Upload LoRA adapter (README written by author)
Browse files- README.md +4 -4
- adapter_config.json +7 -7
- adapter_model.safetensors +2 -2
- all_experiments_details.json +434 -0
README.md
CHANGED
|
@@ -13,7 +13,7 @@ tags:
|
|
| 13 |
- structured-output
|
| 14 |
---
|
| 15 |
|
| 16 |
-
|
| 17 |
|
| 18 |
This repository provides a **LoRA adapter** fine-tuned from
|
| 19 |
**Qwen/Qwen3-4B-Instruct-2507** using **QLoRA (4-bit, Unsloth)**.
|
|
@@ -34,9 +34,9 @@ while intermediate reasoning (Chain-of-Thought) is masked.
|
|
| 34 |
- Base model: Qwen/Qwen3-4B-Instruct-2507
|
| 35 |
- Method: QLoRA (4-bit)
|
| 36 |
- Max sequence length: 512
|
| 37 |
-
- Epochs:
|
| 38 |
-
- Learning rate:
|
| 39 |
-
- LoRA: r=
|
| 40 |
|
| 41 |
## Usage
|
| 42 |
|
|
|
|
| 13 |
- structured-output
|
| 14 |
---
|
| 15 |
|
| 16 |
+
# Qwen3-4B RSLoRA (R=256) for Structured Output
|
| 17 |
|
| 18 |
This repository provides a **LoRA adapter** fine-tuned from
|
| 19 |
**Qwen/Qwen3-4B-Instruct-2507** using **QLoRA (4-bit, Unsloth)**.
|
|
|
|
| 34 |
- Base model: Qwen/Qwen3-4B-Instruct-2507
|
| 35 |
- Method: QLoRA (4-bit)
|
| 36 |
- Max sequence length: 512
|
| 37 |
+
- Epochs: 2
|
| 38 |
+
- Learning rate: 2e-04
|
| 39 |
+
- LoRA: r=256, alpha=32
|
| 40 |
|
| 41 |
## Usage
|
| 42 |
|
adapter_config.json
CHANGED
|
@@ -20,7 +20,7 @@
|
|
| 20 |
"layers_pattern": null,
|
| 21 |
"layers_to_transform": null,
|
| 22 |
"loftq_config": {},
|
| 23 |
-
"lora_alpha":
|
| 24 |
"lora_bias": false,
|
| 25 |
"lora_dropout": 0.0,
|
| 26 |
"megatron_config": null,
|
|
@@ -29,22 +29,22 @@
|
|
| 29 |
"peft_type": "LORA",
|
| 30 |
"peft_version": "0.18.1",
|
| 31 |
"qalora_group_size": 16,
|
| 32 |
-
"r":
|
| 33 |
"rank_pattern": {},
|
| 34 |
"revision": null,
|
| 35 |
"target_modules": [
|
| 36 |
-
"k_proj",
|
| 37 |
-
"v_proj",
|
| 38 |
-
"q_proj",
|
| 39 |
"gate_proj",
|
| 40 |
"down_proj",
|
|
|
|
| 41 |
"up_proj",
|
| 42 |
-
"o_proj"
|
|
|
|
|
|
|
| 43 |
],
|
| 44 |
"target_parameters": null,
|
| 45 |
"task_type": "CAUSAL_LM",
|
| 46 |
"trainable_token_indices": null,
|
| 47 |
"use_dora": false,
|
| 48 |
"use_qalora": false,
|
| 49 |
-
"use_rslora":
|
| 50 |
}
|
|
|
|
| 20 |
"layers_pattern": null,
|
| 21 |
"layers_to_transform": null,
|
| 22 |
"loftq_config": {},
|
| 23 |
+
"lora_alpha": 32,
|
| 24 |
"lora_bias": false,
|
| 25 |
"lora_dropout": 0.0,
|
| 26 |
"megatron_config": null,
|
|
|
|
| 29 |
