Add v8 LoRA adapter + evaluation plots

adapter/adapter_config.json

@@ -10,18 +10,20 @@
   "layers_pattern": null,
   "layers_to_transform": null,
   "loftq_config": {},
-  "lora_alpha": 16,
+  "lora_alpha": 32,
   "lora_dropout": 0.05,
   "megatron_config": null,
   "megatron_core": "megatron.core",
   "modules_to_save": null,
   "peft_type": "LORA",
-  "r": 8,
+  "r": 16,
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
-    "v_proj",
-    "q_proj"
+    "q_proj",
+    "k_proj",
+    "o_proj",
+    "v_proj"
   ],
   "task_type": "CAUSAL_LM",
   "use_dora": false,

adapter/adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:adc0418e4e2f3e610f8389bdc2f89c65039ecc60301144df3809f365f3915006
-size 16794200
+oid sha256:6dbe3323b64f9793bd42f6b4ea712a45a75cc6c1123fefd854d9a9f27f3ffba6
+size 67143296

adapter/eval_plots/eval_log.json

@@ -1,50 +1,42 @@
 [
   {
     "step": 100,
-    "eval_loss": 1.8240644931793213
+    "eval_loss": 0.48696422576904297
   },
   {
     "step": 200,
-    "eval_loss": 1.0799707174301147
+    "eval_loss": 0.34586554765701294
   },
   {
     "step": 300,
-    "eval_loss": 0.9546574950218201
+    "eval_loss": 0.28842663764953613
   },
   {
     "step": 400,
-    "eval_loss": 0.8702907562255859
+    "eval_loss": 0.2540067136287689
   },
   {
     "step": 500,
-    "eval_loss": 0.787815272808075
+    "eval_loss": 0.2295776903629303
   },
   {
     "step": 600,
-    "eval_loss": 0.6270484328269958
+    "eval_loss": 0.21396693587303162
   },
   {
     "step": 700,
-    "eval_loss": 0.5892859101295471
+    "eval_loss": 0.19977937638759613
   },
   {
     "step": 800,
-    "eval_loss": 0.5660054087638855
+    "eval_loss": 0.19195251166820526
   },
   {
     "step": 900,
-    "eval_loss": 0.5496686697006226
+    "eval_loss": 0.18890142440795898
   },
   {
     "step": 1000,
-    "eval_loss": 0.5432103872299194
-  },
-  {
-    "step": 1100,
-    "eval_loss": 0.5394924879074097
-  },
-  {
-    "step": 1200,
-    "eval_loss": 0.5386649966239929
+    "eval_loss": 0.18816642463207245
   }
 ]

adapter/eval_plots/eval_results.json

@@ -1,56 +1,56 @@
 {
   "summary": {
-    "keyword_accuracy": 0.7058823529411765,
+    "keyword_accuracy": 0.7647058823529411,
     "format_accuracy": 0.0,
     "precision": 1.0,
-    "recall": 0.7058823529411765,
-    "f1": 0.8275862068965517,
-    "avg_rouge1": 0.25364705882352945,
-    "avg_rouge2": 0.06123529411764707,
-    "avg_rougeL": 0.25364705882352945,
-    "avg_response_time": 16.716294117647056
+    "recall": 0.7647058823529411,
+    "f1": 0.8666666666666667,
+    "avg_rouge1": 0.03211764705882353,
+    "avg_rouge2": 0.006058823529411765,
+    "avg_rougeL": 0.030941176470588236,
+    "avg_response_time": 16.512235294117644
   },
   "per_category": {
     "DS Logic": {
       "accuracy": 1.0,
       "format": 0.0,
-      "rougeL": 0.0,
-      "avg_time": 16.393,
+      "rougeL": 0.022,
+      "avg_time": 15.978,
       "n": 2
     },
     "LangChain": {
-      "accuracy": 0.0,
+      "accuracy": 0.25,
       "format": 0.0,
-      "rougeL": 0.0,
-      "avg_time": 16.93,
+      "rougeL": 0.022,
+      "avg_time": 17.225,
       "n": 4
     },
     "ML": {
       "accuracy": 1.0,
       "format": 0.0,
-      "rougeL": 0.43,
-      "avg_time": 16.838,
+      "rougeL": 0.031,
+      "avg_time": 22.483,
       "n": 3
     },
     "Pandas": {
       "accuracy": 1.0,
