Commit
•
b15d515
1
Parent(s):
537144c
Upload folder using huggingface_hub
Browse files- README.md +9 -0
- adapter_config.json +26 -0
- adapter_model.bin +3 -0
- config.json +45 -0
- non_lora_trainables.bin +3 -0
- trainer_state.json +445 -0
README.md
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
library_name: peft
|
| 3 |
+
---
|
| 4 |
+
## Training procedure
|
| 5 |
+
|
| 6 |
+
### Framework versions
|
| 7 |
+
|
| 8 |
+
|
| 9 |
+
- PEFT 0.4.0
|
adapter_config.json
ADDED
|
@@ -0,0 +1,26 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"auto_mapping": null,
|
| 3 |
+
"base_model_name_or_path": "/home/ubuntu/LLaVA/checkpoints/llava-v1.5-13b-steer-lora-helpsteerformat-interleaf-lrv3067llava3067-likert-1222_interleavedgpt4v_responseonly-pure-merged",
|
| 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": 256,
|
| 11 |
+
"lora_dropout": 0.05,
|
| 12 |
+
"modules_to_save": null,
|
| 13 |
+
"peft_type": "LORA",
|
| 14 |
+
"r": 128,
|
| 15 |
+
"revision": null,
|
| 16 |
+
"target_modules": [
|
| 17 |
+
"v_proj",
|
| 18 |
+
"up_proj",
|
| 19 |
+
"o_proj",
|
| 20 |
+
"q_proj",
|
| 21 |
+
"down_proj",
|
| 22 |
+
"k_proj",
|
| 23 |
+
"gate_proj"
|
| 24 |
+
],
|
| 25 |
+
"task_type": "CAUSAL_LM"
|
| 26 |
+
}
|
adapter_model.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8d065e19e157f93c9160b378365b1bc385869cc6031e1fe7aa96574cfd73ad61
|
| 3 |
+
size 1001584333
|
config.json
ADDED
|
@@ -0,0 +1,45 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_name_or_path": "/home/ubuntu/LLaVA/checkpoints/llava-v1.5-13b-steer-lora-helpsteerformat-interleaf-lrv3067llava3067-likert-1222_interleavedgpt4v_responseonly-pure-merged",
|
| 3 |
+
"architectures": [
|
| 4 |
+
"LlavaLlamaForCausalLM"
|
| 5 |
+
],
|
| 6 |
+
"bos_token_id": 1,
|
| 7 |
+
"eos_token_id": 2,
|
| 8 |
+
"freeze_mm_mlp_adapter": false,
|
| 9 |
+
"freeze_mm_vision_resampler": false,
|
| 10 |
+
"hidden_act": "silu",
|
| 11 |
+
"hidden_size": 5120,
|
| 12 |
+
"image_aspect_ratio": "pad",
|
| 13 |
+
"initializer_range": 0.02,
|
| 14 |
+
"intermediate_size": 13824,
|
| 15 |
+
"max_length": 4096,
|
| 16 |
+
"max_position_embeddings": 4096,
|
| 17 |
+
"mm_hidden_size": 1024,
|
| 18 |
+
"mm_projector_lr": 2e-05,
|
| 19 |
+
"mm_projector_type": "mlp2x_gelu",
|
| 20 |
+
"mm_resampler_type": null,
|
| 21 |
+
"mm_use_im_patch_token": false,
|
| 22 |
+
"mm_use_im_start_end": false,
|
| 23 |
+
"mm_vision_select_feature": "patch",
|
| 24 |
+
"mm_vision_select_layer": -2,
|
| 25 |
+
"mm_vision_tower": "openai/clip-vit-large-patch14-336",
|
| 26 |
+
"model_type": "llava",