"peft_type": "LORA",
|
| 30 |
"peft_version": "0.18.1",
|
| 31 |
"qalora_group_size": 16,
|
| 32 |
+
"r": 256,
|
| 33 |
"rank_pattern": {},
|
| 34 |
"revision": null,
|
| 35 |
"target_modules": [
|
|
|
|
|
|
|
|
|
|
| 36 |
"gate_proj",
|
| 37 |
"down_proj",
|
| 38 |
+
"k_proj",
|
| 39 |
"up_proj",
|
| 40 |
+
"o_proj",
|
| 41 |
+
"v_proj",
|
| 42 |
+
"q_proj"
|
| 43 |
],
|
| 44 |
"target_parameters": null,
|
| 45 |
"task_type": "CAUSAL_LM",
|
| 46 |
"trainable_token_indices": null,
|
| 47 |
"use_dora": false,
|
| 48 |
"use_qalora": false,
|
| 49 |
+
"use_rslora": true
|
| 50 |
}
|
adapter_model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9f7e32a8d79b5e697d84e7807625629d764aff33c2a8a303e7ff31d5cb7fe96d
|
| 3 |
+
size 2113998360
|
all_experiments_details.json
ADDED
|
@@ -0,0 +1,434 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"Experiment_1": {
|
| 3 |
+
"config": {
|
| 4 |
+
"BASE_MODEL_ID": "Qwen/Qwen3-4B-Instruct-2507",
|
| 5 |
+
"DATASET_ID": "u-10bei/structured_data_with_cot_dataset_512_v2",
|
| 6 |
+
"BASE_OUT_DIR": "./lora_experiments",
|
| 7 |
+
"SEED": 3407,
|
| 8 |
+
"VAL_RATIO": 0.05,
|
| 9 |
+
"MAX_SEQ_LEN": 512,
|
| 10 |
+
"LORA_R": 256,
|
| 11 |
+
"LORA_ALPHA": 32,
|
| 12 |
+
"LORA_DROPOUT": 0.0,
|
| 13 |
+
"LORA_TARGET_MODULES": [
|
| 14 |
+
"q_proj",
|
| 15 |
+
"k_proj",
|
| 16 |
+
"v_proj",
|
| 17 |
+
"o_proj",
|
| 18 |
+
"gate_proj",
|
| 19 |
+
"up_proj",
|
| 20 |
+
"down_proj"
|
| 21 |
+
],
|
| 22 |
+
"EPOCHS": 2,
|
| 23 |
+
"PER_DEVICE_TRAIN_BS": 2,
|
| 24 |
+
"PER_DEVICE_EVAL_BS": 2,
|
| 25 |
+
"GRAD_ACCUM": 8,
|
| 26 |
+
"LR": 0.0002,
|
| 27 |
+
"WARMUP_RATIO": 0.1,
|
| 28 |
+
"WEIGHT_DECAY": 0.05,
|
| 29 |
+
"MAX_STEPS": -1,
|
| 30 |
+
"LOGGING_STEPS": 10,
|
| 31 |
+
"EVAL_STEPS": 50,
|
| 32 |
+
"SAVE_STEPS": 100,
|
| 33 |
+
"SAVE_TOTAL_LIMIT": 2,
|
| 34 |
+
"MASK_COT": true,
|
| 35 |
+
"OUTPUT_MARKERS": [
|
| 36 |
+
"Output:",
|
| 37 |
+
"OUTPUT:",
|
| 38 |
+
"Final:",
|
| 39 |
+
"Answer:",
|
| 40 |
+
"Result:",
|
| 41 |
+
"Response:"
|
| 42 |
+
],
|
| 43 |
+
"OUTPUT_LEARN_MODE": "after_marker",
|
| 44 |
+
"UPSAMPLE_ENABLE": false,
|
| 45 |
+
"UPSAMPLE_RULES_JSON": "{\"xml_to_yaml\": 2.0}",
|
| 46 |
+
"Experiment_Name": "Method_RSLoRA_R256",
|
| 47 |
+
"USE_RSLORA": true
|
| 48 |
+
},
|
| 49 |
+
"history": [
|
| 50 |
+
{
|
| 51 |
+
"loss": 1.3966,
|
| 52 |
+
"grad_norm": 1.0872466564178467,
|
| 53 |
+
"learning_rate": 4e-05,
|
| 54 |
+