       "format": 0.0,
-      "rougeL": 0.347,
-      "avg_time": 17.03,
+      "rougeL": 0.031,
+      "avg_time": 17.45,
       "n": 4
     },
     "R": {
       "accuracy": 0.5,
       "format": 0.0,
-      "rougeL": 0.486,
-      "avg_time": 15.999,
+      "rougeL": 0.039,
+      "avg_time": 12.786,
       "n": 2
     },
     "Viz": {
       "accuracy": 1.0,
       "format": 0.0,
-      "rougeL": 0.333,
-      "avg_time": 16.522,
+      "rougeL": 0.05,
+      "avg_time": 8.516,
       "n": 2
     }
   },
@@ -59,74 +59,74 @@
       "category": "LangChain",
       "prompt": "What function limits the number of messages passed to a model by counting tokens in Langchain? in Python ?",
       "keyword": "trim_messages",
-      "library": "import nltk",
-      "code": "nltk.LangChain(5)  \n.add_rule('[A-Za-z]+', '[A-Za-z]+')  \n.add_rule('[0-9]+', '[0-9]+')  \n.add_rule('[A-Za-z0-9]+', '[A-Za-z0-9]+')  \n.add_rule('[A-Za-z0-9\\.]+', '[A-Za-z0-9\\.]+')  \n.add_rule('[A-Za-z",
+      "library": "",
+      "code": "",
       "math": "",
       "rouge1": 0.0,
       "rouge2": 0.0,
       "rougeL": 0.0,
       "keyword_match": 0,
-      "response_time": 16.873,
-      "output_tokens": 200,
+      "response_time": 20.159,
+      "output_tokens": 190,
       "format_correct": 0
     },
     {
       "category": "LangChain",
       "prompt": "What argument enables jq schema parsing for content_key in JSONLoader in Langchain? in Python ?",
       "keyword": "is_content_key_jq_parsable",
-      "library": "import pandas as pd",
-      "code": "df = pd.read_json('file.json', orient='records')",
+      "library": "",
+      "code": "",
       "math": "",
-      "rouge1": 0.0,
-      "rouge2": 0.0,
-      "rougeL": 0.0,
+      "rouge1": 0.081,
+      "rouge2": 0.021,
+      "rougeL": 0.061,
       "keyword_match": 0,
-      "response_time": 17.915,
-      "output_tokens": 200,
+      "response_time": 17.154,
+      "output_tokens": 137,
       "format_correct": 0
     },
     {
       "category": "LangChain",
       "prompt": "What callback handler tracks token usage in Langchain? in Python ?",
       "keyword": "UsageMetadataCallbackHandler",
-      "library": "TensorFlow",
-      "code": "tf.keras.callbacks.Callback()",
+      "library": "",
+      "code": "",
       "math": "",
       "rouge1": 0.0,
       "rouge2": 0.0,
       "rougeL": 0.0,
       "keyword_match": 0,
-      "response_time": 16.22,
-      "output_tokens": 200,
+      "response_time": 19.01,
+      "output_tokens": 183,
       "format_correct": 0
     },
     {
       "category": "LangChain",
       "prompt": "What taxonomy is used on GitHub Issues for sorting and discovery in Langchain? in Python ?",
       "keyword": "Labels",
-      "library": "import pandas as pd",
-      "code": "df.sort_values('col', ascending=True).head()",
+      "library": "",
+      "code": "",
       "math": "",
-      "rouge1": 0.0,
+      "rouge1": 0.028,
       "rouge2": 0.0,
-      "rougeL": 0.0,
-      "keyword_match": 0,
-      "response_time": 16.711,
-      "output_tokens": 200,
+      "rougeL": 0.028,
+      "keyword_match": 1,
+      "response_time": 12.577,
+      "output_tokens": 123,
       "format_correct": 0
     },
     {
       "category": "Pandas",
       "prompt": "How to drop missing values in a pandas DataFrame? in Python ?",
       "keyword": "dropna",
-      "library": "pd",
