07,
|
| 2418 |
+
"step": 232
|
| 2419 |
+
},
|
| 2420 |
+
{
|
| 2421 |
+
"loss": 0.0008,
|
| 2422 |
+
"learning_rate": 0.0002,
|
| 2423 |
+
"epoch": 4.09,
|
| 2424 |
+
"step": 233
|
| 2425 |
+
},
|
| 2426 |
+
{
|
| 2427 |
+
"loss": 0.0008,
|
| 2428 |
+
"learning_rate": 0.0002,
|
| 2429 |
+
"epoch": 4.11,
|
| 2430 |
+
"step": 234
|
| 2431 |
+
},
|
| 2432 |
+
{
|
| 2433 |
+
"loss": 0.0001,
|
| 2434 |
+
"learning_rate": 0.0002,
|
| 2435 |
+
"epoch": 4.12,
|
| 2436 |
+
"step": 235
|
| 2437 |
+
},
|
| 2438 |
+
{
|
| 2439 |
+
"loss": 0.0004,
|
| 2440 |
+
"learning_rate": 0.0002,
|
| 2441 |
+
"epoch": 4.14,
|
| 2442 |
+
"step": 236
|
| 2443 |
+
},
|
| 2444 |
+
{
|
| 2445 |
+
"loss": 0.0073,
|
| 2446 |
+
"learning_rate": 0.0002,
|
| 2447 |
+
"epoch": 4.16,
|
| 2448 |
+
"step": 237
|
| 2449 |
+
},
|
| 2450 |
+
{
|
| 2451 |
+
"loss": 0.0005,
|
| 2452 |
+
"learning_rate": 0.0002,
|
| 2453 |
+
"epoch": 4.18,
|
| 2454 |
+
"step": 238
|
| 2455 |
+
},
|
| 2456 |
+
{
|
| 2457 |
+
"loss": 0.0011,
|
| 2458 |
+
"learning_rate": 0.0002,
|
| 2459 |
+
"epoch": 4.19,
|
| 2460 |
+
"step": 239
|
| 2461 |
+
},
|
| 2462 |
+
{
|
| 2463 |
+
"loss": 0.0004,
|
| 2464 |
+
"learning_rate": 0.0002,
|
| 2465 |
+
"epoch": 4.21,
|
| 2466 |
+
"step": 240
|
| 2467 |
+
},
|
| 2468 |
+
{
|
| 2469 |
+
"loss": 0.0001,
|
| 2470 |
+
"learning_rate": 0.0002,
|
| 2471 |
+
"epoch": 4.23,
|
| 2472 |
+
"step": 241
|
| 2473 |
+
},
|
| 2474 |
+
{
|
| 2475 |
+
"loss": 0.0005,
|
| 2476 |
+
"learning_rate": 0.0002,
|
| 2477 |
+
"epoch": 4.25,
|
| 2478 |
+
"step": 242
|
| 2479 |
+
},
|
| 2480 |
+
{
|
| 2481 |
+
"loss": 0.0005,
|
| 2482 |
+
"learning_rate": 0.0002,
|
| 2483 |
+
"epoch": 4.26,
|
| 2484 |
+
"step": 243
|
| 2485 |
+
},
|
| 2486 |
+
{
|
| 2487 |
+
"loss": 0.0009,
|
| 2488 |
+
"learning_rate": 0.0002,
|
| 2489 |
+
"epoch": 4.28,
|
| 2490 |
+
"step": 244
|
| 2491 |
+
},
|
| 2492 |
+
{
|
| 2493 |
+
"loss": 0.0026,
|
| 2494 |
+
"learning_rate": 0.0002,
|
| 2495 |
+
"epoch": 4.3,
|
| 2496 |
+
"step": 245
|
| 2497 |
+
},
|
| 2498 |
+
{
|
| 2499 |
+
"loss": 0.0003,
|
| 2500 |
+
"learning_rate": 0.0002,
|
| 2501 |
+
"epoch": 4.32,
|
| 2502 |
+
"step": 246
|
| 2503 |
+
},
|
| 2504 |
+
{
|
| 2505 |
+
"loss": 0.0,
|
| 2506 |
+
"learning_rate": 0.0002,
|
| 2507 |
+
"epoch": 4.33,
|
| 2508 |
+
"step": 247
|
| 2509 |
+
},
|
| 2510 |
+
{
|
| 2511 |
+
"loss": 0.0172,
|
| 2512 |
+
"learning_rate": 0.0002,
|
| 2513 |
+
"epoch": 4.35,
|