|
| 27 |
+
"num_attention_heads": 40,
|
| 28 |
+
"num_hidden_layers": 40,
|
| 29 |
+
"num_key_value_heads": 40,
|
| 30 |
+
"pad_token_id": 0,
|
| 31 |
+
"pretraining_tp": 1,
|
| 32 |
+
"rms_norm_eps": 1e-05,
|
| 33 |
+
"rope_scaling": null,
|
| 34 |
+
"tie_word_embeddings": false,
|
| 35 |
+
"tokenizer_model_max_length": 2048,
|
| 36 |
+
"tokenizer_padding_side": "right",
|
| 37 |
+
"torch_dtype": "float16",
|
| 38 |
+
"transformers_version": "4.31.0",
|
| 39 |
+
"tune_mm_mlp_adapter": false,
|
| 40 |
+
"tune_mm_vision_resampler": false,
|
| 41 |
+
"unfreeze_mm_vision_tower": false,
|
| 42 |
+
"use_cache": true,
|
| 43 |
+
"use_mm_proj": true,
|
| 44 |
+
"vocab_size": 32000
|
| 45 |
+
}
|
non_lora_trainables.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d08a173a526e8a979404fc833be6312a1638363d2545270d93170a3e3ebf41c8
|
| 3 |
+
size 62936807
|
trainer_state.json
ADDED
|
@@ -0,0 +1,445 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 1.8666666666666667,
|
| 5 |
+
"global_step": 14,
|
| 6 |
+
"is_hyper_param_search": false,
|
| 7 |
+
"is_local_process_zero": true,
|
| 8 |
+
"is_world_process_zero": true,
|
| 9 |
+
"log_history": [
|
| 10 |
+
{
|
| 11 |
+
"epoch": 0,
|
| 12 |
+
"logps_train/chosen": -100.49485778808594,
|
| 13 |
+
"logps_train/ref_chosen": -100.5,
|
| 14 |
+
"logps_train/ref_rejected": -105.0,
|
| 15 |
+
"logps_train/rejected": -104.80752563476562,
|
| 16 |
+
"rewards_train/accuracies": 0.46875,
|
| 17 |
+
"rewards_train/chosen": -0.020498279482126236,
|
| 18 |
+
"rewards_train/margins": -0.021032828837633133,
|
| 19 |
+
"rewards_train/rejected": 0.000534549355506897,
|
| 20 |
+
"step": 0
|
| 21 |
+
},
|
| 22 |
+
{
|
| 23 |
+
"epoch": 0,
|
| 24 |
+
"logps_train/chosen": -89.90950012207031,
|
| 25 |
+
"logps_train/ref_chosen": -90.0,
|
| 26 |
+
"logps_train/ref_rejected": -101.0,
|
| 27 |
+
"logps_train/rejected": -100.86872863769531,
|
| 28 |
+
"rewards_train/accuracies": 0.5,
|
| 29 |
+
"rewards_train/chosen": 0.0191262885928154,
|
| 30 |
+
"rewards_train/margins": 0.0020734891295433044,
|
| 31 |
+
"rewards_train/rejected": 0.017052799463272095,
|
| 32 |
+
"step": 0
|
| 33 |
+
},
|
| 34 |
+
{
|
| 35 |
+
"epoch": 0.13,
|
| 36 |
+
"learning_rate": 5e-06,
|
| 37 |
+
"loss": 0.7043,
|
| 38 |
+
"step": 1
|
| 39 |
+
},
|
| 40 |
+
{
|
| 41 |
+
"epoch": 0.13,
|
| 42 |
+
"logps_train/chosen": -104.38166809082031,
|
| 43 |
+
"logps_train/ref_chosen": -104.5,
|
| 44 |
+
"logps_train/ref_rejected": -98.0,
|
| 45 |
+
"logps_train/rejected": -97.89907836914062,
|
| 46 |
+
