diff --git a/app.py b/app.py new file mode 100644 index 0000000000000000000000000000000000000000..f1aa638fb888d1db8de9ac11314d6a8a47d520b8 --- /dev/null +++ b/app.py @@ -0,0 +1,38 @@ +import gradio as gr +from torch.nn.functional import softmax +import torch +from transformers import ViTFeatureExtractor +from transformers import MobileViTFeatureExtractor +from transformers import MobileViTForImageClassification +from transformers import ViTForImageClassification + + +def predict(model_type, inp): + + if model_type == "ViT": + model_name_or_path = './models/vit-base-garbage/' + feature_extractor = ViTFeatureExtractor.from_pretrained(model_name_or_path) + model = ViTForImageClassification.from_pretrained(model_name_or_path) + elif model_type == "MobileViT": + model_name_or_path = './models/apple/mobilevit-small-garbage/' + feature_extractor = MobileViTFeatureExtractor.from_pretrained(model_name_or_path) + model = MobileViTForImageClassification.from_pretrained(model_name_or_path) + inputs = feature_extractor(inp, return_tensors="pt") + LABELS = list(model.config.label2id.keys()) + with torch.no_grad(): + logits = model(**inputs) + print(logits[0]) + probability = torch.nn.functional.softmax(logits[0], dim=-1) + + confidences = {LABELS[i]:(float(probability[0][i])) for i in range(6)} + # print(confidences) + return confidences + + +demo = gr.Interface(fn=predict, + inputs=[gr.Dropdown(["ViT", "MobileViT"], label="Model Name", value='ViT'),gr.inputs.Image(type="pil")], + outputs=gr.outputs.Label(num_top_classes=3), + examples=[["ViT","paper567.jpg"],["ViT","trash105.jpg"],["ViT","plastic202.jpg"],["MobileViT","metal382.jpg"]], + ) + +demo.launch() \ No newline at end of file diff --git a/cardboard373.jpg b/cardboard373.jpg new file mode 100644 index 0000000000000000000000000000000000000000..972747533332bf05b11bb1583beaba19eae0bea1 Binary files /dev/null and b/cardboard373.jpg differ diff --git a/glass456.jpg b/glass456.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b0567693203315f4f188dfb25fc47d8f7bbb8e6e Binary files /dev/null and b/glass456.jpg differ diff --git a/metal382.jpg b/metal382.jpg new file mode 100644 index 0000000000000000000000000000000000000000..1c3190e3f789188598b7c4df3d807821259389c2 Binary files /dev/null and b/metal382.jpg differ diff --git a/models/.DS_Store b/models/.DS_Store new file mode 100644 index 0000000000000000000000000000000000000000..fd4647df87fab4aae1967d8fe89d6cfab4bf5c26 Binary files /dev/null and b/models/.DS_Store differ diff --git a/models/apple/.DS_Store b/models/apple/.DS_Store new file mode 100644 index 0000000000000000000000000000000000000000..fd1add4bcbd8496d8dfe266cfccf9dbc37375c95 Binary files /dev/null and b/models/apple/.DS_Store differ diff --git a/models/apple/mobilevit-small-garbage/all_results.json b/models/apple/mobilevit-small-garbage/all_results.json new file mode 100644 index 0000000000000000000000000000000000000000..8bf5ba49cb4cb321e7bb5557cf1879e88bf3ae9e --- /dev/null +++ b/models/apple/mobilevit-small-garbage/all_results.json @@ -0,0 +1,16 @@ +{ + "epoch": 5.0, + "eval_accuracy": 0.9396751740139211, + "eval_f1": 0.9320563646123423, + "eval_loss": 0.17442849278450012, + "eval_precision": 0.9422527093650638, + "eval_recall": 0.9244185656020448, + "eval_runtime": 7.4671, + "eval_samples_per_second": 57.72, + "eval_steps_per_second": 7.232, + "total_flos": 5.153016901533696e+16, + "train_loss": 0.5183452816696854, + "train_runtime": 159.3173, + "train_samples_per_second": 55.487, + "train_steps_per_second": 3.484 +} \ No newline at end of file diff --git a/models/apple/mobilevit-small-garbage/config.json b/models/apple/mobilevit-small-garbage/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7c761149d1cf6a62904148c5af147acb12dce65d --- /dev/null +++ b/models/apple/mobilevit-small-garbage/config.json @@ -0,0 +1,63 @@ +{ + "_name_or_path": "apple/mobilevit-small", + "architectures": [ + "MobileViTForImageClassification" + ], + "aspp_dropout_prob": 0.1, + "aspp_out_channels": 256, + "atrous_rates": [ + 6, + 12, + 18 + ], + "attention_probs_dropout_prob": 0.0, + "classifier_dropout_prob": 0.1, + "conv_kernel_size": 3, + "expand_ratio": 4.0, + "hidden_act": "silu", + "hidden_dropout_prob": 0.1, + "hidden_sizes": [ + 144, + 192, + 240 + ], + "id2label": { + "0": "cardboard", + "1": "glass", + "2": "metal", + "3": "paper", + "4": "plastic", + "5": "trash" + }, + "image_size": 256, + "initializer_range": 0.02, + "label2id": { + "cardboard": "0", + "glass": "1", + "metal": "2", + "paper": "3", + "plastic": "4", + "trash": "5" + }, + "layer_norm_eps": 1e-05, + "mlp_ratio": 2.0, + "model_type": "mobilevit", + "neck_hidden_sizes": [ + 16, + 32, + 64, + 96, + 128, + 160, + 640 + ], + "num_attention_heads": 4, + "num_channels": 3, + "output_stride": 32, + "patch_size": 2, + "problem_type": "single_label_classification", + "qkv_bias": true, + "semantic_loss_ignore_index": 255, + "torch_dtype": "float32", + "transformers_version": "4.24.0" +} diff --git a/models/apple/mobilevit-small-garbage/preprocessor_config.json b/models/apple/mobilevit-small-garbage/preprocessor_config.json new file mode 100644 index 0000000000000000000000000000000000000000..44670ed45552d237cc7c55d06ee19d15e734a766 --- /dev/null +++ b/models/apple/mobilevit-small-garbage/preprocessor_config.json @@ -0,0 +1,10 @@ +{ + "crop_size": 256, + "do_center_crop": true, + "do_flip_channel_order": true, + "do_flip_channels": true, + "do_resize": true, + "feature_extractor_type": "MobileViTFeatureExtractor", + "resample": 2, + "size": 288 +} diff --git a/models/apple/mobilevit-small-garbage/pytorch_model.bin b/models/apple/mobilevit-small-garbage/pytorch_model.bin new file mode 100644 index 0000000000000000000000000000000000000000..053a310f25cd7d41bd1b7f6c81ccde68184839ad --- /dev/null +++ b/models/apple/mobilevit-small-garbage/pytorch_model.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3de3d2c2a6114c79a2a1814b21ee1b602f130851a3f70611e55b52d652d5dd31 +size 19937419 diff --git a/models/apple/mobilevit-small-garbage/runs/Dec01_01-56-59_48dca185d887/1669859844.908253/events.out.tfevents.1669859844.48dca185d887.5019.7 b/models/apple/mobilevit-small-garbage/runs/Dec01_01-56-59_48dca185d887/1669859844.908253/events.out.tfevents.1669859844.48dca185d887.5019.7 new file mode 100644 index 0000000000000000000000000000000000000000..89edb832e9ea3f2fc3edafc36ef0300c949346e1 --- /dev/null +++ b/models/apple/mobilevit-small-garbage/runs/Dec01_01-56-59_48dca185d887/1669859844.908253/events.out.tfevents.1669859844.48dca185d887.5019.7 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9029ef393ea8d118116c66afed5937f1749a702460878cb2cb78ad3c8232e568 +size 5520 diff --git a/models/apple/mobilevit-small-garbage/runs/Dec01_01-56-59_48dca185d887/events.out.tfevents.1669859844.48dca185d887.5019.6 b/models/apple/mobilevit-small-garbage/runs/Dec01_01-56-59_48dca185d887/events.out.tfevents.1669859844.48dca185d887.5019.6 new file mode 100644 index 