-      "code": "df.dropna(inplace=True)  \n.",
+      "library": "",
+      "code": "",
       "math": "",
-      "rouge1": 0.4,
+      "rouge1": 0.021,
       "rouge2": 0.0,
-      "rougeL": 0.4,
+      "rougeL": 0.021,
       "keyword_match": 1,
-      "response_time": 18.061,
+      "response_time": 20.368,
       "output_tokens": 200,
       "format_correct": 0
     },
@@ -134,44 +134,44 @@
       "category": "Pandas",
       "prompt": "How to read a CSV file using pandas? in Python ?",
       "keyword": "read_csv",
-      "library": "import pandas as pd",
-      "code": "df = pd.read_csv('file.csv')  \n.",
+      "library": "",
+      "code": "",
       "math": "",
-      "rouge1": 0.5,
-      "rouge2": 0.333,
-      "rougeL": 0.5,
+      "rouge1": 0.061,
+      "rouge2": 0.031,
+      "rougeL": 0.061,
       "keyword_match": 1,
-      "response_time": 16.448,
-      "output_tokens": 200,
+      "response_time": 10.138,
+      "output_tokens": 98,
       "format_correct": 0
     },
     {
       "category": "Pandas",
       "prompt": "How to rename a column in pandas? in Python ?",
       "keyword": "rename",
-      "library": "import pandas as pd",
-      "code": "df.rename(columns={'old_col': 'new_col'}, inplace=True)  \n.",
+      "library": "",
+      "code": "",
       "math": "",
-      "rouge1": 0.2,
+      "rouge1": 0.023,
       "rouge2": 0.0,
-      "rougeL": 0.2,
+      "rougeL": 0.023,
       "keyword_match": 1,
-      "response_time": 17.085,
-      "output_tokens": 200,
+      "response_time": 19.195,
+      "output_tokens": 182,
       "format_correct": 0
     },
     {
       "category": "Pandas",
       "prompt": "How to group rows and compute aggregate in pandas? in Python ?",
       "keyword": "groupby",
-      "library": "import pandas as pd",
-      "code": "df.groupby('category').agg({'value': 'sum'})  \n.",
+      "library": "",
+      "code": "",
       "math": "",
-      "rouge1": 0.286,
+      "rouge1": 0.019,
       "rouge2": 0.0,
-      "rougeL": 0.286,
+      "rougeL": 0.019,
       "keyword_match": 1,
-      "response_time": 16.524,
+      "response_time": 20.097,
       "output_tokens": 200,
       "format_correct": 0
     },
@@ -179,14 +179,14 @@
       "category": "ML",
       "prompt": "How to train a decision tree classifier in sklearn? in Python ?",
       "keyword": "DecisionTreeClassifier",
-      "library": "from sklearn.tree import DecisionTreeClassifier",
-      "code": "clf = DecisionTreeClassifier()",
+      "library": "",
+      "code": "",
       "math": "",
-      "rouge1": 0.667,
+      "rouge1": 0.019,
       "rouge2": 0.0,
-      "rougeL": 0.667,
+      "rougeL": 0.019,
       "keyword_match": 1,
-      "response_time": 16.391,
+      "response_time": 20.672,
       "output_tokens": 200,
       "format_correct": 0
     },
@@ -194,14 +194,14 @@
       "category": "ML",
       "prompt": "How to split dataset into train and test in sklearn? in Python ?",
       "keyword": "train_test_split",
-      "library": "from sklearn.model_selection import train_test_split",
-      "code": "X_train, X_test = train_test_split(df, test_size=0.2)",
+      "library": "",
+      "code": "",
       "math": "",
-      "rouge1": 0.4,
-      "rouge2": 0.308,
-      "rougeL": 0.4,
+      "rouge1": 0.048,
+      "rouge2": 0.033,
+      "rougeL": 0.048,
       "keyword_match": 1,