| 2514 |
+
"step": 248
|
| 2515 |
+
},
|
| 2516 |
+
{
|
| 2517 |
+
"loss": 0.0003,
|
| 2518 |
+
"learning_rate": 0.0002,
|
| 2519 |
+
"epoch": 4.37,
|
| 2520 |
+
"step": 249
|
| 2521 |
+
},
|
| 2522 |
+
{
|
| 2523 |
+
"loss": 0.0003,
|
| 2524 |
+
"learning_rate": 0.0002,
|
| 2525 |
+
"epoch": 4.39,
|
| 2526 |
+
"step": 250
|
| 2527 |
+
},
|
| 2528 |
+
{
|
| 2529 |
+
"eval_code_loss": 0.1744336634874344,
|
| 2530 |
+
"eval_code_score": -0.03369342163205147,
|
| 2531 |
+
"eval_code_brier_score": 0.03369342163205147,
|
| 2532 |
+
"eval_code_average_probability": 0.9495358467102051,
|
| 2533 |
+
"eval_code_accuracy": 0.97,
|
| 2534 |
+
"eval_code_probabilities": [
|
| 2535 |
+
1.0,
|
| 2536 |
+
1.0,
|
| 2537 |
+
1.0,
|
| 2538 |
+
1.0,
|
| 2539 |
+
1.0,
|
| 2540 |
+
1.0,
|
| 2541 |
+
1.0,
|
| 2542 |
+
1.0,
|
| 2543 |
+
1.0,
|
| 2544 |
+
1.0,
|
| 2545 |
+
1.0,
|
| 2546 |
+
1.0,
|
| 2547 |
+
1.0,
|
| 2548 |
+
1.0,
|
| 2549 |
+
1.0,
|
| 2550 |
+
0.9451481699943542,
|
| 2551 |
+
0.5480159521102905,
|
| 2552 |
+
0.9964591860771179,
|
| 2553 |
+
1.0,
|
| 2554 |
+
1.0,
|
| 2555 |
+
1.0,
|
| 2556 |
+
1.0,
|
| 2557 |
+
1.0,
|
| 2558 |
+
1.0,
|
| 2559 |
+
1.0,
|
| 2560 |
+
1.0,
|
| 2561 |
+
1.0,
|
| 2562 |
+
0.7053650617599487,
|
| 2563 |
+
0.8087413311004639,
|
| 2564 |
+
0.9013814330101013,
|
| 2565 |
+
0.0008012360776774585,
|
| 2566 |
+
1.0,
|
| 2567 |
+
0.9999994039535522,
|
| 2568 |
+
1.0,
|
| 2569 |
+
0.9998193383216858,
|
| 2570 |
+
0.9999997615814209,
|
| 2571 |
+
1.0,
|
| 2572 |
+
1.0,
|
| 2573 |
+
1.0,
|
| 2574 |
+
1.0,
|
| 2575 |
+
1.0,
|
| 2576 |
+
1.0,
|
| 2577 |
+
1.0,
|
| 2578 |
+
1.0,
|
| 2579 |
+
1.0,
|
| 2580 |
+
0.9999983310699463,
|
| 2581 |
+
1.0,
|
| 2582 |
+
1.0,
|
| 2583 |
+
1.0,
|
| 2584 |
+
1.0,
|
| 2585 |
+
1.0,
|
| 2586 |
+
1.0,
|
| 2587 |
+
1.0,
|
| 2588 |
+
1.0,
|
| 2589 |
+
0.9992625117301941,
|
| 2590 |
+
0.9287911653518677,
|
| 2591 |
+
0.872682511806488,
|
| 2592 |
+
1.0,
|
| 2593 |
+
0.9996117949485779,
|
| 2594 |
+
1.0,
|
| 2595 |
+
1.0,
|
| 2596 |
+
1.0,
|
| 2597 |
+
1.0,
|
| 2598 |
+
1.0,
|
| 2599 |
+
1.0,
|
| 2600 |
+
0.9999651908874512,
|
| 2601 |
+
1.0,
|
| 2602 |
+
0.9498048424720764,
|
| 2603 |
+
0.9871782064437866,
|
| 2604 |
+
0.9963241815567017,
|
| 2605 |
+
0.0026193673256784678,
|
| 2606 |
+
0.999943733215332,
|
| 2607 |
+
0.9986830353736877,
|
| 2608 |
+
0.9999673366546631,
|
| 2609 |
+
0.9674615859985352,
|
| 2610 |
+
1.0,
|
| 2611 |
+
1.0,
