Training in progress, step 500

config.json

@@ -0,0 +1,40 @@
+{
+  "architectures": [
+    "T5ForConditionalGeneration"
+  ],
+  "classifier_dropout": 0.0,
+  "d_ff": 2048,
+  "d_kv": 64,
+  "d_model": 512,
+  "decoder_start_token_id": 0,
+  "dense_act_fn": "relu",
+  "dropout_rate": 0.1,
+  "eos_token_id": 1,
+  "feed_forward_proj": "relu",
+  "initializer_factor": 1.0,
+  "is_encoder_decoder": true,
+  "is_gated_act": false,
+  "layer_norm_epsilon": 1e-06,
+  "model_type": "t5",
+  "n_positions": 512,
+  "num_decoder_layers": 6,
+  "num_heads": 8,
+  "num_layers": 6,
+  "output_past": true,
+  "pad_token_id": 0,
+  "relative_attention_max_distance": 128,
+  "relative_attention_num_buckets": 32,
+  "task_specific_params": {
+    "caption": {
+      "early_stopping": true,
+      "length_penalty": 1.0,
+      "max_length": 64,
+      "num_beams": 4,
+      "prefix": "caption: "
+    }
+  },
+  "torch_dtype": "float32",
+  "transformers_version": "4.51.3",
+  "use_cache": true,
+  "vocab_size": 32128
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5eedf6cb6f677aed32193d5fef6ec706f72626b4da9d67be177e473a9abe09a7
+size 242041896

samples/100.txt

@@ -0,0 +1,19 @@
+[RAW] a dog laying under a glass table
+[NOISE] caption: <extra_id_0> under <extra_id_1> <extra_id_2> table
+[OUT] a table the table the a table.
+
+[RAW] a person sitting next to a broken vending machine
+[NOISE] caption: a person sitting <extra_id_0> a <extra_id_1> <extra_id_2>
+[OUT] in chair.
+
+[RAW] a white couch with red stains on the cushions
+[NOISE] caption: a <extra_id_0> couch with <extra_id_1> stains <extra_id_2> the cushions
+[OUT] couch with a stains and on couch with on couch stains and couch with stains and on couch with stains and on on.
+
+[RAW] an old laptop glowing in the dark room
+[NOISE] caption: an old laptop glowing <extra_id_0> <extra_id_1> <extra_id_2>
+[OUT] .
+
+[RAW] a stack of books placed under a hanging bulb
+[NOISE] caption: a stack of <extra_id_0> <extra_id_1> hanging <extra_id_2>
+[OUT] stacks of stacks of stacks of stacks of stacks of stacks of stacks of stacks of stacks.

samples/200.txt

@@ -0,0 +1,19 @@
+[RAW] a dog laying under a glass table
+[NOISE] caption: <extra_id_0> dog laying under a <extra_id_1> <extra_id_2>
+[OUT] a tree.
+
+[RAW] a person sitting next to a broken vending machine
+[NOISE] caption: a <extra_id_0> sitting <extra_id_1> <extra_id_2> broken vending machine
+[OUT] woman on a a woman sitting on a broken vending machine woman sitting on a woman on a woman woman sitting on a broken vending machine machine on a broken vending machine.
+
+[RAW] a white couch with red stains on the cushions
+[NOISE] caption: a white couch <extra_id_0> <extra_id_1> <extra_id_2>
+[OUT] .
+
+[RAW] an old laptop glowing in the dark room
+[NOISE] caption: an old laptop glowing <extra_id_0> <extra_id_1> <extra_id_2>
+[OUT] .
+
+[RAW] a stack of books placed under a hanging bulb
+[NOISE] caption: a stack of books <extra_id_0> under a <extra_id_1> <extra_id_2>
+[OUT] stacked shelf.

samples/300.txt

@@ -0,0 +1,19 @@
+[RAW] a dog laying under a glass table
+[NOISE] caption: a dog <extra_id_0> <extra_id_1> a <extra_id_2>
+[OUT] , or dog.
+
+[RAW] a person sitting next to a broken vending machine
+[NOISE] caption: <extra_id_0> person <extra_id_1> a <extra_id_2>
+[OUT] a who is person.
+
+[RAW] a white couch with red stains on the cushions
+[NOISE] caption: a <extra_id_0> with red <extra_id_1> <extra_id_2>
+[OUT] photo.
+
+[RAW] an old laptop glowing in the dark room
+[NOISE] caption: <extra_id_0> old <extra_id_1> <extra_id_2> dark room
+[OUT] The room in a room.
+
+[RAW] a stack of books placed under a hanging bulb
+[NOISE] caption: a <extra_id_0> of <extra_id_1> placed <extra_id_2> hanging bulb
+[OUT] piece paper in a piece in a piece paper piece in a piece piece in a hanging bulb.