"rewards_train/accuracies": 0.515625,
|
| 47 |
+
"rewards_train/chosen": 0.03059956058859825,
|
| 48 |
+
"rewards_train/margins": 0.004916800186038017,
|
| 49 |
+
"rewards_train/rejected": 0.025682760402560234,
|
| 50 |
+
"step": 1
|
| 51 |
+
},
|
| 52 |
+
{
|
| 53 |
+
"epoch": 0.13,
|
| 54 |
+
"logps_train/chosen": -94.88069152832031,
|
| 55 |
+
"logps_train/ref_chosen": -95.0,
|
| 56 |
+
"logps_train/ref_rejected": -95.0,
|
| 57 |
+
"logps_train/rejected": -95.29293823242188,
|
| 58 |
+
"rewards_train/accuracies": 0.546875,
|
| 59 |
+
"rewards_train/chosen": -0.009438544511795044,
|
| 60 |
+
"rewards_train/margins": -0.020688317716121674,
|
| 61 |
+
"rewards_train/rejected": 0.01124977320432663,
|
| 62 |
+
"step": 1
|
| 63 |
+
},
|
| 64 |
+
{
|
| 65 |
+
"epoch": 0.27,
|
| 66 |
+
"learning_rate": 4.927354543565131e-06,
|
| 67 |
+
"loss": 0.7061,
|
| 68 |
+
"step": 2
|
| 69 |
+
},
|
| 70 |
+
{
|
| 71 |
+
"epoch": 0.27,
|
| 72 |
+
"logps_train/chosen": -82.45372009277344,
|
| 73 |
+
"logps_train/ref_chosen": -82.5,
|
| 74 |
+
"logps_train/ref_rejected": -89.0,
|
| 75 |
+
"logps_train/rejected": -89.00653839111328,
|
| 76 |
+
"rewards_train/accuracies": 0.5625,
|
| 77 |
+
"rewards_train/chosen": 0.0465165413916111,
|
| 78 |
+
"rewards_train/margins": 0.0453295623883605,
|
| 79 |
+
"rewards_train/rejected": 0.001186979003250599,
|
| 80 |
+
"step": 2
|
| 81 |
+
},
|
| 82 |
+
{
|
| 83 |
+
"epoch": 0.27,
|
| 84 |
+
"logps_train/chosen": -100.2186508178711,
|
| 85 |
+
"logps_train/ref_chosen": -100.0,
|
| 86 |
+
"logps_train/ref_rejected": -106.0,
|
| 87 |
+
"logps_train/rejected": -106.00666809082031,
|
| 88 |
+
"rewards_train/accuracies": 0.59375,
|
| 89 |
+
"rewards_train/chosen": -0.029247179627418518,
|
| 90 |
+
"rewards_train/margins": -0.0325072705745697,
|
| 91 |
+
"rewards_train/rejected": 0.003260090947151184,
|
| 92 |
+
"step": 2
|
| 93 |
+
},
|
| 94 |
+
{
|
| 95 |
+
"epoch": 0.4,
|
| 96 |
+
"learning_rate": 4.7136400641330245e-06,
|
| 97 |
+
"loss": 0.6969,
|
| 98 |
+
"step": 3
|
| 99 |
+
},
|
| 100 |
+
{
|
| 101 |
+
"epoch": 0.4,
|
| 102 |
+
"logps_train/chosen": -85.843994140625,
|
| 103 |
+
"logps_train/ref_chosen": -86.0,
|
| 104 |
+
"logps_train/ref_rejected": -90.0,
|
| 105 |
+
"logps_train/rejected": -90.01055908203125,
|
| 106 |
+
"rewards_train/accuracies": 0.5625,
|
| 107 |
+
"rewards_train/chosen": 0.0008565783500671387,
|
| 108 |
+
"rewards_train/margins": -0.012054374441504478,
|
| 109 |
+
"rewards_train/rejected": 0.012910952791571617,
|
| 110 |
+
"step": 3
|
| 111 |
+
},
|
| 112 |
+
{
|
| 113 |
+