|
| 2612 |
+
1.0,
|
| 2613 |
+
1.0,
|
| 2614 |
+
1.0,
|
| 2615 |
+
1.0,
|
| 2616 |
+
0.9999785423278809,
|
| 2617 |
+
0.9016892313957214,
|
| 2618 |
+
0.9997751116752625,
|
| 2619 |
+
1.0,
|
| 2620 |
+
0.9999997615814209,
|
| 2621 |
+
1.0,
|
| 2622 |
+
0.9999815225601196,
|
| 2623 |
+
0.9999998807907104,
|
| 2624 |
+
0.9999561309814453,
|
| 2625 |
+
1.0,
|
| 2626 |
+
0.9956650137901306,
|
| 2627 |
+
1.0,
|
| 2628 |
+
0.08844359964132309,
|
| 2629 |
+
0.9999833106994629,
|
| 2630 |
+
0.9990620017051697,
|
| 2631 |
+
0.9781332612037659,
|
| 2632 |
+
0.747072160243988,
|
| 2633 |
+
0.6776840090751648,
|
| 2634 |
+
0.9581350088119507
|
| 2635 |
+
],
|
| 2636 |
+
"eval_code_runtime": 104.7916,
|
| 2637 |
+
"eval_code_samples_per_second": 0.954,
|
| 2638 |
+
"eval_code_steps_per_second": 0.038,
|
| 2639 |
+
"epoch": 4.39,
|
| 2640 |
+
"step": 250
|
| 2641 |
+
},
|
| 2642 |
+
{
|
| 2643 |
+
"loss": 0.0001,
|
| 2644 |
+
"learning_rate": 0.0002,
|
| 2645 |
+
"epoch": 4.4,
|
| 2646 |
+
"step": 251
|
| 2647 |
+
},
|
| 2648 |
+
{
|
| 2649 |
+
"loss": 0.0008,
|
| 2650 |
+
"learning_rate": 0.0002,
|
| 2651 |
+
"epoch": 4.42,
|
| 2652 |
+
"step": 252
|
| 2653 |
+
},
|
| 2654 |
+
{
|
| 2655 |
+
"loss": 0.0004,
|
| 2656 |
+
"learning_rate": 0.0002,
|
| 2657 |
+
"epoch": 4.44,
|
| 2658 |
+
"step": 253
|
| 2659 |
+
},
|
| 2660 |
+
{
|
| 2661 |
+
"loss": 0.0034,
|
| 2662 |
+
"learning_rate": 0.0002,
|
| 2663 |
+
"epoch": 4.46,
|
| 2664 |
+
"step": 254
|
| 2665 |
+
},
|
| 2666 |
+
{
|
| 2667 |
+
"loss": 0.0006,
|
| 2668 |
+
"learning_rate": 0.0002,
|
| 2669 |
+
"epoch": 4.47,
|
| 2670 |
+
"step": 255
|
| 2671 |
+
},
|
| 2672 |
+
{
|
| 2673 |
+
"loss": 0.0,
|
| 2674 |
+
"learning_rate": 0.0002,
|
| 2675 |
+
"epoch": 4.49,
|
| 2676 |
+
"step": 256
|
| 2677 |
+
},
|
| 2678 |
+
{
|
| 2679 |
+
"loss": 0.0004,
|
| 2680 |
+
"learning_rate": 0.0002,
|
| 2681 |
+
"epoch": 4.51,
|
| 2682 |
+
"step": 257
|
| 2683 |
+
},
|
| 2684 |
+
{
|
| 2685 |
+
"loss": 0.0009,
|
| 2686 |
+
"learning_rate": 0.0002,
|
| 2687 |
+
"epoch": 4.53,
|
| 2688 |
+
"step": 258
|
| 2689 |
+
},
|
| 2690 |
+
{
|
| 2691 |
+
"loss": 0.0042,
|
| 2692 |
+
"learning_rate": 0.0002,
|
| 2693 |
+
"epoch": 4.54,
|
| 2694 |
+
"step": 259
|
| 2695 |
+
},
|
| 2696 |
+
{
|
| 2697 |
+
"loss": 0.0003,
|
| 2698 |
+
"learning_rate": 0.0002,
|
| 2699 |
+
"epoch": 4.56,
|
| 2700 |
+
"step": 260
|
| 2701 |
+
},
|
| 2702 |
+
{
|
| 2703 |
+
"loss": 0.0001,