0000000000000000000000000000000000000000..b7fa606cc2be1c5438d04df0a9504d0b9518ba5b --- /dev/null +++ b/models/apple/mobilevit-small-garbage/runs/Dec01_01-56-59_48dca185d887/events.out.tfevents.1669859844.48dca185d887.5019.6 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d1ef8ff7f11f7bdbec4874fbf7309af69251cfad89d4db2f9b936ef27ab03e82 +size 5775 diff --git a/models/apple/mobilevit-small-garbage/runs/Dec01_01-58-12_48dca185d887/1669859892.1544209/events.out.tfevents.1669859892.48dca185d887.5019.9 b/models/apple/mobilevit-small-garbage/runs/Dec01_01-58-12_48dca185d887/1669859892.1544209/events.out.tfevents.1669859892.48dca185d887.5019.9 new file mode 100644 index 0000000000000000000000000000000000000000..c4aa5ac9d816e7ff0bb4781aea127a193416bffa --- /dev/null +++ b/models/apple/mobilevit-small-garbage/runs/Dec01_01-58-12_48dca185d887/1669859892.1544209/events.out.tfevents.1669859892.48dca185d887.5019.9 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fb6cd0a21c8eedb632c85fcff05dacc7bf3428542263c618958d97749d88e625 +size 5520 diff --git a/models/apple/mobilevit-small-garbage/runs/Dec01_01-58-12_48dca185d887/events.out.tfevents.1669859892.48dca185d887.5019.8 b/models/apple/mobilevit-small-garbage/runs/Dec01_01-58-12_48dca185d887/events.out.tfevents.1669859892.48dca185d887.5019.8 new file mode 100644 index 0000000000000000000000000000000000000000..bd4c6bdfe2dddf4b0b9c49c972d4129d13586caf --- /dev/null +++ b/models/apple/mobilevit-small-garbage/runs/Dec01_01-58-12_48dca185d887/events.out.tfevents.1669859892.48dca185d887.5019.8 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3226a08a0d3739a8da4b6ef5e37f81701bf60a5f4c4e7bbe2678e7b4bc887072 +size 10697 diff --git a/models/apple/mobilevit-small-garbage/runs/Dec01_02-03-03_48dca185d887/1669860183.7440953/events.out.tfevents.1669860183.48dca185d887.5019.11 b/models/apple/mobilevit-small-garbage/runs/Dec01_02-03-03_48dca185d887/1669860183.7440953/events.out.tfevents.1669860183.48dca185d887.5019.11 new file mode 100644 index 0000000000000000000000000000000000000000..d208c7da41b695a2e1f92e3eb82a06ecb42110e0 --- /dev/null +++ b/models/apple/mobilevit-small-garbage/runs/Dec01_02-03-03_48dca185d887/1669860183.7440953/events.out.tfevents.1669860183.48dca185d887.5019.11 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0b6dfd2faa33c11bbb5f9638c7ed47fcce4499450a2912573286a751c1fdd878 +size 5520 diff --git a/models/apple/mobilevit-small-garbage/runs/Dec01_02-03-03_48dca185d887/events.out.tfevents.1669860183.48dca185d887.5019.10 b/models/apple/mobilevit-small-garbage/runs/Dec01_02-03-03_48dca185d887/events.out.tfevents.1669860183.48dca185d887.5019.10 new file mode 100644 index 0000000000000000000000000000000000000000..1c6c0008c6ea264381649b9cc03eb8ba40e42850 --- /dev/null +++ b/models/apple/mobilevit-small-garbage/runs/Dec01_02-03-03_48dca185d887/events.out.tfevents.1669860183.48dca185d887.5019.10 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a69e9734c9a1d66df23d6d2326cf01248e11e73a5f757604b12fd305e896e3bf +size 14797 diff --git a/models/apple/mobilevit-small-garbage/runs/Dec12_21-56-05_30c96569bd60/events.out.tfevents.1670882184.30c96569bd60.71.2 b/models/apple/mobilevit-small-garbage/runs/Dec12_21-56-05_30c96569bd60/events.out.tfevents.1670882184.30c96569bd60.71.2 new file mode 100644 index 0000000000000000000000000000000000000000..de2bf89f3279930415e53a0191d8efcc498d07d6 --- /dev/null +++ b/models/apple/mobilevit-small-garbage/runs/Dec12_21-56-05_30c96569bd60/events.out.tfevents.1670882184.30c96569bd60.71.2 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dc7b9c1ba45264fb14f86d79d0ea4d1ef43bbf7b7de5386e19de02b293293250 +size 438 diff --git a/models/apple/mobilevit-small-garbage/test_results.json b/models/apple/mobilevit-small-garbage/test_results.json new file mode 100644 index 0000000000000000000000000000000000000000..50c56aa7e7639a55e93bb27fdac1fc0cecce6927 --- /dev/null +++ b/models/apple/mobilevit-small-garbage/test_results.json @@ -0,0 +1,10 @@ +{ + "eval_accuracy": 0.9396751740139211, + "eval_f1": 0.9320563646123423, + "eval_loss": 0.17442849278450012, + "eval_precision": 0.9422527093650638, + "eval_recall": 0.9244185656020448, + "eval_runtime": 7.4671, + "eval_samples_per_second": 57.72, + "eval_steps_per_second": 7.232 +} \ No newline at end of file diff --git a/models/apple/mobilevit-small-garbage/train_results.json b/models/apple/mobilevit-small-garbage/train_results.json new file mode 100644 index 0000000000000000000000000000000000000000..bd06913a47c67ea8b87f63f3a6c2af81fb521eba --- /dev/null +++ b/models/apple/mobilevit-small-garbage/train_results.json @@ -0,0 +1,8 @@ +{ + "epoch": 5.0, + "total_flos": 5.153016901533696e+16, + "train_loss": 0.5183452816696854, + "train_runtime": 159.3173, + "train_samples_per_second": 55.487, + "train_steps_per_second": 3.484 +} \ No newline at end of file diff --git a/models/apple/mobilevit-small-garbage/trainer_state.json b/models/apple/mobilevit-small-garbage/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..dc54ff1861c04e0cdee6b781fcd4b3b951629691 --- /dev/null +++ b/models/apple/mobilevit-small-garbage/trainer_state.json @@ -0,0 +1,400 @@ +{ + "best_metric": 0.26617246866226196, + "best_model_checkpoint": "./apple/mobilevit-small-garbage/checkpoint-500", + "epoch": 5.0, + "global_step": 555, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.09, + "learning_rate": 0.0001963963963963964, + "loss": 1.7495, + "step": 10 + }, + { + "epoch": 0.18, + "learning_rate": 0.00019279279279279282, + "loss": 1.6591, + "step": 20 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001891891891891892, + "loss": 1.5418, + "step": 30 + }, + { + "epoch": 0.36, + "learning_rate": 0.00018594594594594597, + "loss": 1.3934, + "step": 40 + }, + { + "epoch": 0.45, + "learning_rate": 0.00018234234234234235, + "loss": 1.2802, + "step": 50 + }, + { + "epoch": 0.54, + "learning_rate": 0.00017873873873873876, + "loss": 1.18, + "step": 60 + }, + { + "epoch": 0.63, + "learning_rate": 0.00017513513513513516, + "loss": 0.9854, + "step": 70 + }, + { + "epoch": 0.72, + "learning_rate": 0.00017153153153153154, + "loss": 0.9527, + "step": 80 + }, + { + "epoch": 0.81, + "learning_rate": 0.00016792792792792794, + "loss": 0.8574, + "step": 90 + }, + { + "epoch": 0.9, + "learning_rate": 0.00016432432432432435, + "loss": 0.8589, + "step": 100 + }, + { + "epoch": 0.9, + "eval_accuracy": 0.8079268292682927, + "eval_loss": 0.661000669002533, + "eval_runtime": 4.4994, + "eval_samples_per_second": 72.898, + "eval_steps_per_second": 9.112, + "step": 100 + }, + { + "epoch": 0.99, + "learning_rate": 0.0001610810810810811, + "loss": 0.7726, + "step": 110 + }, + { + "epoch": 1.08, + "learning_rate": 0.00015747747747747747, + "loss": 0.74, + "step": 120 + }, + { + "epoch": 1.17, + "learning_rate": 0.00015387387387387388, + "loss": 0.6325, + "step": 130 + }, + { + "epoch": 1.26, + "learning_rate": 0.00015063063063063063, + "loss": 0.5792, + "step": 140 + }, + { + "epoch": 