"epoch": 0.4,
|
| 114 |
+
"logps_train/chosen": -92.26002502441406,
|
| 115 |
+
"logps_train/ref_chosen": -92.0,
|
| 116 |
+
"logps_train/ref_rejected": -77.5,
|
| 117 |
+
"logps_train/rejected": -77.61531829833984,
|
| 118 |
+
"rewards_train/accuracies": 0.546875,
|
| 119 |
+
"rewards_train/chosen": 0.024773243814706802,
|
| 120 |
+
"rewards_train/margins": 0.04251495748758316,
|
| 121 |
+
"rewards_train/rejected": -0.017741713672876358,
|
| 122 |
+
"step": 3
|
| 123 |
+
},
|
| 124 |
+
{
|
| 125 |
+
"epoch": 0.53,
|
| 126 |
+
"learning_rate": 4.3712768704277535e-06,
|
| 127 |
+
"loss": 0.6936,
|
| 128 |
+
"step": 4
|
| 129 |
+
},
|
| 130 |
+
{
|
| 131 |
+
"epoch": 0.53,
|
| 132 |
+
"logps_train/chosen": -88.51109313964844,
|
| 133 |
+
"logps_train/ref_chosen": -88.5,
|
| 134 |
+
"logps_train/ref_rejected": -91.5,
|
| 135 |
+
"logps_train/rejected": -91.55830383300781,
|
| 136 |
+
"rewards_train/accuracies": 0.53125,
|
| 137 |
+
"rewards_train/chosen": 0.044210705906152725,
|
| 138 |
+
"rewards_train/margins": 0.0598596166819334,
|
| 139 |
+
"rewards_train/rejected": -0.015648910775780678,
|
| 140 |
+
"step": 4
|
| 141 |
+
},
|
| 142 |
+
{
|
| 143 |
+
"epoch": 0.53,
|
| 144 |
+
"logps_train/chosen": -104.84616088867188,
|
| 145 |
+
"logps_train/ref_chosen": -105.0,
|
| 146 |
+
"logps_train/ref_rejected": -104.5,
|
| 147 |
+
"logps_train/rejected": -104.39971923828125,
|
| 148 |
+
"rewards_train/accuracies": 0.515625,
|
| 149 |
+
"rewards_train/chosen": 0.04029738903045654,
|
| 150 |
+
"rewards_train/margins": 0.04008368402719498,
|
| 151 |
+
"rewards_train/rejected": 0.00021370500326156616,
|
| 152 |
+
"step": 4
|
| 153 |
+
},
|
| 154 |
+
{
|
| 155 |
+
"epoch": 0.67,
|
| 156 |
+
"learning_rate": 3.92016186682789e-06,
|
| 157 |
+
"loss": 0.6761,
|
| 158 |
+
"step": 5
|
| 159 |
+
},
|
| 160 |
+
{
|
| 161 |
+
"epoch": 0.67,
|
| 162 |
+
"logps_train/chosen": -109.26580047607422,
|
| 163 |
+
"logps_train/ref_chosen": -109.0,
|
| 164 |
+
"logps_train/ref_rejected": -99.5,
|
| 165 |
+
"logps_train/rejected": -99.34159851074219,
|
| 166 |
+
"rewards_train/accuracies": 0.40625,
|
| 167 |
+
"rewards_train/chosen": 0.017489098012447357,
|
| 168 |
+
"rewards_train/margins": -0.0021405071020126343,
|
| 169 |
+
"rewards_train/rejected": 0.01962960511445999,
|
| 170 |
+
"step": 5
|
| 171 |
+
},
|
| 172 |
+
{
|
| 173 |
+
"epoch": 0.67,
|
| 174 |
+
"logps_train/chosen": -82.48857116699219,
|
| 175 |
+
"logps_train/ref_chosen": -82.5,
|
| 176 |
+
"logps_train/ref_rejected": -80.0,
|
| 177 |
+
"logps_train/rejected": -79.8382568359375,