|
| 2704 |
+
"learning_rate": 0.0002,
|
| 2705 |
+
"epoch": 4.58,
|
| 2706 |
+
"step": 261
|
| 2707 |
+
},
|
| 2708 |
+
{
|
| 2709 |
+
"loss": 0.0002,
|
| 2710 |
+
"learning_rate": 0.0002,
|
| 2711 |
+
"epoch": 4.6,
|
| 2712 |
+
"step": 262
|
| 2713 |
+
},
|
| 2714 |
+
{
|
| 2715 |
+
"loss": 0.0023,
|
| 2716 |
+
"learning_rate": 0.0002,
|
| 2717 |
+
"epoch": 4.61,
|
| 2718 |
+
"step": 263
|
| 2719 |
+
},
|
| 2720 |
+
{
|
| 2721 |
+
"loss": 0.0001,
|
| 2722 |
+
"learning_rate": 0.0002,
|
| 2723 |
+
"epoch": 4.63,
|
| 2724 |
+
"step": 264
|
| 2725 |
+
},
|
| 2726 |
+
{
|
| 2727 |
+
"loss": 0.0001,
|
| 2728 |
+
"learning_rate": 0.0002,
|
| 2729 |
+
"epoch": 4.65,
|
| 2730 |
+
"step": 265
|
| 2731 |
+
},
|
| 2732 |
+
{
|
| 2733 |
+
"loss": 0.0005,
|
| 2734 |
+
"learning_rate": 0.0002,
|
| 2735 |
+
"epoch": 4.67,
|
| 2736 |
+
"step": 266
|
| 2737 |
+
},
|
| 2738 |
+
{
|
| 2739 |
+
"loss": 0.0003,
|
| 2740 |
+
"learning_rate": 0.0002,
|
| 2741 |
+
"epoch": 4.68,
|
| 2742 |
+
"step": 267
|
| 2743 |
+
},
|
| 2744 |
+
{
|
| 2745 |
+
"loss": 0.0,
|
| 2746 |
+
"learning_rate": 0.0002,
|
| 2747 |
+
"epoch": 4.7,
|
| 2748 |
+
"step": 268
|
| 2749 |
+
},
|
| 2750 |
+
{
|
| 2751 |
+
"loss": 0.0,
|
| 2752 |
+
"learning_rate": 0.0002,
|
| 2753 |
+
"epoch": 4.72,
|
| 2754 |
+
"step": 269
|
| 2755 |
+
},
|
| 2756 |
+
{
|
| 2757 |
+
"loss": 0.0,
|
| 2758 |
+
"learning_rate": 0.0002,
|
| 2759 |
+
"epoch": 4.74,
|
| 2760 |
+
"step": 270
|
| 2761 |
+
},
|
| 2762 |
+
{
|
| 2763 |
+
"loss": 0.0013,
|
| 2764 |
+
"learning_rate": 0.0002,
|
| 2765 |
+
"epoch": 4.75,
|
| 2766 |
+
"step": 271
|
| 2767 |
+
},
|
| 2768 |
+
{
|
| 2769 |
+
"loss": 0.006,
|
| 2770 |
+
"learning_rate": 0.0002,
|
| 2771 |
+
"epoch": 4.77,
|
| 2772 |
+
"step": 272
|
| 2773 |
+
},
|
| 2774 |
+
{
|
| 2775 |
+
"loss": 0.0002,
|
| 2776 |
+
"learning_rate": 0.0002,
|
| 2777 |
+
"epoch": 4.79,
|
| 2778 |
+
"step": 273
|
| 2779 |
+
},
|
| 2780 |
+
{
|
| 2781 |
+
"loss": 0.0008,
|
| 2782 |
+
"learning_rate": 0.0002,
|
| 2783 |
+
"epoch": 4.81,
|
| 2784 |
+
"step": 274
|
| 2785 |
+
},
|
| 2786 |
+
{
|
| 2787 |
+
"loss": 0.0,
|
| 2788 |
+
"learning_rate": 0.0002,
|
| 2789 |
+
"epoch": 4.82,
|
| 2790 |
+
"step": 275
|
| 2791 |
+
},
|
| 2792 |
+
{
|
| 2793 |
+
"eval_code_loss": 0.18856269121170044,
|
| 2794 |
+
"eval_code_score": -0.03526609018445015,
|
| 2795 |
+
"eval_code_brier_score": 0.03526609018445015,
|
| 2796 |
+