-      "response_time": 17.126,
+      "response_time": 20.794,
       "output_tokens": 200,
       "format_correct": 0
     },
@@ -209,45 +209,45 @@
       "category": "ML",
       "prompt": "How to normalize features using StandardScaler in sklearn? in Python ?",
       "keyword": "StandardScaler",
-      "library": "from sklearn.preprocessing import StandardScaler",
-      "code": "scaler = StandardScaler(); X_scaled = scaler.fit_transform(X)",
+      "library": "",
+      "code": "",
       "math": "",
-      "rouge1": 0.222,
+      "rouge1": 0.025,
       "rouge2": 0.0,
-      "rougeL": 0.222,
+      "rougeL": 0.025,
       "keyword_match": 1,
-      "response_time": 16.996,
-      "output_tokens": 200,
+      "response_time": 25.984,
+      "output_tokens": 136,
       "format_correct": 0
     },
     {
       "category": "Viz",
       "prompt": "How to plot a heatmap using seaborn? in Python ?",
       "keyword": "heatmap",
-      "library": "import seaborn as sns",
-      "code": "sns.heatmap(df, annot=True)",
+      "library": "",
+      "code": "",
       "math": "",
-      "rouge1": 0.333,
+      "rouge1": 0.043,
       "rouge2": 0.0,
-      "rougeL": 0.333,
+      "rougeL": 0.043,
       "keyword_match": 1,
-      "response_time": 16.279,
-      "output_tokens": 200,
+      "response_time": 9.108,
+      "output_tokens": 87,
       "format_correct": 0
     },
     {
       "category": "Viz",
       "prompt": "How to create a histogram using matplotlib? in Python ?",
       "keyword": "hist",
-      "library": "import matplotlib.pyplot as plt",
-      "code": "plt.hist(data, bins=5)",
+      "library": "",
+      "code": "",
       "math": "",
-      "rouge1": 0.333,
+      "rouge1": 0.057,
       "rouge2": 0.0,
-      "rougeL": 0.333,
+      "rougeL": 0.057,
       "keyword_match": 1,
-      "response_time": 16.764,
-      "output_tokens": 200,
+      "response_time": 7.924,
+      "output_tokens": 77,
       "format_correct": 0
     },
     {
@@ -255,59 +255,59 @@
       "prompt": "How to get the maximum value of a vector in R ?",
       "keyword": "max(",
       "library": "",
-      "code": "max_value <- max(vector)",
+      "code": "",
       "math": "",
-      "rouge1": 0.4,
+      "rouge1": 0.043,
       "rouge2": 0.0,
-      "rougeL": 0.4,
+      "rougeL": 0.043,
       "keyword_match": 1,
-      "response_time": 15.953,
-      "output_tokens": 200,
+      "response_time": 8.151,
+      "output_tokens": 75,
       "format_correct": 0
     },
     {
       "category": "R",
       "prompt": "How to read a CSV file in R ?",
       "keyword": "read.csv",
-      "library": "readr",
-      "code": "df <- read_csv(\"file.csv\")",
+      "library": "",
+      "code": "",
       "math": "",
-      "rouge1": 0.571,
-      "rouge2": 0.4,
-      "rougeL": 0.571,
+      "rouge1": 0.035,
+      "rouge2": 0.018,
+      "rougeL": 0.035,
       "keyword_match": 0,
-      "response_time": 16.045,
-      "output_tokens": 200,
+      "response_time": 17.421,
+      "output_tokens": 172,
       "format_correct": 0
     },
     {
       "category": "DS Logic",
       "prompt": "What does overfitting mean in machine learning? in Python ?",
       "keyword": "overfit",
-      "library": "sklearn.linear_model",
-      "code": "X = df[col_num].values; y = df['target'].values; lr = LogisticRegression(penalty='l2', C=10); lr.fit(X, y)",