|
| 178 |
+
"rewards_train/accuracies": 0.46875,
|
| 179 |
+
"rewards_train/chosen": 0.05613037571310997,
|
| 180 |
+
"rewards_train/margins": 0.01352725550532341,
|
| 181 |
+
"rewards_train/rejected": 0.04260312020778656,
|
| 182 |
+
"step": 5
|
| 183 |
+
},
|
| 184 |
+
{
|
| 185 |
+
"epoch": 0.8,
|
| 186 |
+
"learning_rate": 3.386512217606339e-06,
|
| 187 |
+
"loss": 0.6952,
|
| 188 |
+
"step": 6
|
| 189 |
+
},
|
| 190 |
+
{
|
| 191 |
+
"epoch": 0.8,
|
| 192 |
+
"logps_train/chosen": -92.07105255126953,
|
| 193 |
+
"logps_train/ref_chosen": -92.0,
|
| 194 |
+
"logps_train/ref_rejected": -89.5,
|
| 195 |
+
"logps_train/rejected": -89.48802185058594,
|
| 196 |
+
"rewards_train/accuracies": 0.53125,
|
| 197 |
+
"rewards_train/chosen": 0.07213925570249557,
|
| 198 |
+
"rewards_train/margins": -0.003675796091556549,
|
| 199 |
+
"rewards_train/rejected": 0.07581505179405212,
|
| 200 |
+
"step": 6
|
| 201 |
+
},
|
| 202 |
+
{
|
| 203 |
+
"epoch": 0.8,
|
| 204 |
+
"logps_train/chosen": -88.25656127929688,
|
| 205 |
+
"logps_train/ref_chosen": -88.5,
|
| 206 |
+
"logps_train/ref_rejected": -94.0,
|
| 207 |
+
"logps_train/rejected": -93.6957778930664,
|
| 208 |
+
"rewards_train/accuracies": 0.46875,
|
| 209 |
+
"rewards_train/chosen": 0.025041330605745316,
|
| 210 |
+
"rewards_train/margins": -0.009836137294769287,
|
| 211 |
+
"rewards_train/rejected": 0.0348774679005146,
|
| 212 |
+
"step": 6
|
| 213 |
+
},
|
| 214 |
+
{
|
| 215 |
+
"epoch": 0.93,
|
| 216 |
+
"learning_rate": 2.8013417006383078e-06,
|
| 217 |
+
"loss": 0.7033,
|
| 218 |
+
"step": 7
|
| 219 |
+
},
|
| 220 |
+
{
|
| 221 |
+
"epoch": 0.93,
|
| 222 |
+
"logps_train/chosen": -106.97582244873047,
|
| 223 |
+
"logps_train/ref_chosen": -107.0,
|
| 224 |
+
"logps_train/ref_rejected": -104.0,
|
| 225 |
+
"logps_train/rejected": -104.00101470947266,
|
| 226 |
+
"rewards_train/accuracies": 0.78125,
|
| 227 |
+
"rewards_train/chosen": 0.1645529866218567,
|
| 228 |
+
"rewards_train/margins": 0.235374353826046,
|
| 229 |
+
"rewards_train/rejected": -0.0708213672041893,
|
| 230 |
+
"step": 7
|
| 231 |
+
},
|
| 232 |
+
{
|
| 233 |
+
"epoch": 0.93,
|
| 234 |
+
"logps_train/chosen": -95.53459167480469,
|
| 235 |
+
"logps_train/ref_chosen": -96.0,
|
| 236 |
+
"logps_train/ref_rejected": -90.5,
|
| 237 |
+
"logps_train/rejected": -90.36405944824219,
|
| 238 |
+
"rewards_train/accuracies": 0.734375,
|
| 239 |
+
"rewards_train/chosen": 0.18082404136657715,
|
| 240 |
+
"rewards_train/margins": 0.21460064873099327,
|
| 241 |
+
"rewards_train/rejected": -0.03377660736441612,