"epoch": 0.04501969611705121,
|
| 55 |
+
"step": 10
|
| 56 |
+
},
|
| 57 |
+
{
|
| 58 |
+
"loss": 0.9325,
|
| 59 |
+
"grad_norm": 0.9557391405105591,
|
| 60 |
+
"learning_rate": 8.444444444444444e-05,
|
| 61 |
+
"epoch": 0.09003939223410241,
|
| 62 |
+
"step": 20
|
| 63 |
+
},
|
| 64 |
+
{
|
| 65 |
+
"loss": 0.8011,
|
| 66 |
+
"grad_norm": 0.513979971408844,
|
| 67 |
+
"learning_rate": 0.00012888888888888892,
|
| 68 |
+
"epoch": 0.13505908835115363,
|
| 69 |
+
"step": 30
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"loss": 0.8125,
|
| 73 |
+
"grad_norm": 0.5614296197891235,
|
| 74 |
+
"learning_rate": 0.00017333333333333334,
|
| 75 |
+
"epoch": 0.18007878446820483,
|
| 76 |
+
"step": 40
|
| 77 |
+
},
|
| 78 |
+
{
|
| 79 |
+
"loss": 0.7486,
|
| 80 |
+
"grad_norm": 0.32493776082992554,
|
| 81 |
+
"learning_rate": 0.0001999509018141497,
|
| 82 |
+
"epoch": 0.22509848058525606,
|
| 83 |
+
"step": 50
|
| 84 |
+
},
|
| 85 |
+
{
|
| 86 |
+
"eval_loss": 0.8527934551239014,
|
| 87 |
+
"eval_runtime": 25.8751,
|
| 88 |
+
"eval_samples_per_second": 7.227,
|
| 89 |
+
"eval_steps_per_second": 3.633,
|
| 90 |
+
"epoch": 0.22509848058525606,
|
| 91 |
+
"step": 50
|
| 92 |
+
},
|
| 93 |
+
{
|
| 94 |
+
"loss": 0.7415,
|
| 95 |
+
"grad_norm": 0.2533496022224426,
|
| 96 |
+
"learning_rate": 0.00019939910076582706,
|
| 97 |
+
"epoch": 0.27011817670230726,
|
| 98 |
+
"step": 60
|
| 99 |
+
},
|
| 100 |
+
{
|
| 101 |
+
"loss": 0.8629,
|
| 102 |
+
"grad_norm": 0.3107249438762665,
|
| 103 |
+
"learning_rate": 0.00019823752233636866,
|
| 104 |
+
"epoch": 0.31513787281935846,
|
| 105 |
+
"step": 70
|
| 106 |
+
},
|
| 107 |
+
{
|
| 108 |
+
"loss": 0.7151,
|
| 109 |
+
"grad_norm": 0.32290422916412354,
|
| 110 |
+
"learning_rate": 0.00019647329238755036,
|
| 111 |
+
"epoch": 0.36015756893640966,
|
| 112 |
+
"step": 80
|
| 113 |
+
},
|
| 114 |
+
{
|
| 115 |
+
"loss": 0.7843,
|
| 116 |
+
"grad_norm": 0.3160030245780945,
|
| 117 |
+
"learning_rate": 0.0001941172338293343,
|
| 118 |
+
"epoch": 0.4051772650534609,
|
| 119 |
+
"step": 90
|
| 120 |
+
},
|
| 121 |
+
{
|
| 122 |
+
"loss": 0.7931,
|
| 123 |
+
"grad_norm": 0.40372225642204285,
|
| 124 |
+
"learning_rate": 0.00019118380022524738,
|
| 125 |
+
"epoch": 0.4501969611705121,
|
| 126 |
+
"step": 100
|
| 127 |
+
},
|
| 128 |
+
{
|
| 129 |
+
"eval_loss": 0.7929844260215759,
|
| 130 |
+
"eval_runtime": 25.6248,
|
| 131 |
+