"eval_code_average_probability": 0.949932873249054,
|
| 2797 |
+
"eval_code_accuracy": 0.96,
|
| 2798 |
+
"eval_code_probabilities": [
|
| 2799 |
+
1.0,
|
| 2800 |
+
1.0,
|
| 2801 |
+
1.0,
|
| 2802 |
+
1.0,
|
| 2803 |
+
1.0,
|
| 2804 |
+
1.0,
|
| 2805 |
+
1.0,
|
| 2806 |
+
1.0,
|
| 2807 |
+
1.0,
|
| 2808 |
+
1.0,
|
| 2809 |
+
1.0,
|
| 2810 |
+
1.0,
|
| 2811 |
+
1.0,
|
| 2812 |
+
1.0,
|
| 2813 |
+
1.0,
|
| 2814 |
+
0.9667010307312012,
|
| 2815 |
+
0.4576479494571686,
|
| 2816 |
+
0.9961231350898743,
|
| 2817 |
+
1.0,
|
| 2818 |
+
1.0,
|
| 2819 |
+
1.0,
|
| 2820 |
+
1.0,
|
| 2821 |
+
1.0,
|
| 2822 |
+
1.0,
|
| 2823 |
+
1.0,
|
| 2824 |
+
1.0,
|
| 2825 |
+
1.0,
|
| 2826 |
+
0.7199408411979675,
|
| 2827 |
+
0.8098323345184326,
|
| 2828 |
+
0.8810715079307556,
|
| 2829 |
+
0.00041508462163619697,
|
| 2830 |
+
1.0,
|
| 2831 |
+
1.0,
|
| 2832 |
+
1.0,
|
| 2833 |
+
0.9999997615814209,
|
| 2834 |
+
1.0,
|
| 2835 |
+
1.0,
|
| 2836 |
+
1.0,
|
| 2837 |
+
1.0,
|
| 2838 |
+
1.0,
|
| 2839 |
+
1.0,
|
| 2840 |
+
1.0,
|
| 2841 |
+
1.0,
|
| 2842 |
+
1.0,
|
| 2843 |
+
1.0,
|
| 2844 |
+
1.0,
|
| 2845 |
+
1.0,
|
| 2846 |
+
1.0,
|
| 2847 |
+
1.0,
|
| 2848 |
+
1.0,
|
| 2849 |
+
1.0,
|
| 2850 |
+
1.0,
|
| 2851 |
+
1.0,
|
| 2852 |
+
1.0,
|
| 2853 |
+
0.9999830722808838,
|
| 2854 |
+
0.9893618226051331,
|
| 2855 |
+
0.948550820350647,
|
| 2856 |
+
1.0,
|
| 2857 |
+
0.9997448325157166,
|
| 2858 |
+
1.0,
|
| 2859 |
+
1.0,
|
| 2860 |
+
1.0,
|
| 2861 |
+
1.0,
|
| 2862 |
+
1.0,
|
| 2863 |
+
1.0,
|
| 2864 |
+
0.9999991655349731,
|
| 2865 |
+
1.0,
|
| 2866 |
+
0.9401667714118958,
|
| 2867 |
+
0.9961704611778259,
|
| 2868 |
+
0.9991104006767273,
|
| 2869 |
+
0.0038628315087407827,
|
| 2870 |
+
0.9999805688858032,
|
| 2871 |
+
0.9974443912506104,
|
| 2872 |
+
0.9999669790267944,
|
| 2873 |
+
0.9498675465583801,
|
| 2874 |
+
1.0,
|
| 2875 |
+
1.0,
|
| 2876 |
+
1.0,
|
| 2877 |
+
1.0,
|
| 2878 |
+
1.0,
|
| 2879 |
+
1.0,
|
| 2880 |
+
0.9999998807907104,
|
| 2881 |
+
0.9562873244285583,
|
| 2882 |
+
0.9999939203262329,
|
| 2883 |
+
1.0,
|
| 2884 |
+
0.9999998807907104,
|
| 2885 |
+
1.0,
|
| 2886 |
+
0.9999774694442749,
|
| 2887 |
+
0.9999997615814209,
|
| 2888 |
+
0.9999823570251465,
|
| 2889 |
+
1.0,
|
| 2890 |
+
0.9991405010223389,
|
| 2891 |
+
1.0,
|
| 2892 |
+
0.027186574414372444,
|
| 2893 |
+
0.9999986886978149,
|
| 2894 |
+
0.9999057054519653,
|
| 2895 |
+
0.9804957509040833,
|
| 2896 |
+
0.7432851791381836,
|
| 2897 |
+