1.35, + "learning_rate": 0.00014702702702702703, + "loss": 0.5582, + "step": 150 + }, + { + "epoch": 1.44, + "learning_rate": 0.00014342342342342344, + "loss": 0.5644, + "step": 160 + }, + { + "epoch": 1.53, + "learning_rate": 0.00013981981981981982, + "loss": 0.4776, + "step": 170 + }, + { + "epoch": 1.62, + "learning_rate": 0.00013621621621621622, + "loss": 0.4537, + "step": 180 + }, + { + "epoch": 1.71, + "learning_rate": 0.00013261261261261263, + "loss": 0.4813, + "step": 190 + }, + { + "epoch": 1.8, + "learning_rate": 0.000129009009009009, + "loss": 0.4676, + "step": 200 + }, + { + "epoch": 1.8, + "eval_accuracy": 0.8597560975609756, + "eval_loss": 0.4162694811820984, + "eval_runtime": 4.3992, + "eval_samples_per_second": 74.559, + "eval_steps_per_second": 9.32, + "step": 200 + }, + { + "epoch": 1.89, + "learning_rate": 0.0001254054054054054, + "loss": 0.5374, + "step": 210 + }, + { + "epoch": 1.98, + "learning_rate": 0.0001218018018018018, + "loss": 0.4006, + "step": 220 + }, + { + "epoch": 2.07, + "learning_rate": 0.00011855855855855858, + "loss": 0.4236, + "step": 230 + }, + { + "epoch": 2.16, + "learning_rate": 0.00011495495495495497, + "loss": 0.4494, + "step": 240 + }, + { + "epoch": 2.25, + "learning_rate": 0.00011135135135135135, + "loss": 0.3139, + "step": 250 + }, + { + "epoch": 2.34, + "learning_rate": 0.00010774774774774776, + "loss": 0.3151, + "step": 260 + }, + { + "epoch": 2.43, + "learning_rate": 0.00010414414414414416, + "loss": 0.2666, + "step": 270 + }, + { + "epoch": 2.52, + "learning_rate": 0.00010054054054054053, + "loss": 0.404, + "step": 280 + }, + { + "epoch": 2.61, + "learning_rate": 9.693693693693694e-05, + "loss": 0.3745, + "step": 290 + }, + { + "epoch": 2.7, + "learning_rate": 9.333333333333334e-05, + "loss": 0.3111, + "step": 300 + }, + { + "epoch": 2.7, + "eval_accuracy": 0.8932926829268293, + "eval_loss": 0.3442399501800537, + "eval_runtime": 4.2748, + "eval_samples_per_second": 76.729, + "eval_steps_per_second": 9.591, + "step": 300 + }, + { + "epoch": 2.79, + "learning_rate": 8.972972972972973e-05, + "loss": 0.3372, + "step": 310 + }, + { + "epoch": 2.88, + "learning_rate": 8.612612612612613e-05, + "loss": 0.4432, + "step": 320 + }, + { + "epoch": 2.97, + "learning_rate": 8.252252252252253e-05, + "loss": 0.3007, + "step": 330 + }, + { + "epoch": 3.06, + "learning_rate": 7.891891891891892e-05, + "loss": 0.2942, + "step": 340 + }, + { + "epoch": 3.15, + "learning_rate": 7.531531531531531e-05, + "loss": 0.2356, + "step": 350 + }, + { + "epoch": 3.24, + "learning_rate": 7.171171171171172e-05, + "loss": 0.2971, + "step": 360 + }, + { + "epoch": 3.33, + "learning_rate": 6.810810810810811e-05, + "loss": 0.2004, + "step": 370 + }, + { + "epoch": 3.42, + "learning_rate": 6.45045045045045e-05, + "loss": 0.3405, + "step": 380 + }, + { + "epoch": 3.51, + "learning_rate": 6.09009009009009e-05, + "loss": 0.3609, + "step": 390 + }, + { + "epoch": 3.6, + "learning_rate": 5.7297297297297305e-05, + "loss": 0.3709, + "step": 400 + }, + { + "epoch": 3.6, + "eval_accuracy": 0.9146341463414634, + "eval_loss": 0.3073720932006836, + "eval_runtime": 4.3122, + "eval_samples_per_second": 76.064, + "eval_steps_per_second": 9.508, + "step": 400 + }, + { + "epoch": 3.69, + "learning_rate": 5.369369369369369e-05, + "loss": 0.3094, + "step": 410 + }, + { + "epoch": 3.78, + "learning_rate": 5.009009009009009e-05, + "loss": 0.2294, + "step": 420 + }, + { + "epoch": 3.87, + "learning_rate": 4.648648648648649e-05, + "loss": 0.3965, + "step": 430 + }, + { + "epoch": 3.96, + "learning_rate": 4.2882882882882885e-05, + "loss": 0.2148, + "step": 440 + }, + { + "epoch": 4.05, + "learning_rate": 3.927927927927928e-05, + "loss": 0.3167, + "step": 450 + }, + { + "epoch": 4.14, + "learning_rate": 3.567567567567568e-05, + "loss": 0.1351, + "step": 460 + }, + { + "epoch": 4.23, + "learning_rate": 3.207207207207207e-05, + "loss": 0.275, + "step": 470 + }, + { + "epoch": 4.32, + "learning_rate": 2.8468468468468467e-05, + "loss": 0.1743, + "step": 480 + }, + { + "epoch": 4.41, + "learning_rate": 2.486486486486487e-05, + "loss": 0.1459, + "step": 490 + }, + { + "epoch": 4.5, + "learning_rate": 2.126126126126126e-05, + "loss": 0.1942, + "step": 500 + }, + { + "epoch": 4.5, + "eval_accuracy": 0.9115853658536586, + "eval_loss": 0.26617246866226196, + "eval_runtime": 4.2639, + "eval_samples_per_second": 76.925, + "eval_steps_per_second": 9.616, + "step": 500 + }, + { + "epoch": 4.59, + "learning_rate": 1.765765765765766e-05, + "loss": 0.1843, + "step": 510 + }, + { + "epoch": 4.68, + "learning_rate": 1.4054054054054055e-05, + "loss": 0.3644, + "step": 520 + }, + { + "epoch": 4.77, + "learning_rate": 1.0450450450450452e-05, + "loss": 0.1587, + "step": 530 + }, + { + "epoch": 4.86, + "learning_rate": 6.846846846846847e-06, + "loss": 0.1534, + "step": 540 + }, + { + "epoch": 4.95, + "learning_rate": 3.2432432432432437e-06, + "loss": 0.1954, + "step": 550 + }, + { + "epoch": 5.0, + "step": 555, + "total_flos": 5.153016901533696e+16, + "train_loss": 0.5183452816696854, + "train_runtime": 159.3173, + "train_samples_per_second": 55.487, + "train_steps_per_second": 3.484 + } + ], + "max_steps": 555, + "num_train_epochs": 5, + "total_flos": 5.153016901533696e+16, + "trial_name": null, + "trial_params": null +} diff --git a/models/apple/mobilevit-small-garbage/training_args.bin b/models/apple/mobilevit-small-garbage/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..fe239bd4ed2e2186ab566ad1c492723d402ea6ae --- /dev/null +++ b/models/apple/mobilevit-small-garbage/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:caccd1c746faf29e8da18b5a35181f77e8b3e0ebbfc29f59627f281fbd6e21d0 +size 3439 diff --git a/models/vit-base-garbage/.DS_Store b/models/vit-base-garbage/.DS_Store new file mode 100644 index 0000000000000000000000000000000000000000..85f2f944694cc1191e86f080e287661d39ef0d0d Binary files /dev/null and b/models/vit-base-garbage/.DS_Store differ diff --git a/models/vit-base-garbage/all_results.json b/models/vit-base-garbage/all_results.json new file mode 100644 index 0000000000000000000000000000000000000000..c20ababe2fc000a5ca5a78bec3aa1c6105bd5208 --- /dev/null +++ b/models/vit-base-garbage/all_results.json @@ -0,0 +1,16 @@ +{ + "epoch": 4.0, + "eval_accuracy": 0.9651972157772621, + "eval_f1": 0.9598970898619156, + "eval_loss": 0.15073566138744354, + "eval_precision": 0.9580334762339824, + "eval_recall": 0.9624065831267469, + "eval_runtime": 8.4033, + "eval_samples_per_second": 51.289, + "eval_steps_per_second": 6.426, + "total_flos": 5.480429975941939e+17, + "train_loss": 0.2510145430092339, + "train_runtime": 644.0352, + "train_samples_per_second": 10.981, + "train_steps_per_second": 0.689 +} \ No newline at end of file diff --git a/models/vit-base-garbage/checkpoint-300/config.json b/models/vit-base-garbage/checkpoint-300/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1f477f30ccda8e2f179f7673b37ea15af01399e9 --- /dev/null +++ b/models/vit-base-garbage/checkpoint-300/config.json @@ -0,0 +1,40 @@ +{ + "_name_or_path": "google/vit-base-patch16-224-in21k", + "architectures": [ + "ViTForImageClassification" + ], + "attention_probs_dropout_prob": 0.0, + "encoder_stride": 16, + "hidden_act": "gelu", + "hidden_dropout_prob": 0.0, + "hidden_size": 768, + "id2label": { + "0": "cardboard", + "1": "glass", + "2": "metal", + "3": "paper", + "4": "plastic", + "5": "trash" + }, + "image_size": 224, + "initializer_range": 0.02, + "intermediate_size": 3072, + "label2id": { + "cardboard": "0", + "glass": "1", + "metal": "2", + "paper": "3", + "plastic": "4", + "trash": "5" + }, + "layer_norm_eps": 1e-12, + "model_type": "vit", + "num_attention_heads": 12, + "num_channels": 3, + "num_hidden_layers": 12, + "patch_size": 16, + "problem_type": "single_label_classification", + "qkv_bias": true, + "torch_dtype": "float32", + "transformers_version": "4.24.0" +} diff --git a/models/vit-base-garbage/checkpoint-300/optimizer.pt b/models/vit-base-garbage/checkpoint-300/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..0e38e8eca59b2c8aa15bc24f1fb5d2b47b236c41 --- /dev/null +++ b/models/vit-base-garbage/checkpoint-300/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:45d91074a9863dc2fcba115a7d96f102c35d41ab1052c1055d4c491aa3157eb5 +size 686542177 diff --git a/models/vit-base-garbage/checkpoint-300/preprocessor_config.json b/models/vit-base-garbage/checkpoint-300/preprocessor_config.json new file mode 100644 index 0000000000000000000000000000000000000000..b7414e73cf93e2818ed2c82d3d7bfc0d85991c13 --- /dev/null +++ b/models/vit-base-garbage/checkpoint-300/preprocessor_config.json @@ -0,0 +1,17 @@ +{ + "do_normalize": true, + "do_resize": true, + "feature_extractor_type": "ViTFeatureExtractor", + "image_mean": [ + 0.5, + 0.5, + 0.5 + ], + "image_std": [ + 0.5, + 0.5, + 0.5 + ], + "resample": 2, + "size": 224 +} diff --git a/models/vit-base-garbage/checkpoint-300/pytorch_model.bin b/models/vit-base-garbage/checkpoint-300/pytorch_model.bin new file mode 100644 index 0000000000000000000000000000000000000000..0ff631430341e1dac7e71ff465692ec60f82edc3 --- /dev/null +++ b/models/vit-base-garbage/checkpoint-300/pytorch_model.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b2bd1080c7a3858edda617a6b2e4814025d90da3f24591c916c4ca3618c247a6 +size 343279281 diff --git a/models/vit-base-garbage/checkpoint-300/rng_state.pth b/models/vit-base-garbage/checkpoint-300/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..3d890aa3229c86cd01884d1a2db43f99eea0f032 --- /dev/null +++ b/models/vit-base-garbage/checkpoint-300/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c2ab9b52e364e49a6593a765f851264b1dd5e9b0207961295e70ed4788ef648c +size 14503 diff --git a/models/vit-base-garbage/checkpoint-300/scaler.pt b/models/vit-base-garbage/checkpoint-300/scaler.pt new file mode 100644 index 0000000000000000000000000000000000000000..01066cf4761ea9d2f7962f5181762f7b08690b79 --- /dev/null +++ b/models/vit-base-garbage/checkpoint-300/scaler.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0967b9f865f16344c55f5ccc3cf7d6e8e97ca61dda304e931ca6bad130f48dd1 +size 559 diff --git a/models/vit-base-garbage/checkpoint-300/scheduler.pt b/models/vit-base-garbage/checkpoint-300/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..687646fadbbc37d5044bc5c689c761a1cc30c7ed --- /dev/null +++ b/models/vit-base-garbage/checkpoint-300/scheduler.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7d806fe60265194e6eba06de83a8b30480bcb818797ce556e7298374edc1993a +size 623 diff --git a/models/vit-base-garbage/checkpoint-300/trainer_state.json b/models/vit-base-garbage/checkpoint-300/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..e36d5ac475a55595b7df0dbf0399d049f62957e7 --- /dev/null +++ b/models/vit-base-garbage/checkpoint-300/trainer_state.json @@ -0,0 +1,223 @@ +{ + "best_metric": 0.2087339609861374, + "best_model_checkpoint": "./vit-base-beans/checkpoint-300", + "epoch": 2.7027027027027026, + "global_step": 300, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.09, + "learning_rate": 0.0001954954954954955, + "loss": 1.552, + "step": 10 + }, + { + "epoch": 0.18, + "learning_rate": 0.000190990990990991, + "loss": 1.0968, + "step": 20 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001864864864864865, + "loss": 0.8214, + "step": 30 + }, + { + "epoch": 0.36, + "learning_rate": 0.000181981981981982, + "loss": 0.655, + "step": 40 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001774774774774775, + "loss": 0.5131, + "step": 50 + }, + { + "epoch": 0.54, + "learning_rate": 0.000172972972972973, + "loss": 0.5863, + "step": 60 + }, + { + "epoch": 0.63, + "learning_rate": 0.00016846846846846846, + "loss": 0.4228, + "step": 70 + }, + { + "epoch": 0.72, + "learning_rate": 0.00016396396396396395, + "loss": 0.4562, + "step": 80 + }, + { + "epoch": 0.81, + "learning_rate": 0.00015945945945945947, + "loss": 0.4817, + "step": 90 + }, + { + "epoch": 0.9, + "learning_rate": 0.00015495495495495496, + "loss": 0.5217, + "step": 100 + }, + { + "epoch": 0.9, + "eval_accuracy": 0.8871951219512195, + "eval_loss": 0.37025415897369385, + "eval_runtime": 82.1306, + "eval_samples_per_second": 3.994, + "eval_steps_per_second": 0.499, + "step": 100 + }, + { + "epoch": 0.99, + "learning_rate": 0.00015045045045045046, + "loss": 0.4543, + "step": 110 + }, + { + "epoch": 1.08, + "learning_rate": 0.00014594594594594595, + "loss": 0.2195, + "step": 120 + }, + { + "epoch": 1.17, + "learning_rate": 0.00014144144144144144, + "loss": 0.2413, + "step": 130 + }, + { + "epoch": 1.26, + "learning_rate": 0.00013693693693693693, + "loss": 0.2153, + "step": 140 + }, + { + "epoch": 1.35, + "learning_rate": 0.00013243243243243243, + "loss": 0.2376, + "step": 150 + }, + { + "epoch": 1.44, + "learning_rate": 0.00012792792792792795, + "loss": 0.215, + "step": 160 + }, + { + "epoch": 1.53, + "learning_rate": 0.00012342342342342344, + "loss": 0.1317, + "step": 170 + }, + { + "epoch": 1.62, + "learning_rate": 0.00011891891891891893, + "loss": 0.143, + "step": 180 + }, + { + "epoch": 1.71, + "learning_rate": 0.00011441441441441443, + "loss": 0.2173, + "step": 190 + }, + { + "epoch": 1.8, + "learning_rate": 0.00010990990990990993, + "loss": 0.1637, + "step": 200 + }, + { + "epoch": 1.8, + "eval_accuracy": 0.8597560975609756, + "eval_loss": 0.5076180696487427, + "eval_runtime": 4.074, + "eval_samples_per_second": 80.511, + "eval_steps_per_second": 10.064, + "step": 200 + }, + { + "epoch": 1.89, + "learning_rate": 0.0001054054054054054, + "loss": 0.3325, + "step": 210 + }, + { + "epoch": 1.98, + "learning_rate": 0.00010090090090090089, + "loss": 0.2099, + "step": 220 + }, + { + "epoch": 2.07, + "learning_rate": 9.639639639639641e-05, + "loss": 0.1132, + "step": 230 + }, + { + "epoch": 2.16, + "learning_rate": 9.18918918918919e-05, + "loss": 0.1517, + "step": 240 + }, + { + "epoch": 2.25, + "learning_rate": 8.738738738738738e-05, + "loss": 0.0624, + "step": 250 + }, + { + "epoch": 