"eval_samples_per_second": 7.298,
|
| 132 |
+
"eval_steps_per_second": 3.668,
|
| 133 |
+
"epoch": 0.4501969611705121,
|
| 134 |
+
"step": 100
|
| 135 |
+
},
|
| 136 |
+
{
|
| 137 |
+
"loss": 0.812,
|
| 138 |
+
"grad_norm": 0.4371040165424347,
|
| 139 |
+
"learning_rate": 0.0001876909871250184,
|
| 140 |
+
"epoch": 0.4952166572875633,
|
| 141 |
+
"step": 110
|
| 142 |
+
},
|
| 143 |
+
{
|
| 144 |
+
"loss": 0.7258,
|
| 145 |
+
"grad_norm": 0.28696581721305847,
|
| 146 |
+
"learning_rate": 0.00018366022166841676,
|
| 147 |
+
"epoch": 0.5402363534046145,
|
| 148 |
+
"step": 120
|
| 149 |
+
},
|
| 150 |
+
{
|
| 151 |
+
"loss": 0.7824,
|
| 152 |
+
"grad_norm": 0.9101247787475586,
|
| 153 |
+
"learning_rate": 0.0001791162311375321,
|
| 154 |
+
"epoch": 0.5852560495216658,
|
| 155 |
+
"step": 130
|
| 156 |
+
},
|
| 157 |
+
{
|
| 158 |
+
"loss": 0.7643,
|
| 159 |
+
"grad_norm": 0.2658868730068207,
|
| 160 |
+
"learning_rate": 0.00017408689126387995,
|
| 161 |
+
"epoch": 0.6302757456387169,
|
| 162 |
+
"step": 140
|
| 163 |
+
},
|
| 164 |
+
{
|
| 165 |
+
"loss": 0.711,
|
| 166 |
+
"grad_norm": 0.43262964487075806,
|
| 167 |
+
"learning_rate": 0.0001686030552209133,
|
| 168 |
+
"epoch": 0.6752954417557682,
|
| 169 |
+
"step": 150
|
| 170 |
+
},
|
| 171 |
+
{
|
| 172 |
+
"eval_loss": 0.7660654783248901,
|
| 173 |
+
"eval_runtime": 24.682,
|
| 174 |
+
"eval_samples_per_second": 7.576,
|
| 175 |
+
"eval_steps_per_second": 3.808,
|
| 176 |
+
"epoch": 0.6752954417557682,
|
| 177 |
+
"step": 150
|
| 178 |
+
},
|
| 179 |
+
{
|
| 180 |
+
"loss": 0.6915,
|
| 181 |
+
"grad_norm": 0.36354902386665344,
|
| 182 |
+
"learning_rate": 0.00016269836435100934,
|
| 183 |
+
"epoch": 0.7203151378728193,
|
| 184 |
+
"step": 160
|
| 185 |
+
},
|
| 186 |
+
{
|
| 187 |
+
"loss": 0.7233,
|
| 188 |
+
"grad_norm": 0.3520168662071228,
|
| 189 |
+
"learning_rate": 0.0001564090417880529,
|
| 190 |
+
"epoch": 0.7653348339898706,
|
| 191 |
+
"step": 170
|
| 192 |
+
},
|
| 193 |
+
{
|
| 194 |
+
"loss": 0.6626,
|
| 195 |
+
"grad_norm": 0.33897820115089417,
|
| 196 |
+
"learning_rate": 0.0001497736702416662,
|
| 197 |
+
"epoch": 0.8103545301069218,
|
| 198 |
+
"step": 180
|
| 199 |
+
},
|
| 200 |
+
{
|
| 201 |
+
"loss": 0.6801,
|
| 202 |
+
"grad_norm": 0.4690793752670288,
|
| 203 |
+
"learning_rate": 0.00014283295530629877,
|
| 204 |
+
"epoch": 0.855374226223973,
|
| 205 |
+
"step": 190
|
| 206 |
+
},
|
| 207 |
+
{
|