0.7164115905761719,
|
| 2898 |
+
0.9146708250045776
|
| 2899 |
+
],
|
| 2900 |
+
"eval_code_runtime": 104.8208,
|
| 2901 |
+
"eval_code_samples_per_second": 0.954,
|
| 2902 |
+
"eval_code_steps_per_second": 0.038,
|
| 2903 |
+
"epoch": 4.82,
|
| 2904 |
+
"step": 275
|
| 2905 |
+
},
|
| 2906 |
+
{
|
| 2907 |
+
"loss": 0.0018,
|
| 2908 |
+
"learning_rate": 0.0002,
|
| 2909 |
+
"epoch": 4.84,
|
| 2910 |
+
"step": 276
|
| 2911 |
+
},
|
| 2912 |
+
{
|
| 2913 |
+
"loss": 0.0001,
|
| 2914 |
+
"learning_rate": 0.0002,
|
| 2915 |
+
"epoch": 4.86,
|
| 2916 |
+
"step": 277
|
| 2917 |
+
},
|
| 2918 |
+
{
|
| 2919 |
+
"loss": 0.0005,
|
| 2920 |
+
"learning_rate": 0.0002,
|
| 2921 |
+
"epoch": 4.88,
|
| 2922 |
+
"step": 278
|
| 2923 |
+
},
|
| 2924 |
+
{
|
| 2925 |
+
"loss": 0.0019,
|
| 2926 |
+
"learning_rate": 0.0002,
|
| 2927 |
+
"epoch": 4.89,
|
| 2928 |
+
"step": 279
|
| 2929 |
+
},
|
| 2930 |
+
{
|
| 2931 |
+
"loss": 0.0009,
|
| 2932 |
+
"learning_rate": 0.0002,
|
| 2933 |
+
"epoch": 4.91,
|
| 2934 |
+
"step": 280
|
| 2935 |
+
},
|
| 2936 |
+
{
|
| 2937 |
+
"loss": 0.0,
|
| 2938 |
+
"learning_rate": 0.0002,
|
| 2939 |
+
"epoch": 4.93,
|
| 2940 |
+
"step": 281
|
| 2941 |
+
},
|
| 2942 |
+
{
|
| 2943 |
+
"loss": 0.0008,
|
| 2944 |
+
"learning_rate": 0.0002,
|
| 2945 |
+
"epoch": 4.95,
|
| 2946 |
+
"step": 282
|
| 2947 |
+
},
|
| 2948 |
+
{
|
| 2949 |
+
"loss": 0.0001,
|
| 2950 |
+
"learning_rate": 0.0002,
|
| 2951 |
+
"epoch": 4.96,
|
| 2952 |
+
"step": 283
|
| 2953 |
+
},
|
| 2954 |
+
{
|
| 2955 |
+
"loss": 0.0,
|
| 2956 |
+
"learning_rate": 0.0002,
|
| 2957 |
+
"epoch": 4.98,
|
| 2958 |
+
"step": 284
|
| 2959 |
+
},
|
| 2960 |
+
{
|
| 2961 |
+
"loss": 0.0,
|
| 2962 |
+
"learning_rate": 0.0002,
|
| 2963 |
+
"epoch": 5.0,
|
| 2964 |
+
"step": 285
|
| 2965 |
+
},
|
| 2966 |
+
{
|
| 2967 |
+
"loss": 0.0021,
|
| 2968 |
+
"learning_rate": 0.0002,
|
| 2969 |
+
"epoch": 5.02,
|
| 2970 |
+
"step": 286
|
| 2971 |
+
},
|
| 2972 |
+
{
|
| 2973 |
+
"loss": 0.0,
|
| 2974 |
+
"learning_rate": 0.0002,
|
| 2975 |
+
"epoch": 5.04,
|
| 2976 |
+
"step": 287
|
| 2977 |
+
},
|
| 2978 |
+
{
|
| 2979 |
+
"loss": 0.0,
|
| 2980 |
+
"learning_rate": 0.0002,
|
| 2981 |
+
"epoch": 5.05,
|
| 2982 |
+
"step": 288
|
| 2983 |
+
},
|
| 2984 |
+
{
|
| 2985 |
+
"loss": 0.0003,
|
| 2986 |
+
"learning_rate": 0.0002,
|
| 2987 |
+
"epoch": 5.07,
|
| 2988 |
+
"step": 289
|
| 2989 |
+
},
|
| 2990 |
+
{
|
| 2991 |
+
"loss": 0.0005,
|