2.34, + "learning_rate": 8.288288288288289e-05, + "loss": 0.0905, + "step": 260 + }, + { + "epoch": 2.43, + "learning_rate": 7.837837837837838e-05, + "loss": 0.0586, + "step": 270 + }, + { + "epoch": 2.52, + "learning_rate": 7.387387387387387e-05, + "loss": 0.071, + "step": 280 + }, + { + "epoch": 2.61, + "learning_rate": 6.936936936936938e-05, + "loss": 0.0393, + "step": 290 + }, + { + "epoch": 2.7, + "learning_rate": 6.486486486486487e-05, + "loss": 0.0291, + "step": 300 + }, + { + "epoch": 2.7, + "eval_accuracy": 0.9420731707317073, + "eval_loss": 0.2087339609861374, + "eval_runtime": 4.0634, + "eval_samples_per_second": 80.72, + "eval_steps_per_second": 10.09, + "step": 300 + } + ], + "max_steps": 444, + "num_train_epochs": 4, + "total_flos": 3.7073496896077824e+17, + "trial_name": null, + "trial_params": null +} diff --git a/models/vit-base-garbage/checkpoint-300/training_args.bin b/models/vit-base-garbage/checkpoint-300/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..bcf44ad847b818b017359c02a0e900aa60a1322d --- /dev/null +++ b/models/vit-base-garbage/checkpoint-300/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b388b03a0bc352599c3f1a7b29932b4fb98d8f608da2c3b16a52bf3c7baf4496 +size 3375 diff --git a/models/vit-base-garbage/checkpoint-400/config.json b/models/vit-base-garbage/checkpoint-400/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1f477f30ccda8e2f179f7673b37ea15af01399e9 --- /dev/null +++ b/models/vit-base-garbage/checkpoint-400/config.json @@ -0,0 +1,40 @@ +{ + "_name_or_path": "google/vit-base-patch16-224-in21k", + "architectures": [ + "ViTForImageClassification" + ], + "attention_probs_dropout_prob": 0.0, + "encoder_stride": 16, + "hidden_act": "gelu", + "hidden_dropout_prob": 0.0, + "hidden_size": 768, + "id2label": { + "0": "cardboard", + "1": "glass", + "2": "metal", + "3": "paper", + "4": "plastic", + "5": "trash" + }, + "image_size": 224, + "initializer_range": 0.02, + "intermediate_size": 3072, + "label2id": { + "cardboard": "0", + "glass": "1", + "metal": "2", + "paper": "3", + "plastic": "4", + "trash": "5" + }, + "layer_norm_eps": 1e-12, + "model_type": "vit", + "num_attention_heads": 12, + "num_channels": 3, + "num_hidden_layers": 12, + "patch_size": 16, + "problem_type": "single_label_classification", + "qkv_bias": true, + "torch_dtype": "float32", + "transformers_version": "4.24.0" +} diff --git a/models/vit-base-garbage/checkpoint-400/optimizer.pt b/models/vit-base-garbage/checkpoint-400/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..c24df391a6b2465290d87f78961dcfa9d0a6a853 --- /dev/null +++ b/models/vit-base-garbage/checkpoint-400/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bd9df6a4f363cff65999f1dd47fad0cee5d4a5a09ffe3efb82953939a0f36237 +size 686542177 diff --git a/models/vit-base-garbage/checkpoint-400/preprocessor_config.json b/models/vit-base-garbage/checkpoint-400/preprocessor_config.json new file mode 100644 index 0000000000000000000000000000000000000000..b7414e73cf93e2818ed2c82d3d7bfc0d85991c13 --- /dev/null +++ b/models/vit-base-garbage/checkpoint-400/preprocessor_config.json @@ -0,0 +1,17 @@ +{ + "do_normalize": true, + "do_resize": true, + "feature_extractor_type": "ViTFeatureExtractor", + "image_mean": [ + 0.5, + 0.5, + 0.5 + ], + "image_std": [ + 0.5, + 0.5, + 0.5 + ], + "resample": 2, + "size": 224 +} diff --git a/models/vit-base-garbage/checkpoint-400/pytorch_model.bin b/models/vit-base-garbage/checkpoint-400/pytorch_model.bin new file mode 100644 index 0000000000000000000000000000000000000000..642eaf0c8cf588ee787f521e0ea7b6471e7b1114 --- /dev/null +++ b/models/vit-base-garbage/checkpoint-400/pytorch_model.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b61389241626facd22ef1a4236cf35bfbdcd4c59cf1e2a55cd438f5472ee1ab0 +size 343279281 diff --git a/models/vit-base-garbage/checkpoint-400/rng_state.pth b/models/vit-base-garbage/checkpoint-400/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..968c4beab115410a386c306620bb113cb38161a1 --- /dev/null +++ b/models/vit-base-garbage/checkpoint-400/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cc81a4ed07eaae97e601e6dad9894c368270413e9040807580143d45f03df338 +size 14503 diff --git a/models/vit-base-garbage/checkpoint-400/scaler.pt b/models/vit-base-garbage/checkpoint-400/scaler.pt new file mode 100644 index 0000000000000000000000000000000000000000..9c7aef4199e98d81152810b661dbaffc01963383 --- /dev/null +++ b/models/vit-base-garbage/checkpoint-400/scaler.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:476e510c8ea7edbd2b51d1e76a4e037820a5639381c0d8b5d32dafa492795a1e +size 559 diff --git a/models/vit-base-garbage/checkpoint-400/scheduler.pt b/models/vit-base-garbage/checkpoint-400/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..6a433d0484cf188147ef42d2f239947877a40f02 --- /dev/null +++ b/models/vit-base-garbage/checkpoint-400/scheduler.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c5a43e941fea784321070831dd2f6b1bc4f69813f39be747b387fb6a717115bf +size 623 diff --git a/models/vit-base-garbage/checkpoint-400/trainer_state.json b/models/vit-base-garbage/checkpoint-400/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..db7c7b8d9ace2363b0cac89b8672c46f5a29a554 --- /dev/null +++ b/models/vit-base-garbage/checkpoint-400/trainer_state.json @@ -0,0 +1,292 @@ +{ + "best_metric": 0.172527477145195, + "best_model_checkpoint": "./vit-base-beans/checkpoint-400", + "epoch": 3.6036036036036037, + "global_step": 400, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.09, + "learning_rate": 0.0001954954954954955, + "loss": 1.552, + "step": 10 + }, + { + "epoch": 0.18, + "learning_rate": 0.000190990990990991, + "loss": 1.0968, + "step": 20 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001864864864864865, + "loss": 0.8214, + "step": 30 + }, + { + "epoch": 0.36, + "learning_rate": 0.000181981981981982, + "loss": 0.655, + "step": 40 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001774774774774775, + "loss": 0.5131, + "step": 50 + }, + { + "epoch": 0.54, + "learning_rate": 0.000172972972972973, + "loss": 0.5863, + "step": 60 + }, + { + "epoch": 0.63, + "learning_rate": 0.00016846846846846846, + "loss": 0.4228, + "step": 70 + }, + { + "epoch": 0.72, + "learning_rate": 0.00016396396396396395, + "loss": 0.4562, + "step": 80 + }, + { + "epoch": 0.81, + "learning_rate": 0.00015945945945945947, + "loss": 0.4817, + "step": 90 + }, + { + "epoch": 0.9, + "learning_rate": 0.00015495495495495496, + "loss": 0.5217, + "step": 100 + }, + { + "epoch": 0.9, + "eval_accuracy": 0.8871951219512195, + "eval_loss": 0.37025415897369385, + "eval_runtime": 82.1306, + "eval_samples_per_second": 3.994, + "eval_steps_per_second": 0.499, + "step": 100 + }, + { + "epoch": 0.99, + "learning_rate": 0.00015045045045045046, + "loss": 0.4543, + "step": 110 + }, + { + "epoch": 1.08, + "learning_rate": 0.00014594594594594595, + "loss": 0.2195, + "step": 120 + }, + { + "epoch": 1.17, + "learning_rate": 0.00014144144144144144, + "loss": 