| 208 |
+
"loss": 0.6281,
|
| 209 |
+
"grad_norm": 0.42146065831184387,
|
| 210 |
+
"learning_rate": 0.00013562947574718976,
|
| 211 |
+
"epoch": 0.9003939223410242,
|
| 212 |
+
"step": 200
|
| 213 |
+
},
|
| 214 |
+
{
|
| 215 |
+
"eval_loss": 0.6899478435516357,
|
| 216 |
+
"eval_runtime": 24.6215,
|
| 217 |
+
"eval_samples_per_second": 7.595,
|
| 218 |
+
"eval_steps_per_second": 3.818,
|
| 219 |
+
"epoch": 0.9003939223410242,
|
| 220 |
+
"step": 200
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"loss": 0.6237,
|
| 224 |
+
"grad_norm": 0.38645192980766296,
|
| 225 |
+
"learning_rate": 0.00012820742229510817,
|
| 226 |
+
"epoch": 0.9454136184580754,
|
| 227 |
+
"step": 210
|
| 228 |
+
},
|
| 229 |
+
{
|
| 230 |
+
"loss": 0.5856,
|
| 231 |
+
"grad_norm": 0.4482150375843048,
|
| 232 |
+
"learning_rate": 0.00012061232655226964,
|
| 233 |
+
"epoch": 0.9904333145751266,
|
| 234 |
+
"step": 220
|
| 235 |
+
},
|
| 236 |
+
{
|
| 237 |
+
"loss": 0.4553,
|
| 238 |
+
"grad_norm": 0.5086686015129089,
|
| 239 |
+
"learning_rate": 0.00011289078167249402,
|
| 240 |
+
"epoch": 1.0315137872819358,
|
| 241 |
+
"step": 230
|
| 242 |
+
},
|
| 243 |
+
{
|
| 244 |
+
"loss": 0.5029,
|
| 245 |
+
"grad_norm": 0.8159873485565186,
|
| 246 |
+
"learning_rate": 0.00010509015652912966,
|
| 247 |
+
"epoch": 1.076533483398987,
|
| 248 |
+
"step": 240
|
| 249 |
+
},
|
| 250 |
+
{
|
| 251 |
+
"loss": 0.5179,
|
| 252 |
+
"grad_norm": 0.5635101199150085,
|
| 253 |
+
"learning_rate": 9.72583051242198e-05,
|
| 254 |
+
"epoch": 1.1215531795160383,
|
| 255 |
+
"step": 250
|
| 256 |
+
},
|
| 257 |
+
{
|
| 258 |
+
"eval_loss": 0.5836588144302368,
|
| 259 |
+
"eval_runtime": 24.7575,
|
| 260 |
+
"eval_samples_per_second": 7.553,
|
| 261 |
+
"eval_steps_per_second": 3.797,
|
| 262 |
+
"epoch": 1.1215531795160383,
|
| 263 |
+
"step": 250
|
| 264 |
+
},
|
| 265 |
+
{
|
| 266 |
+
"loss": 0.514,
|
| 267 |
+
"grad_norm": 0.661852240562439,
|
| 268 |
+
"learning_rate": 8.944327302158073e-05,
|
| 269 |
+
"epoch": 1.1665728756330895,
|
| 270 |
+
"step": 260
|
| 271 |
+
},
|
| 272 |
+
{
|
| 273 |
+
"loss": 0.4472,
|
| 274 |
+
"grad_norm": 0.4757942259311676,
|
| 275 |
+
"learning_rate": 8.169300260471818e-05,
|
| 276 |
+
"epoch": 1.2115925717501406,
|
| 277 |
+
"step": 270
|
| 278 |
+
},
|
| 279 |
+
{
|
| 280 |
+
"loss": 0.4286,
|
| 281 |
+
"grad_norm": 0.6126232743263245,
|
| 282 |
+
"learning_rate": 7.405503896771729e-05,