|
| 242 |
+
"step": 7
|
| 243 |
+
},
|
| 244 |
+
{
|
| 245 |
+
"epoch": 1.07,
|
| 246 |
+
"learning_rate": 2.1986582993616926e-06,
|
| 247 |
+
"loss": 0.5967,
|
| 248 |
+
"step": 8
|
| 249 |
+
},
|
| 250 |
+
{
|
| 251 |
+
"epoch": 1.07,
|
| 252 |
+
"logps_train/chosen": -103.63533020019531,
|
| 253 |
+
"logps_train/ref_chosen": -104.0,
|
| 254 |
+
"logps_train/ref_rejected": -91.0,
|
| 255 |
+
"logps_train/rejected": -91.11985778808594,
|
| 256 |
+
"rewards_train/accuracies": 0.6875,
|
| 257 |
+
"rewards_train/chosen": 0.211873859167099,
|
| 258 |
+
"rewards_train/margins": 0.20722994953393936,
|
| 259 |
+
"rewards_train/rejected": 0.0046439096331596375,
|
| 260 |
+
"step": 8
|
| 261 |
+
},
|
| 262 |
+
{
|
| 263 |
+
"epoch": 1.07,
|
| 264 |
+
"logps_train/chosen": -89.88359069824219,
|
| 265 |
+
"logps_train/ref_chosen": -90.0,
|
| 266 |
+
"logps_train/ref_rejected": -91.0,
|
| 267 |
+
"logps_train/rejected": -91.22651672363281,
|
| 268 |
+
"rewards_train/accuracies": 0.71875,
|
| 269 |
+
"rewards_train/chosen": 0.1387726217508316,
|
| 270 |
+
"rewards_train/margins": 0.168781116604805,
|
| 271 |
+
"rewards_train/rejected": -0.03000849485397339,
|
| 272 |
+
"step": 8
|
| 273 |
+
},
|
| 274 |
+
{
|
| 275 |
+
"epoch": 1.2,
|
| 276 |
+
"learning_rate": 1.613487782393661e-06,
|
| 277 |
+
"loss": 0.6165,
|
| 278 |
+
"step": 9
|
| 279 |
+
},
|
| 280 |
+
{
|
| 281 |
+
"epoch": 1.2,
|
| 282 |
+
"logps_train/chosen": -99.79483795166016,
|
| 283 |
+
"logps_train/ref_chosen": -100.0,
|
| 284 |
+
"logps_train/ref_rejected": -91.5,
|
| 285 |
+
"logps_train/rejected": -91.58244323730469,
|
| 286 |
+
"rewards_train/accuracies": 0.78125,
|
| 287 |
+
"rewards_train/chosen": 0.2085389792919159,
|
| 288 |
+
"rewards_train/margins": 0.2784496992826462,
|
| 289 |
+
"rewards_train/rejected": -0.06991071999073029,
|
| 290 |
+
"step": 9
|
| 291 |
+
},
|
| 292 |
+
{
|
| 293 |
+
"epoch": 1.2,
|
| 294 |
+
"logps_train/chosen": -97.40536499023438,
|
| 295 |
+
"logps_train/ref_chosen": -98.0,
|
| 296 |
+
"logps_train/ref_rejected": -95.5,
|
| 297 |
+
"logps_train/rejected": -95.38594055175781,
|
| 298 |
+
"rewards_train/accuracies": 0.765625,
|
| 299 |
+
"rewards_train/chosen": 0.16743528842926025,
|
| 300 |
+
"rewards_train/margins": 0.1775425188243389,
|
| 301 |
+
"rewards_train/rejected": -0.010107230395078659,
|
| 302 |
+
"step": 9
|
| 303 |
+
},
|
| 304 |
+
{
|
| 305 |
+
"epoch": 1.33,
|
| 306 |
+
"learning_rate": 1.079838133172111e-06,
|
| 307 |
+
"loss": 0.5951,
|
| 308 |
+
"step": 10
|
| 309 |
+
},
|
| 310 |
+
{
|
| 311 |