| 2992 |
+
"learning_rate": 0.0002,
|
| 2993 |
+
"epoch": 5.09,
|
| 2994 |
+
"step": 290
|
| 2995 |
+
},
|
| 2996 |
+
{
|
| 2997 |
+
"loss": 0.0005,
|
| 2998 |
+
"learning_rate": 0.0002,
|
| 2999 |
+
"epoch": 5.11,
|
| 3000 |
+
"step": 291
|
| 3001 |
+
},
|
| 3002 |
+
{
|
| 3003 |
+
"loss": 0.0,
|
| 3004 |
+
"learning_rate": 0.0002,
|
| 3005 |
+
"epoch": 5.12,
|
| 3006 |
+
"step": 292
|
| 3007 |
+
},
|
| 3008 |
+
{
|
| 3009 |
+
"loss": 0.0,
|
| 3010 |
+
"learning_rate": 0.0002,
|
| 3011 |
+
"epoch": 5.14,
|
| 3012 |
+
"step": 293
|
| 3013 |
+
},
|
| 3014 |
+
{
|
| 3015 |
+
"loss": 0.0,
|
| 3016 |
+
"learning_rate": 0.0002,
|
| 3017 |
+
"epoch": 5.16,
|
| 3018 |
+
"step": 294
|
| 3019 |
+
},
|
| 3020 |
+
{
|
| 3021 |
+
"loss": 0.0,
|
| 3022 |
+
"learning_rate": 0.0002,
|
| 3023 |
+
"epoch": 5.18,
|
| 3024 |
+
"step": 295
|
| 3025 |
+
},
|
| 3026 |
+
{
|
| 3027 |
+
"loss": 0.0,
|
| 3028 |
+
"learning_rate": 0.0002,
|
| 3029 |
+
"epoch": 5.19,
|
| 3030 |
+
"step": 296
|
| 3031 |
+
},
|
| 3032 |
+
{
|
| 3033 |
+
"loss": 0.0001,
|
| 3034 |
+
"learning_rate": 0.0002,
|
| 3035 |
+
"epoch": 5.21,
|
| 3036 |
+
"step": 297
|
| 3037 |
+
},
|
| 3038 |
+
{
|
| 3039 |
+
"loss": 0.0001,
|
| 3040 |
+
"learning_rate": 0.0002,
|
| 3041 |
+
"epoch": 5.23,
|
| 3042 |
+
"step": 298
|
| 3043 |
+
},
|
| 3044 |
+
{
|
| 3045 |
+
"loss": 0.0008,
|
| 3046 |
+
"learning_rate": 0.0002,
|
| 3047 |
+
"epoch": 5.25,
|
| 3048 |
+
"step": 299
|
| 3049 |
+
},
|
| 3050 |
+
{
|
| 3051 |
+
"loss": 0.0001,
|
| 3052 |
+
"learning_rate": 0.0002,
|
| 3053 |
+
"epoch": 5.26,
|
| 3054 |
+
"step": 300
|
| 3055 |
+
},
|
| 3056 |
+
{
|
| 3057 |
+
"eval_code_loss": 0.17653299868106842,
|
| 3058 |
+
"eval_code_score": -0.035850197076797485,
|
| 3059 |
+
"eval_code_brier_score": 0.035850197076797485,
|
| 3060 |
+
"eval_code_average_probability": 0.949984610080719,
|
| 3061 |
+
"eval_code_accuracy": 0.96,
|
| 3062 |
+
"eval_code_probabilities": [
|
| 3063 |
+
1.0,
|
| 3064 |
+
1.0,
|
| 3065 |
+
1.0,
|
| 3066 |
+
1.0,
|
| 3067 |
+
1.0,
|
| 3068 |
+
1.0,
|
| 3069 |
+
1.0,
|
| 3070 |
+
1.0,
|
| 3071 |
+
1.0,
|
| 3072 |
+
1.0,
|
| 3073 |
+
1.0,
|
| 3074 |
+
1.0,
|
| 3075 |
+
1.0,
|
| 3076 |
+
1.0,
|
| 3077 |
+
1.0,
|
| 3078 |
+
0.9785318374633789,
|
| 3079 |
+
0.42346829175949097,
|
| 3080 |
+
0.9971503615379333,
|
| 3081 |
+
1.0,
|
| 3082 |
+
1.0,
|
| 3083 |
+
0.9999997615814209,
|
| 3084 |
+
1.0,
|
| 3085 |
+
1.0,
|
| 3086 |
+