+      "library": "",
+      "code": "",
       "math": "",
-      "rouge1": 0.0,
+      "rouge1": 0.026,
       "rouge2": 0.0,
-      "rougeL": 0.0,
+      "rougeL": 0.026,
       "keyword_match": 1,
-      "response_time": 16.685,
-      "output_tokens": 200,
+      "response_time": 11.186,
+      "output_tokens": 106,
       "format_correct": 0
     },
     {
       "category": "DS Logic",
       "prompt": "What is gradient descent in machine learning? in Python ?",
       "keyword": "gradient",
-      "library": "from sklearn.linear_model import LinearRegression",
-      "code": "reg = LinearRegression(); reg.fit(X, y)",
+      "library": "",
+      "code": "",
       "math": "",
-      "rouge1": 0.0,
+      "rouge1": 0.017,
       "rouge2": 0.0,
-      "rougeL": 0.0,
+      "rougeL": 0.017,
       "keyword_match": 1,
-      "response_time": 16.101,
-      "output_tokens": 200,
+      "response_time": 20.77,
+      "output_tokens": 196,
       "format_correct": 0
     }
   ]

adapter/eval_plots/loss_log.json

@@ -1,482 +1,402 @@
 [
   {
     "step": 10,
-    "loss": 3.7145
+    "loss": 2.4541
   },
   {
     "step": 20,
-    "loss": 3.8108
+    "loss": 2.1867
   },
   {
     "step": 30,
-    "loss": 3.8272
+    "loss": 1.4349
   },
   {
     "step": 40,
-    "loss": 3.6441
+    "loss": 0.9306
   },
   {
     "step": 50,
-    "loss": 3.6304
+    "loss": 0.7783
   },
   {
     "step": 60,
-    "loss": 3.4205
+    "loss": 0.6622
   },
   {
     "step": 70,
-    "loss": 3.0848
+    "loss": 0.5654
   },
   {
     "step": 80,
-    "loss": 2.7674
+    "loss": 0.5339
   },
   {
     "step": 90,
-    "loss": 2.6102
+    "loss": 0.467
   },
   {
     "step": 100,
-    "loss": 2.1779
+    "loss": 0.4574
   },
   {
     "step": 110,
-    "loss": 1.6865
+    "loss": 0.4523
   },
   {
     "step": 120,
-    "loss": 1.4528
+    "loss": 0.4217
   },
   {
     "step": 130,
-    "loss": 1.3545
+    "loss": 0.4343
   },
   {
     "step": 140,
-    "loss": 1.2961
+    "loss": 0.4261
   },
   {
     "step": 150,
-    "loss": 1.2643
+    "loss": 0.4277
   },
   {
     "step": 160,
-    "loss": 1.1677
+    "loss": 0.3847
   },
   {
     "step": 170,
-    "loss": 1.1044
+    "loss": 0.344
   },
   {
     "step": 180,
-    "loss": 1.1668
+    "loss": 0.3867
   },
   {
     "step": 190,
-    "loss": 1.098
+    "loss": 0.3597
   },
   {
     "step": 200,
-    "loss": 1.0887
+    "loss": 0.3223
   },
   {
     "step": 210,
-    "loss": 1.0874
+    "loss": 0.323
   },
   {
     "step": 220,
-    "loss": 1.1033
+    "loss": 0.3337
   },
   {
     "step": 230,
-    "loss": 1.0045
+    "loss": 0.2679
   },
   {
     "step": 240,
-    "loss": 1.0166
+    "loss": 0.3112
   },
   {
     "step": 250,
-    "loss": 1.0397
+    "loss": 0.3013
   },
   {
     "step": 260,
-    "loss": 1.02
+    "loss": 0.2929
   },
   {
     "step": 270,
-    "loss": 0.9788
+    "loss": 0.27
   },
   {
     "step": 280,
-    "loss": 1.0221
+    "loss": 0.3208
   },
   {
     "step": 290,
-    "loss": 0.9507
+    "loss": 0.2797
   },
   {
     "step": 300,
-    "loss": 0.9196
+    "loss": 0.2468
   },
   {
     "step": 310,
-    "loss": 0.9043
+    "loss": 0.27
   },
   {
     "step": 320,
-    "loss": 0.9055
+    "loss": 0.2513
   },
   {
     "step": 330,
-    "loss": 0.9389
+    "loss": 0.254
   },
   {
     "step": 340,
-    "loss": 0.9211
+    "loss": 0.2899
   },
   {
     "step": 350,