+
"epoch": 1.33,
|
| 312 |
+
"logps_train/chosen": -99.17311096191406,
|
| 313 |
+
"logps_train/ref_chosen": -99.5,
|
| 314 |
+
"logps_train/ref_rejected": -97.0,
|
| 315 |
+
"logps_train/rejected": -96.72930908203125,
|
| 316 |
+
"rewards_train/accuracies": 0.703125,
|
| 317 |
+
"rewards_train/chosen": 0.21544356644153595,
|
| 318 |
+
"rewards_train/margins": 0.18458961695432663,
|
| 319 |
+
"rewards_train/rejected": 0.03085394948720932,
|
| 320 |
+
"step": 10
|
| 321 |
+
},
|
| 322 |
+
{
|
| 323 |
+
"epoch": 1.33,
|
| 324 |
+
"logps_train/chosen": -90.94091796875,
|
| 325 |
+
"logps_train/ref_chosen": -91.0,
|
| 326 |
+
"logps_train/ref_rejected": -97.5,
|
| 327 |
+
"logps_train/rejected": -97.65658569335938,
|
| 328 |
+
"rewards_train/accuracies": 0.8125,
|
| 329 |
+
"rewards_train/chosen": 0.21173310279846191,
|
| 330 |
+
"rewards_train/margins": 0.18907413445413113,
|
| 331 |
+
"rewards_train/rejected": 0.022658968344330788,
|
| 332 |
+
"step": 10
|
| 333 |
+
},
|
| 334 |
+
{
|
| 335 |
+
"epoch": 1.47,
|
| 336 |
+
"learning_rate": 6.28723129572247e-07,
|
| 337 |
+
"loss": 0.6138,
|
| 338 |
+
"step": 11
|
| 339 |
+
},
|
| 340 |
+
{
|
| 341 |
+
"epoch": 1.47,
|
| 342 |
+
"logps_train/chosen": -96.39669799804688,
|
| 343 |
+
"logps_train/ref_chosen": -97.0,
|
| 344 |
+
"logps_train/ref_rejected": -99.0,
|
| 345 |
+
"logps_train/rejected": -98.8724136352539,
|
| 346 |
+
"rewards_train/accuracies": 0.8125,
|
| 347 |
+
"rewards_train/chosen": 0.16395366191864014,
|
| 348 |
+
"rewards_train/margins": 0.25006232410669327,
|
| 349 |
+
"rewards_train/rejected": -0.08610866218805313,
|
| 350 |
+
"step": 11
|
| 351 |
+
},
|
| 352 |
+
{
|
| 353 |
+
"epoch": 1.47,
|
| 354 |
+
"logps_train/chosen": -80.8447036743164,
|
| 355 |
+
"logps_train/ref_chosen": -81.0,
|
| 356 |
+
"logps_train/ref_rejected": -90.0,
|
| 357 |
+
"logps_train/rejected": -90.13240814208984,
|
| 358 |
+
"rewards_train/accuracies": 0.734375,
|
| 359 |
+
"rewards_train/chosen": 0.13546136021614075,
|
| 360 |
+
"rewards_train/margins": 0.16997329890727997,
|
| 361 |
+
"rewards_train/rejected": -0.03451193869113922,
|
| 362 |
+
"step": 11
|
| 363 |
+
},
|
| 364 |
+
{
|
| 365 |
+
"epoch": 1.6,
|
| 366 |
+
"learning_rate": 2.8635993586697555e-07,
|
| 367 |
+
"loss": 0.6012,
|
| 368 |
+
"step": 12
|
| 369 |
+
},
|
| 370 |
+
{
|
| 371 |
+
"epoch": 1.6,
|
| 372 |
+
"logps_train/chosen": -89.55223846435547,
|
| 373 |
+
"logps_train/ref_chosen": -90.0,
|
| 374 |
+
"logps_train/ref_rejected": -96.0,
|
| 375 |
+
"logps_train/rejected": -95.89089965820312,
|