0.2413, + "step": 130 + }, + { + "epoch": 1.26, + "learning_rate": 0.00013693693693693693, + "loss": 0.2153, + "step": 140 + }, + { + "epoch": 1.35, + "learning_rate": 0.00013243243243243243, + "loss": 0.2376, + "step": 150 + }, + { + "epoch": 1.44, + "learning_rate": 0.00012792792792792795, + "loss": 0.215, + "step": 160 + }, + { + "epoch": 1.53, + "learning_rate": 0.00012342342342342344, + "loss": 0.1317, + "step": 170 + }, + { + "epoch": 1.62, + "learning_rate": 0.00011891891891891893, + "loss": 0.143, + "step": 180 + }, + { + "epoch": 1.71, + "learning_rate": 0.00011441441441441443, + "loss": 0.2173, + "step": 190 + }, + { + "epoch": 1.8, + "learning_rate": 0.00010990990990990993, + "loss": 0.1637, + "step": 200 + }, + { + "epoch": 1.8, + "eval_accuracy": 0.8597560975609756, + "eval_loss": 0.5076180696487427, + "eval_runtime": 4.074, + "eval_samples_per_second": 80.511, + "eval_steps_per_second": 10.064, + "step": 200 + }, + { + "epoch": 1.89, + "learning_rate": 0.0001054054054054054, + "loss": 0.3325, + "step": 210 + }, + { + "epoch": 1.98, + "learning_rate": 0.00010090090090090089, + "loss": 0.2099, + "step": 220 + }, + { + "epoch": 2.07, + "learning_rate": 9.639639639639641e-05, + "loss": 0.1132, + "step": 230 + }, + { + "epoch": 2.16, + "learning_rate": 9.18918918918919e-05, + "loss": 0.1517, + "step": 240 + }, + { + "epoch": 2.25, + "learning_rate": 8.738738738738738e-05, + "loss": 0.0624, + "step": 250 + }, + { + "epoch": 2.34, + "learning_rate": 8.288288288288289e-05, + "loss": 0.0905, + "step": 260 + }, + { + "epoch": 2.43, + "learning_rate": 7.837837837837838e-05, + "loss": 0.0586, + "step": 270 + }, + { + "epoch": 2.52, + "learning_rate": 7.387387387387387e-05, + "loss": 0.071, + "step": 280 + }, + { + "epoch": 2.61, + "learning_rate": 6.936936936936938e-05, + "loss": 0.0393, + "step": 290 + }, + { + "epoch": 2.7, + "learning_rate": 6.486486486486487e-05, + "loss": 0.0291, + "step": 300 + }, + { + "epoch": 2.7, + "eval_accuracy": 0.9420731707317073, + "eval_loss": 0.2087339609861374, + "eval_runtime": 4.0634, + "eval_samples_per_second": 80.72, + "eval_steps_per_second": 10.09, + "step": 300 + }, + { + "epoch": 2.79, + "learning_rate": 6.0360360360360365e-05, + "loss": 0.075, + "step": 310 + }, + { + "epoch": 2.88, + "learning_rate": 5.585585585585585e-05, + "loss": 0.1357, + "step": 320 + }, + { + "epoch": 2.97, + "learning_rate": 5.135135135135135e-05, + "loss": 0.0312, + "step": 330 + }, + { + "epoch": 3.06, + "learning_rate": 4.684684684684685e-05, + "loss": 0.0753, + "step": 340 + }, + { + "epoch": 3.15, + "learning_rate": 4.234234234234234e-05, + "loss": 0.0231, + "step": 350 + }, + { + "epoch": 3.24, + "learning_rate": 3.783783783783784e-05, + "loss": 0.0266, + "step": 360 + }, + { + "epoch": 3.33, + "learning_rate": 3.3333333333333335e-05, + "loss": 0.023, + "step": 370 + }, + { + "epoch": 3.42, + "learning_rate": 2.882882882882883e-05, + "loss": 0.0224, + "step": 380 + }, + { + "epoch": 3.51, + "learning_rate": 2.4324324324324327e-05, + "loss": 0.0251, + "step": 390 + }, + { + "epoch": 3.6, + "learning_rate": 1.981981981981982e-05, + "loss": 0.025, + "step": 400 + }, + { + "epoch": 3.6, + "eval_accuracy": 0.9573170731707317, + "eval_loss": 0.172527477145195, + "eval_runtime": 4.0556, + "eval_samples_per_second": 80.877, + "eval_steps_per_second": 10.11, + "step": 400 + } + ], + "max_steps": 444, + "num_train_epochs": 4, + "total_flos": 4.9410663923367936e+17, + "trial_name": null, + "trial_params": null +} diff --git a/models/vit-base-garbage/checkpoint-400/training_args.bin b/models/vit-base-garbage/checkpoint-400/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..bcf44ad847b818b017359c02a0e900aa60a1322d --- /dev/null +++ b/models/vit-base-garbage/checkpoint-400/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b388b03a0bc352599c3f1a7b29932b4fb98d8f608da2c3b16a52bf3c7baf4496 +size 3375 diff --git a/models/vit-base-garbage/config.json b/models/vit-base-garbage/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1f477f30ccda8e2f179f7673b37ea15af01399e9 --- /dev/null +++ b/models/vit-base-garbage/config.json @@ -0,0 +1,40 @@ +{ + "_name_or_path": "google/vit-base-patch16-224-in21k", + "architectures": [ + "ViTForImageClassification" + ], + "attention_probs_dropout_prob": 0.0, + "encoder_stride": 16, + "hidden_act": "gelu", + "hidden_dropout_prob": 0.0, + "hidden_size": 768, + "id2label": { + "0": "cardboard", + "1": "glass", + "2": "metal", + "3": "paper", + "4": "plastic", + "5": "trash" + }, + "image_size": 224, + "initializer_range": 0.02, + "intermediate_size": 3072, + "label2id": { + "cardboard": "0", + "glass": "1", + "metal": "2", + "paper": "3", + "plastic": "4", + "trash": "5" + }, + "layer_norm_eps": 1e-12, + "model_type": "vit", + "num_attention_heads": 12, + "num_channels": 3, + "num_hidden_layers": 12, + "patch_size": 16, + "problem_type": "single_label_classification", + "qkv_bias": true, + "torch_dtype": "float32", + "transformers_version": "4.24.0" +} diff --git a/models/vit-base-garbage/preprocessor_config.json b/models/vit-base-garbage/preprocessor_config.json new file mode 100644 index 0000000000000000000000000000000000000000..b7414e73cf93e2818ed2c82d3d7bfc0d85991c13 --- /dev/null +++ b/models/vit-base-garbage/preprocessor_config.json @@ -0,0 +1,17 @@ +{ + "do_normalize": true, + "do_resize": true, + "feature_extractor_type": "ViTFeatureExtractor", + "image_mean": [ + 0.5, + 0.5, + 0.5 + ], + "image_std": [ + 0.5, + 0.5, + 0.5 + ], + "resample": 2, + "size": 224 +} diff --git a/models/vit-base-garbage/pytorch_model.bin b/models/vit-base-garbage/pytorch_model.bin new file mode 100644 index 0000000000000000000000000000000000000000..642eaf0c8cf588ee787f521e0ea7b6471e7b1114 --- /dev/null +++ b/models/vit-base-garbage/pytorch_model.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b61389241626facd22ef1a4236cf35bfbdcd4c59cf1e2a55cd438f5472ee1ab0 +size 343279281 diff --git a/models/vit-base-garbage/runs/Dec12_19-10-35_ed23ba8f6546/1670872238.0395455/events.out.tfevents.1670872238.ed23ba8f6546.76.2 b/models/vit-base-garbage/runs/Dec12_19-10-35_ed23ba8f6546/1670872238.0395455/events.out.tfevents.1670872238.ed23ba8f6546.76.2 new file mode 100644 index 0000000000000000000000000000000000000000..8068e0c5e6301e6c9a5ab544d7f29c307c2bfaf2 --- /dev/null +++ b/models/vit-base-garbage/runs/Dec12_19-10-35_ed23ba8f6546/1670872238.0395455/events.out.tfevents.1670872238.ed23ba8f6546.76.2 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c6fad8c207a1a9f50421bd83fe5c5536f74f143ef54da571f20c56126e9e0654 +size 5513 diff --git a/models/vit-base-garbage/runs/Dec12_19-10-35_ed23ba8f6546/events.out.tfevents.1670872238.ed23ba8f6546.76.1 b/models/vit-base-garbage/runs/Dec12_19-10-35_ed23ba8f6546/events.out.tfevents.1670872238.ed23ba8f6546.76.1 new file mode 100644 index 0000000000000000000000000000000000000000..e5a416bde828253bdb9543cf3bf21313773da315 --- /dev/null +++ b/models/vit-base-garbage/runs/Dec12_19-10-35_ed23ba8f6546/events.out.tfevents.1670872238.ed23ba8f6546.76.1 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e69af10d73925b30f110cdad2f552fe6c74f9e26c2b71eea25bd103822c32c6c +size 3944 diff --git a/models/vit-base-garbage/runs/Dec12_19-10-35_ed23ba8f6546/events.out.tfevents.1670872287.ed23ba8f6546.76.3 b/models/vit-base-garbage/runs/Dec12_19-10-35_ed23ba8f6546/events.out.tfevents.1670872287.ed23ba8f6546.76.3 new file mode 100644 index 0000000000000000000000000000000000000000..7264f12f97d80c8a02b1af87e948e70f7ce92378 --- /dev/null +++ b/models/vit-base-garbage/runs/Dec12_19-10-35_ed23ba8f6546/events.out.tfevents.1670872287.ed23ba8f6546.76.3 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:83f3175ad6971770c09cab115bd878851c1b6955a575e58a2f5ecd8f401aba66 +size 298 diff --git a/models/vit-base-garbage/runs/Dec12_21-11-38_30c96569bd60/events.out.tfevents.1670879785.30c96569bd60.71.0 b/models/vit-base-garbage/runs/Dec12_21-11-38_30c96569bd60/events.out.tfevents.1670879785.30c96569bd60.71.0 new file mode 100644 index 0000000000000000000000000000000000000000..51614e2373ea3e78e95933fa0d59a26b7e896c7d --- /dev/null +++ b/models/vit-base-garbage/runs/Dec12_21-11-38_30c96569bd60/events.out.tfevents.1670879785.30c96569bd60.71.0 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ea778ed9b6506059fdb91fa8755db5f01a45726eb172139fc2a361675d576579 +size 298 diff --git a/models/vit-base-garbage/runs/Dec12_21-51-08_30c96569bd60/events.out.tfevents.1670881883.30c96569bd60.71.1 b/models/vit-base-garbage/runs/Dec12_21-51-08_30c96569bd60/events.out.tfevents.1670881883.30c96569bd60.71.1 new file mode 100644 index 0000000000000000000000000000000000000000..717c649d67cac195ced2a761d14ad2194bb3d295 --- /dev/null +++ b/models/vit-base-garbage/runs/Dec12_21-51-08_30c96569bd60/events.out.tfevents.1670881883.30c96569bd60.71.1 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a10eb065b3cacdcdb9c4be462f2ee61e3b114891cc8d09d374632916c531c688 +size 438 diff --git a/models/vit-base-garbage/runs/Nov28_22-29-34_09524550a2a1/1669674600.7132652/events.out.tfevents.1669674600.09524550a2a1.78.1 b/models/vit-base-garbage/runs/Nov28_22-29-34_09524550a2a1/1669674600.7132652/events.out.tfevents.1669674600.09524550a2a1.78.1 new file mode 100644 index 0000000000000000000000000000000000000000..b5b4c010a6d79cd0809fad5f2d301dc0f8b85e8e --- /dev/null +++ b/models/vit-base-garbage/runs/Nov28_22-29-34_09524550a2a1/1669674600.7132652/events.out.tfevents.1669674600.09524550a2a1.78.1 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c3071e168dc29887ffce431e759528cdfe8d5e8eabaeb83255bdb68d496698dd +size 5475 diff --git a/models/vit-base-garbage/runs/Nov28_22-29-34_09524550a2a1/events.out.tfevents.1669674600.09524550a2a1.78.0 b/models/vit-base-garbage/runs/Nov28_22-29-34_09524550a2a1/events.out.tfevents.1669674600.09524550a2a1.78.0 new file mode 100644 index 0000000000000000000000000000000000000000..c16ccbbc43c79e0dcf8655e2a6555d702e87b349 --- /dev/null +++ b/models/vit-base-garbage/runs/Nov28_22-29-34_09524550a2a1/events.out.tfevents.1669674600.09524550a2a1.78.0 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:44c53d7a0c3a26177b471447620ac4b32421f012e1266ab47e00e7e178d8457f +size 12357 diff --git a/models/vit-base-garbage/runs/Nov28_22-29-34_09524550a2a1/events.out.tfevents.1669675373.09524550a2a1.78.2 b/models/vit-base-garbage/runs/Nov28_22-29-34_09524550a2a1/events.out.tfevents.1669675373.09524550a2a1.78.2 new file mode 100644 index 0000000000000000000000000000000000000000..024f815e6bb6bd1187d04c6e35c9a4cc2b75bd27 --- /dev/null +++ b/models/vit-base-garbage/runs/Nov28_22-29-34_09524550a2a1/events.out.tfevents.1669675373.09524550a2a1.78.2 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5c92e28afdccf80c49989824349ed1e0c896187d5c174a4985340e2c792f76b3 +size 363 diff --git a/models/vit-base-garbage/runs/Nov30_07-00-26_a9e9695b4f94/events.out.tfevents.1669791773.a9e9695b4f94.79.0 b/models/vit-base-garbage/runs/Nov30_07-00-26_a9e9695b4f94/events.out.tfevents.1669791773.a9e9695b4f94.79.0 new file mode 100644 index 0000000000000000000000000000000000000000..d7731173d3384c11ae9c09fda102bb5dff6e240e --- /dev/null +++ b/models/vit-base-garbage/runs/Nov30_07-00-26_a9e9695b4f94/events.out.tfevents.1669791773.a9e9695b4f94.79.0 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f19fab374e46c81c6347c683af7886d08eb58fe303e3160506614401c5cd0b41 +size 298 diff --git a/models/vit-base-garbage/runs/Nov30_07-09-00_a9e9695b4f94/events.out.tfevents.1669792153.a9e9695b4f94.79.1 b/models/vit-base-garbage/runs/Nov30_07-09-00_a9e9695b4f94/events.out.tfevents.1669792153.a9e9695b4f94.79.1 new file mode 100644 index 0000000000000000000000000000000000000000..cc7d3eecf668471ed2c8abdc821813e9d5cbd9a4 --- /dev/null +++ b/models/vit-base-garbage/runs/Nov30_07-09-00_a9e9695b4f94/events.out.tfevents.1669792153.a9e9695b4f94.79.1 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b7c77abbcf48a99c88124c17d8c50cf75aaedb8cd913356b5ea9449e4ba14ec3 +size 299 diff --git a/models/vit-base-garbage/runs/Nov30_07-09-38_a9e9695b4f94/events.out.tfevents.1669792190.a9e9695b4f94.79.2 b/models/vit-base-garbage/runs/Nov30_07-09-38_a9e9695b4f94/events.out.tfevents.1669792190.a9e9695b4f94.79.2 new file mode 100644 index 0000000000000000000000000000000000000000..85d1cbd26e73f9e506249009942faedbf755adf8 --- /dev/null +++ b/models/vit-base-garbage/runs/Nov30_07-09-38_a9e9695b4f94/events.out.tfevents.1669792190.a9e9695b4f94.79.2 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:76099c5eab728a25ef961a15e182245cabee638998aab3b37a427fddd913f585 +size 292 diff --git a/models/vit-base-garbage/test_results.json b/models/vit-base-garbage/test_results.json new file mode 100644 index 0000000000000000000000000000000000000000..a72e1389bb34815c773ddf66bf909769223a5348 --- /dev/null +++ b/models/vit-base-garbage/test_results.json @@ -0,0 +1,10 @@ +{ + "eval_accuracy": 0.9651972157772621, + "eval_f1": 0.9598970898619156, + "eval_loss": 0.15073566138744354, + "eval_precision": 0.9580334762339824, + "eval_recall": 0.9624065831267469, + "eval_runtime": 8.4033, + "eval_samples_per_second": 51.289, + "eval_steps_per_second": 6.426 +} \ No newline at end of file diff --git a/models/vit-base-garbage/train_results.json b/models/vit-base-garbage/train_results.json new file mode 100644 index 0000000000000000000000000000000000000000..a4687dd984a78b4490193c8229c5e8cc4263fe65 --- /dev/null +++ b/models/vit-base-garbage/train_results.json @@ -0,0 +1,8 @@ +{ + "epoch": 4.0, + "total_flos": 5.480429975941939e+17, + "train_loss": 0.2510145430092339, + "train_runtime": 644.0352, + "train_samples_per_second": 10.981, + "train_steps_per_second": 0.689 +} \ No newline at end of file diff --git a/models/vit-base-garbage/trainer_state.json b/models/vit-base-garbage/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..5856ac0a4771576f1353c25448f766d4b2155d03 --- /dev/null +++ b/models/vit-base-garbage/trainer_state.json @@ -0,0 +1,325 @@ +{ + "best_metric": 0.172527477145195, + "best_model_checkpoint": "./vit-base-beans/checkpoint-400", + "epoch": 4.0, + "global_step": 444, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.09, + "learning_rate": 0.0001954954954954955, + "loss": 1.552, + "step": 10 + }, + { + "epoch": 0.18, + "learning_rate": 0.000190990990990991, + "loss": 1.0968, + "step": 20 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001864864864864865, + "loss": 0.8214, + "step": 30 + }, + { + "epoch": 0.36, + "learning_rate": 0.000181981981981982, + "loss": 0.655, + "step": 40 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001774774774774775, + "loss": 0.5131, + "step": 50 + }, + { + "epoch": 0.54, + "learning_rate": 0.000172972972972973, + "loss": 0.5863, + "step": 60 + }, + { + "epoch": 0.63, + "learning_rate": 0.00016846846846846846, + "loss": 0.4228, + "step": 70 + }, + { + "epoch": 0.72, + "learning_rate": 0.00016396396396396395, + "loss": 0.4562, + "step": 80 + }, + { + "epoch": 0.81, + "learning_rate": 0.00015945945945945947, + "loss": 0.4817, + "step": 90 + }, + { + "epoch": 0.9, + "learning_rate": 0.00015495495495495496, + "loss": 0.5217, + "step": 100 + }, + { + "epoch": 0.9, + "eval_accuracy": 0.8871951219512195, + "eval_loss": 0.37025415897369385, + "eval_runtime": 82.1306, + "eval_samples_per_second": 3.994, + "eval_steps_per_second": 0.499, + "step": 100 + }, + { + "epoch": 0.99, + "learning_rate": 0.00015045045045045046, + "loss": 0.4543, + "step": 110 + }, + { + "epoch": 1.08, + "learning_rate": 0.00014594594594594595, + "loss": 0.2195, + "step": 120 + }, + { + "epoch": 1.17, + "learning_rate": 0.00014144144144144144, + "loss": 0.2413, + "step": 130 + }, + { + "epoch": 1.26, + "learning_rate": 0.00013693693693693693, + "loss": 0.2153, + "step": 140 + }, + { + "epoch": 1.35, + "learning_rate": 0.00013243243243243243, + "loss": 0.2376, + "step": 150 + }, + { + "epoch": 1.44, + "learning_rate": 0.00012792792792792795, + "loss": 0.215, + "step": 160 + }, + { + "epoch": 1.53, + "learning_rate": 0.00012342342342342344, + "loss": 0.1317, + "step": 170 + }, + { + "epoch": 1.62, + "learning_rate": 0.00011891891891891893, + "loss": 0.143, + "step": 180 + }, + { + "epoch": 1.71, + "learning_rate": 0.00011441441441441443, + "loss": 0.2173, + "step": 190 + }, + { + "epoch": 1.8, + "learning_rate": 0.00010990990990990993, + "loss": 0.1637, + "step": 200 + }, + { + "epoch": 1.8, + "eval_accuracy": 0.8597560975609756, + "eval_loss": 0.5076180696487427, + "eval_runtime": 4.074, + "eval_samples_per_second": 80.511, + "eval_steps_per_second": 10.064, + "step": 200 + }, + { + "epoch": 1.89, + "learning_rate": 0.0001054054054054054, + "loss": 0.3325, + "step": 210 + }, + { + "epoch": 1.98, + "learning_rate": 0.00010090090090090089, + "loss": 0.2099, + "step": 220 + }, + { + "epoch": 2.07, + "learning_rate": 9.639639639639641e-05, + "loss": 0.1132, + "step": 230 + }, + { + "epoch": 2.16, + "learning_rate": 9.18918918918919e-05, + "loss": 0.1517, + "step": 240 + }, + { + "epoch": 2.25, + "learning_rate": 8.738738738738738e-05, + "loss": 0.0624, + "step": 250 + }, + { + "epoch": 2.34, + "learning_rate": 8.288288288288289e-05, + "loss": 0.0905, + "step": 260 + }, + { + "epoch": 2.43, + "learning_rate": 7.837837837837838e-05, + "loss": 0.0586, + "step": 270 + }, + { + "epoch": 2.52, + "learning_rate": 7.387387387387387e-05, + "loss": 0.071, + "step": 280 + }, + { + "epoch": 2.61, + "learning_rate": 6.936936936936938e-05, + "loss": 0.0393, + "step": 290 + }, + { + "epoch": 2.7, + "learning_rate": 6.486486486486487e-05, + "loss": 0.0291, + "step": 300 + }, + { + "epoch": 2.7, + "eval_accuracy": 0.9420731707317073, + "eval_loss": 0.2087339609861374, + "eval_runtime": 4.0634, + "eval_samples_per_second": 80.72, + "eval_steps_per_second": 10.09, + "step": 300 + }, + { + "epoch": 2.79, + "learning_rate": 6.0360360360360365e-05, + "loss": 0.075, + "step": 310 + }, + { + "epoch": 2.88, + "learning_rate": 5.585585585585585e-05, + "loss": 0.1357, + "step": 320 + }, + { + "epoch": 2.97, + "learning_rate": 5.135135135135135e-05, + "loss": 0.0312, + "step": 330 + }, + { + "epoch": 3.06, + "learning_rate": 4.684684684684685e-05, + "loss": 0.0753, + "step": 340 + }, + { + "epoch": 3.15, + "learning_rate": 4.234234234234234e-05, + "loss": 0.0231, + "step": 350 + }, + { + "epoch": 3.24, + "learning_rate": 3.783783783783784e-05, + "loss": 0.0266, + "step": 360 + }, + { + "epoch": 3.33, + "learning_rate": 3.3333333333333335e-05, + "loss": 0.023, + "step": 370 + }, + { + "epoch": 3.42, + "learning_rate": 2.882882882882883e-05, + "loss": 0.0224, + "step": 380 + }, + { + "epoch": 3.51, + "learning_rate": 2.4324324324324327e-05, + "loss": 0.0251, + "step": 390 + }, + { + "epoch": 3.6, + "learning_rate": 1.981981981981982e-05, + "loss": 0.025, + "step": 400 + }, + { + "epoch": 3.6, + "eval_accuracy": 0.9573170731707317, + "eval_loss": 0.172527477145195, + "eval_runtime": 4.0556, + "eval_samples_per_second": 80.877, + "eval_steps_per_second": 10.11, + "step": 400 + }, + { + "epoch": 3.69, + "learning_rate": 1.5315315315315316e-05, + "loss": 0.0544, + "step": 410 + }, + { + "epoch": 3.78, + "learning_rate": 1.0810810810810812e-05, + "loss": 0.0472, + "step": 420 + }, + { + "epoch": 3.87, + "learning_rate": 6.306306306306306e-06, + "loss": 0.0478, + "step": 430 + }, + { + "epoch": 3.96, + "learning_rate": 1.801801801801802e-06, + "loss": 0.0209, + "step": 440 + }, + { + "epoch": 4.0, + "step": 444, + "total_flos": 5.480429975941939e+17, + "train_loss": 0.2510145430092339, + "train_runtime": 644.0352, + "train_samples_per_second": 10.981, + "train_steps_per_second": 0.689 + } + ], + "max_steps": 444, + "num_train_epochs": 4, + "total_flos": 5.480429975941939e+17, + "trial_name": null, + "trial_params": null +} diff --git a/models/vit-base-garbage/training_args.bin b/models/vit-base-garbage/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..bcf44ad847b818b017359c02a0e900aa60a1322d --- /dev/null +++ b/models/vit-base-garbage/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b388b03a0bc352599c3f1a7b29932b4fb98d8f608da2c3b16a52bf3c7baf4496 +size 3375 diff --git a/paper567.jpg b/paper567.jpg new file mode 100644 index 0000000000000000000000000000000000000000..8490f1682db3a27bcd4bee0c91e520683f99bfa2 Binary files /dev/null and b/paper567.jpg differ diff --git a/plastic202.jpg b/plastic202.jpg new file mode 100644 index 0000000000000000000000000000000000000000..57965285fc6a71ee5e3b46b01a338d84ee2413d1 Binary files /dev/null and b/plastic202.jpg differ diff --git a/requirements.txt b/requirements.txt new file mode 100644 index 0000000000000000000000000000000000000000..747b7aa97abc76d0e81e308ad7e8d51a6f91c76f --- /dev/null +++ b/requirements.txt @@ -0,0 +1 @@ +transformers \ No newline at end of file diff --git a/trash105.jpg b/trash105.jpg new file mode 100644 index 0000000000000000000000000000000000000000..03b8072cbbaa2ec42fed5630cc522a868aa1fee4 Binary files /dev/null and b/trash105.jpg differ