|
| 283 |
+
"epoch": 1.2566122678671918,
|
| 284 |
+
"step": 280
|
| 285 |
+
},
|
| 286 |
+
{
|
| 287 |
+
"loss": 0.4862,
|
| 288 |
+
"grad_norm": 0.5707330703735352,
|
| 289 |
+
"learning_rate": 6.65762382433589e-05,
|
| 290 |
+
"epoch": 1.301631963984243,
|
| 291 |
+
"step": 290
|
| 292 |
+
},
|
| 293 |
+
{
|
| 294 |
+
"loss": 0.4352,
|
| 295 |
+
"grad_norm": 0.3476680815219879,
|
| 296 |
+
"learning_rate": 5.930248015776325e-05,
|
| 297 |
+
"epoch": 1.3466516601012943,
|
| 298 |
+
"step": 300
|
| 299 |
+
},
|
| 300 |
+
{
|
| 301 |
+
"eval_loss": 0.4677433371543884,
|
| 302 |
+
"eval_runtime": 24.7074,
|
| 303 |
+
"eval_samples_per_second": 7.569,
|
| 304 |
+
"eval_steps_per_second": 3.805,
|
| 305 |
+
"epoch": 1.3466516601012943,
|
| 306 |
+
"step": 300
|
| 307 |
+
},
|
| 308 |
+
{
|
| 309 |
+
"loss": 0.3118,
|
| 310 |
+
"grad_norm": 0.6370311379432678,
|
| 311 |
+
"learning_rate": 5.227838657493396e-05,
|
| 312 |
+
"epoch": 1.3916713562183456,
|
| 313 |
+
"step": 310
|
| 314 |
+
},
|
| 315 |
+
{
|
| 316 |
+
"loss": 0.3847,
|
| 317 |
+
"grad_norm": 0.5159108638763428,
|
| 318 |
+
"learning_rate": 4.5547047757828985e-05,
|
| 319 |
+
"epoch": 1.4366910523353966,
|
| 320 |
+
"step": 320
|
| 321 |
+
},
|
| 322 |
+
{
|
| 323 |
+
"loss": 0.3931,
|
| 324 |
+
"grad_norm": 0.6911277174949646,
|
| 325 |
+
"learning_rate": 3.914975802524806e-05,
|
| 326 |
+
"epoch": 1.4817107484524479,
|
| 327 |
+
"step": 330
|
| 328 |
+
},
|
| 329 |
+
{
|
| 330 |
+
"loss": 0.5302,
|
| 331 |
+
"grad_norm": 0.5555063486099243,
|
| 332 |
+
"learning_rate": 3.312576242618511e-05,
|
| 333 |
+
"epoch": 1.5267304445694991,
|
| 334 |
+
"step": 340
|
| 335 |
+
},
|
| 336 |
+
{
|
| 337 |
+
"loss": 0.3183,
|
| 338 |
+
"grad_norm": 0.8983607292175293,
|
| 339 |
+
"learning_rate": 2.7512015985706418e-05,
|
| 340 |
+
"epoch": 1.5717501406865504,
|
| 341 |
+
"step": 350
|
| 342 |
+
},
|
| 343 |
+
{
|
| 344 |
+
"eval_loss": 0.3523830473423004,
|
| 345 |
+
"eval_runtime": 24.7895,
|
| 346 |
+
"eval_samples_per_second": 7.544,
|
| 347 |
+
"eval_steps_per_second": 3.792,
|
| 348 |
+
"epoch": 1.5717501406865504,
|
| 349 |
+
"step": 350
|
| 350 |
+
},
|
| 351 |
+
{
|
| 352 |
+
"loss": 0.3694,
|
| 353 |
+
"grad_norm": 0.5024566054344177,
|
| 354 |
+
"learning_rate": 2.234295699929413e-05,
|
| 355 |
+
"epoch": 1.6167698368036016,
|
| 356 |
+
"step": 360
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"loss": 0.3502,