samples/400.txt

@@ -0,0 +1,19 @@
+[RAW] a dog laying under a glass table
+[NOISE] caption: a <extra_id_0> a <extra_id_1>
+[OUT] . a..
+
+[RAW] a person sitting next to a broken vending machine
+[NOISE] caption: a <extra_id_0> <extra_id_1> to a broken vending <extra_id_2>
+[OUT] broken vending machine link machine.
+
+[RAW] a white couch with red stains on the cushions
+[NOISE] caption: <extra_id_0> <extra_id_1> couch <extra_id_2> red stains on the cushions
+[OUT] The The has red and The has red and The is a The has The has red and stains on the couch.
+
+[RAW] an old laptop glowing in the dark room
+[NOISE] caption: an old laptop <extra_id_0> <extra_id_1> dark <extra_id_2>
+[OUT] , in.
+
+[RAW] a stack of books placed under a hanging bulb
+[NOISE] caption: a stack of <extra_id_0> placed under a <extra_id_1> <extra_id_2>
+[OUT] paper stack of paper.

samples/500.txt

@@ -0,0 +1,19 @@
+[RAW] a dog laying under a glass table
+[NOISE] caption: <extra_id_0> <extra_id_1> <extra_id_2> table
+[OUT] a table a table.
+
+[RAW] a person sitting next to a broken vending machine
+[NOISE] caption: <extra_id_0> sitting <extra_id_1> to <extra_id_2> broken vending machine
+[OUT] I’m next a on a broken vending machine a broken vending machine.
+
+[RAW] a white couch with red stains on the cushions
+[NOISE] caption: a white couch <extra_id_0> <extra_id_1> <extra_id_2> cushions
+[OUT] with cushions soft.
+
+[RAW] an old laptop glowing in the dark room
+[NOISE] caption: an old laptop <extra_id_0> <extra_id_1> dark <extra_id_2>
+[OUT] , in.
+
+[RAW] a stack of books placed under a hanging bulb
+[NOISE] caption: <extra_id_0> <extra_id_1> under a hanging <extra_id_2>
+[OUT] a hanging wall.

samples/600.txt

@@ -0,0 +1,19 @@
+[RAW]     a dog laying under a glass table
+[CORRUPT] caption: <extra_id_0> layinge <extra_id_1> <extra_id_2>
+[OUTPUT]  a one man on a striped rug
+
+[RAW]     a person sitting next to a broken vending machine
+[CORRUPT] caption: a <extra_id_0> <extra_id_1> next to a <extra_id_2>
+[OUTPUT]  mrs. seated striped rug
+
+[RAW]     a white couch with red stains on the cushions
+[CORRUPT] caption: a whzite <extra_id_0> <extra_id_1> stains n he <extra_id_2>
+[OUTPUT]  and a woman are seated
+
+[RAW]     an old laptop glowing in the dark room
+[CORRUPT] caption: <extra_id_0> laptop <extra_id_1> dark <extra_id_2>
+[OUTPUT]  a mrs. sitting on a under light
+
+[RAW]     a stack of books placed under a hanging bulb
+[CORRUPT] caption: ar stack <extra_id_0> books placed <extra_id_1> a <extra_id_2> bulb
+[OUTPUT]  and two on striped

samples/700.txt

@@ -0,0 +1,19 @@
+[RAW]     a dog laying under a glass table
+[CORRUPT] caption: a dogz <extra_id_0> <extra_id_1> <extra_id_2>
+[OUTPUT]  and a woman are seated on a striped rug
+
+[RAW]     a person sitting next to a broken vending machine
+[CORRUPT] caption: a person <extra_id_0> <extra_id_1> brioken vending <extra_id_2>
+[OUTPUT]  and a woman are seated on a shoes
+
+[RAW]     a white couch with red stains on the cushions
+[CORRUPT] caption: a <extra_id_0> <extra_id_1> with red <extra_id_2> cushions
+[OUTPUT]  man and a woman are seated striped
+
+[RAW]     an old laptop glowing in the dark room
+[CORRUPT] caption: an ovld <extra_id_0> <extra_id_1>
+[OUTPUT]  and a woman are adjusting their beret
+
+[RAW]     a stack of books placed under a hanging bulb
+[CORRUPT] caption: a stack <extra_id_0> books <extra_id_1> <extra_id_2> hangijng bulb
+[OUTPUT]  and two are seated on a

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4ff2b074f12d595dc1428f7e0d171e3f63fd95b34d5704d181b12fe2841260e1
+size 5368