1.0,
|
| 3087 |
+
1.0,
|
| 3088 |
+
1.0,
|
| 3089 |
+
1.0,
|
| 3090 |
+
0.7221306562423706,
|
| 3091 |
+
0.8104501962661743,
|
| 3092 |
+
0.8751990795135498,
|
| 3093 |
+
0.0004177717200946063,
|
| 3094 |
+
1.0,
|
| 3095 |
+
1.0,
|
| 3096 |
+
1.0,
|
| 3097 |
+
0.9999998807907104,
|
| 3098 |
+
1.0,
|
| 3099 |
+
1.0,
|
| 3100 |
+
1.0,
|
| 3101 |
+
1.0,
|
| 3102 |
+
1.0,
|
| 3103 |
+
1.0,
|
| 3104 |
+
1.0,
|
| 3105 |
+
1.0,
|
| 3106 |
+
1.0,
|
| 3107 |
+
1.0,
|
| 3108 |
+
1.0,
|
| 3109 |
+
1.0,
|
| 3110 |
+
1.0,
|
| 3111 |
+
1.0,
|
| 3112 |
+
1.0,
|
| 3113 |
+
1.0,
|
| 3114 |
+
1.0,
|
| 3115 |
+
1.0,
|
| 3116 |
+
1.0,
|
| 3117 |
+
0.9999911785125732,
|
| 3118 |
+
0.9935014843940735,
|
| 3119 |
+
0.9632013440132141,
|
| 3120 |
+
1.0,
|
| 3121 |
+
0.9997548460960388,
|
| 3122 |
+
1.0,
|
| 3123 |
+
1.0,
|
| 3124 |
+
1.0,
|
| 3125 |
+
1.0,
|
| 3126 |
+
1.0,
|
| 3127 |
+
1.0,
|
| 3128 |
+
0.999998927116394,
|
| 3129 |
+
1.0,
|
| 3130 |
+
0.9625283479690552,
|
| 3131 |
+
0.9952881336212158,
|
| 3132 |
+
0.9994906187057495,
|
| 3133 |
+
0.011133184656500816,
|
| 3134 |
+
0.9999599456787109,
|
| 3135 |
+
0.9987699389457703,
|
| 3136 |
+
0.9999761581420898,
|
| 3137 |
+
0.9492788314819336,
|
| 3138 |
+
1.0,
|
| 3139 |
+
0.9999998807907104,
|
| 3140 |
+
1.0,
|
| 3141 |
+
1.0,
|
| 3142 |
+
1.0,
|
| 3143 |
+
1.0,
|
| 3144 |
+
1.0,
|
| 3145 |
+
0.9763202667236328,
|
| 3146 |
+
0.9999985694885254,
|
| 3147 |
+
1.0,
|
| 3148 |
+
0.9999996423721313,
|
| 3149 |
+
1.0,
|
| 3150 |
+
0.9999876022338867,
|
| 3151 |
+
0.9999998807907104,
|
| 3152 |
+
0.9999854564666748,
|
| 3153 |
+
1.0,
|
| 3154 |
+
0.9988405108451843,
|
| 3155 |
+
1.0,
|
| 3156 |
+
0.03199751302599907,
|
| 3157 |
+
0.9999868869781494,
|
| 3158 |
+
0.9996381998062134,
|
| 3159 |
+
0.9803478121757507,
|
| 3160 |
+
0.6885609030723572,
|
| 3161 |
+
0.6760653853416443,
|
| 3162 |
+
0.9665107727050781
|
| 3163 |
+
],
|
| 3164 |
+
"eval_code_runtime": 104.8074,
|
| 3165 |
+
"eval_code_samples_per_second": 0.954,
|
| 3166 |
+
"eval_code_steps_per_second": 0.038,
|
| 3167 |
+
"epoch": 5.26,
|
| 3168 |
+
"step": 300
|
| 3169 |
+
},
|
| 3170 |
+
{
|
| 3171 |
+
"train_runtime": 13610.3633,
|
| 3172 |
+
"train_samples_per_second": 0.705,
|
| 3173 |
+
"train_steps_per_second": 0.022,
|
| 3174 |
+
"total_flos": 0.0,
|
| 3175 |
+
"train_loss": 0.002647306595269659,
|
| 3176 |
+
"epoch": 5.26,
|
| 3177 |
+
"step": 300
|
| 3178 |
+
}
|
| 3179 |
+
]
|