|
| 360 |
+
"grad_norm": 0.5496794581413269,
|
| 361 |
+
"learning_rate": 1.7650295766411605e-05,
|
| 362 |
+
"epoch": 1.6617895329206527,
|
| 363 |
+
"step": 370
|
| 364 |
+
},
|
| 365 |
+
{
|
| 366 |
+
"loss": 0.3006,
|
| 367 |
+
"grad_norm": 0.4716707766056061,
|
| 368 |
+
"learning_rate": 1.3462820059333403e-05,
|
| 369 |
+
"epoch": 1.7068092290377042,
|
| 370 |
+
"step": 380
|
| 371 |
+
},
|
| 372 |
+
{
|
| 373 |
+
"loss": 0.3471,
|
| 374 |
+
"grad_norm": 0.46408089995384216,
|
| 375 |
+
"learning_rate": 9.80621852061826e-06,
|
| 376 |
+
"epoch": 1.7518289251547552,
|
| 377 |
+
"step": 390
|
| 378 |
+
},
|
| 379 |
+
{
|
| 380 |
+
"loss": 0.2301,
|
| 381 |
+
"grad_norm": 0.41809141635894775,
|
| 382 |
+
"learning_rate": 6.702923072617129e-06,
|
| 383 |
+
"epoch": 1.7968486212718064,
|
| 384 |
+
"step": 400
|
| 385 |
+
},
|
| 386 |
+
{
|
| 387 |
+
"eval_loss": 0.3102871775627136,
|
| 388 |
+
"eval_runtime": 24.3436,
|
| 389 |
+
"eval_samples_per_second": 7.682,
|
| 390 |
+
"eval_steps_per_second": 3.861,
|
| 391 |
+
"epoch": 1.7968486212718064,
|
| 392 |
+
"step": 400
|
| 393 |
+
},
|
| 394 |
+
{
|
| 395 |
+
"loss": 0.2704,
|
| 396 |
+
"grad_norm": 0.4462619423866272,
|
| 397 |
+
"learning_rate": 4.171971305776945e-06,
|
| 398 |
+
"epoch": 1.8418683173888577,
|
| 399 |
+
"step": 410
|
| 400 |
+
},
|
| 401 |
+
{
|
| 402 |
+
"loss": 0.2168,
|
| 403 |
+
"grad_norm": 0.4414360523223877,
|
| 404 |
+
"learning_rate": 2.2288896899377186e-06,
|
| 405 |
+
"epoch": 1.8868880135059087,
|
| 406 |
+
"step": 420
|
| 407 |
+
},
|
| 408 |
+
{
|
| 409 |
+
"loss": 0.2626,
|
| 410 |
+
"grad_norm": 0.3277634084224701,
|
| 411 |
+
"learning_rate": 8.855983250793288e-07,
|
| 412 |
+
"epoch": 1.93190770962296,
|
| 413 |
+
"step": 430
|
| 414 |
+
},
|
| 415 |
+
{
|
| 416 |
+
"loss": 0.342,
|
| 417 |
+
"grad_norm": 0.4310093820095062,
|
| 418 |
+
"learning_rate": 1.5033781583758678e-07,
|
| 419 |
+
"epoch": 1.9769274057400112,
|
| 420 |
+
"step": 440
|
| 421 |
+
},
|
| 422 |
+
{
|
| 423 |
+
"train_runtime": 3776.627,
|
| 424 |
+
"train_samples_per_second": 1.882,
|
| 425 |
+
"train_steps_per_second": 0.118,
|
| 426 |
+
"total_flos": 7.57271106173184e+16,
|
| 427 |
+
"train_loss": 0.5718902958882763,
|
| 428 |
+
"epoch": 2.0,
|
| 429 |
+
"step": 446,
|
| 430 |
+
"total_runtime_sec": 3777.7196531295776
|
| 431 |
+
}
|
| 432 |
+
]
|
| 433 |
+
}
|
| 434 |
+
}
|