| 376 |
+
"rewards_train/accuracies": 0.84375,
|
| 377 |
+
"rewards_train/chosen": 0.20703516900539398,
|
| 378 |
+
"rewards_train/margins": 0.23602662980556488,
|
| 379 |
+
"rewards_train/rejected": -0.0289914608001709,
|
| 380 |
+
"step": 12
|
| 381 |
+
},
|
| 382 |
+
{
|
| 383 |
+
"epoch": 1.6,
|
| 384 |
+
"logps_train/chosen": -82.7729263305664,
|
| 385 |
+
"logps_train/ref_chosen": -83.0,
|
| 386 |
+
"logps_train/ref_rejected": -85.0,
|
| 387 |
+
"logps_train/rejected": -85.27727508544922,
|
| 388 |
+
"rewards_train/accuracies": 0.796875,
|
| 389 |
+
"rewards_train/chosen": 0.17896729707717896,
|
| 390 |
+
"rewards_train/margins": 0.2392353191971779,
|
| 391 |
+
"rewards_train/rejected": -0.06026802211999893,
|
| 392 |
+
"step": 12
|
| 393 |
+
},
|
| 394 |
+
{
|
| 395 |
+
"epoch": 1.73,
|
| 396 |
+
"learning_rate": 7.264545643486997e-08,
|
| 397 |
+
"loss": 0.5902,
|
| 398 |
+
"step": 13
|
| 399 |
+
},
|
| 400 |
+
{
|
| 401 |
+
"epoch": 1.73,
|
| 402 |
+
"logps_train/chosen": -104.69845581054688,
|
| 403 |
+
"logps_train/ref_chosen": -105.0,
|
| 404 |
+
"logps_train/ref_rejected": -101.0,
|
| 405 |
+
"logps_train/rejected": -100.98780822753906,
|
| 406 |
+
"rewards_train/accuracies": 0.796875,
|
| 407 |
+
"rewards_train/chosen": 0.19716176390647888,
|
| 408 |
+
"rewards_train/margins": 0.24550895392894745,
|
| 409 |
+
"rewards_train/rejected": -0.04834719002246857,
|
| 410 |
+
"step": 13
|
| 411 |
+
},
|
| 412 |
+
{
|
| 413 |
+
"epoch": 1.73,
|
| 414 |
+
"logps_train/chosen": -90.20387268066406,
|
| 415 |
+
"logps_train/ref_chosen": -91.0,
|
| 416 |
+
"logps_train/ref_rejected": -85.0,
|
| 417 |
+
"logps_train/rejected": -85.12068939208984,
|
| 418 |
+
"rewards_train/accuracies": 0.65625,
|
| 419 |
+
"rewards_train/chosen": 0.1614937037229538,
|
| 420 |
+
"rewards_train/margins": 0.1774044744670391,
|
| 421 |
+
"rewards_train/rejected": -0.015910770744085312,
|
| 422 |
+
"step": 13
|
| 423 |
+
},
|
| 424 |
+
{
|
| 425 |
+
"epoch": 1.87,
|
| 426 |
+
"learning_rate": 0.0,
|
| 427 |
+
"loss": 0.6056,
|
| 428 |
+
"step": 14
|
| 429 |
+
},
|
| 430 |
+
{
|
| 431 |
+
"epoch": 1.87,
|
| 432 |
+
"step": 14,
|
| 433 |
+
"total_flos": 0.0,
|
| 434 |
+
"train_loss": 0.6496244881834302,
|
| 435 |
+
"train_runtime": 164.1376,
|
| 436 |
+
"train_samples_per_second": 11.076,
|
| 437 |
+
"train_steps_per_second": 0.085
|
| 438 |
+
}
|
| 439 |
+
],
|
| 440 |
+
"max_steps": 14,
|
| 441 |
+
"num_train_epochs": 2,
|
| 442 |
+
"total_flos": 0.0,
|
| 443 |
+
"trial_name": null,
|
| 444 |
+
"trial_params": null
|
| 445 |
+
}
|