-    "loss": 0.9335
+    "loss": 0.2807
   },
   {
     "step": 360,
-    "loss": 0.9127
+    "loss": 0.2676
   },
   {
     "step": 370,
-    "loss": 0.8907
+    "loss": 0.2495
   },
   {
     "step": 380,
-    "loss": 0.8504
+    "loss": 0.2172
   },
   {
     "step": 390,
-    "loss": 0.9536
+    "loss": 0.2786
   },
   {
     "step": 400,
-    "loss": 0.8082
+    "loss": 0.2147
   },
   {
     "step": 410,
-    "loss": 0.8522
+    "loss": 0.235
   },
   {
     "step": 420,
-    "loss": 0.8181
+    "loss": 0.2239
   },
   {
     "step": 430,
-    "loss": 0.7697
+    "loss": 0.1885
   },
   {
     "step": 440,
-    "loss": 0.8197
+    "loss": 0.2431
   },
   {
     "step": 450,
-    "loss": 0.8189
+    "loss": 0.2412
   },
   {
     "step": 460,
-    "loss": 0.7979
+    "loss": 0.2341
   },
   {
     "step": 470,
-    "loss": 0.8282
+    "loss": 0.23
   },
   {
     "step": 480,
-    "loss": 0.8882
+    "loss": 0.2684
   },
   {
     "step": 490,
-    "loss": 0.8076
+    "loss": 0.197
   },
   {
     "step": 500,
-    "loss": 0.799
+    "loss": 0.2476
   },
   {
     "step": 510,
-    "loss": 0.6938
+    "loss": 0.1719
   },
   {
     "step": 520,
-    "loss": 0.7809
+    "loss": 0.2302
   },
   {
     "step": 530,
-    "loss": 0.7753
+    "loss": 0.2177
   },
   {
     "step": 540,
-    "loss": 0.7249
+    "loss": 0.2076
   },
   {
     "step": 550,
-    "loss": 0.7071
+    "loss": 0.2235
   },
   {
     "step": 560,
-    "loss": 0.6511
+    "loss": 0.2208
   },
   {
     "step": 570,
-    "loss": 0.6253
+    "loss": 0.1932
   },
   {
     "step": 580,
-    "loss": 0.6207
+    "loss": 0.1919
   },
   {
     "step": 590,
-    "loss": 0.6019
+    "loss": 0.1824
   },
   {
     "step": 600,
-    "loss": 0.5924
+    "loss": 0.2221
   },
   {
     "step": 610,
-    "loss": 0.5868
+    "loss": 0.2139
   },
   {
     "step": 620,
-    "loss": 0.5878
+    "loss": 0.1968
   },
   {
     "step": 630,
-    "loss": 0.5844
+    "loss": 0.1999
   },
   {
     "step": 640,
-    "loss": 0.592
+    "loss": 0.1835
   },
   {
     "step": 650,
-    "loss": 0.5673
+    "loss": 0.1981
   },
   {
     "step": 660,
-    "loss": 0.5592
+    "loss": 0.2057
   },
   {
     "step": 670,
-    "loss": 0.6151
+    "loss": 0.2158
   },
   {
     "step": 680,
-    "loss": 0.5323
+    "loss": 0.1915
   },
   {
     "step": 690,
-    "loss": 0.5866
+    "loss": 0.2286
   },
   {
     "step": 700,
-    "loss": 0.5334
+    "loss": 0.1697
   },
   {
     "step": 710,
-    "loss": 0.5482
+    "loss": 0.1912
   },
   {
     "step": 720,
-    "loss": 0.5439
+    "loss": 0.1923
   },
   {
     "step": 730,
-    "loss": 0.5287
+    "loss": 0.1601
   },
   {
     "step": 740,
-    "loss": 0.6061
+    "loss": 0.215
   },
   {
     "step": 750,
-    "loss": 0.5285
+    "loss": 0.1742
   },
   {
     "step": 760,
-    "loss": 0.5141
+    "loss": 0.1408
   },
   {
     "step": 770,
-    "loss": 0.4986
+    "loss": 0.1484
   },
   {
     "step": 780,
-    "loss": 0.5414
+    "loss": 0.1607
   },
   {
     "step": 790,
-    "loss": 0.5625
+    "loss": 0.1571
   },
   {
     "step": 800,
-    "loss": 0.5384
+    "loss": 0.1546
   },
   {
     "step": 810,
-    "loss": 0.5576
+    "loss": 0.1793
   },
   {
     "step": 820,
-    "loss": 0.5088
+    "loss": 0.1523
   },
   {
     "step": 830,
-    "loss": 0.5248
+    "loss": 0.1682
   },
   {
     "step": 840,
-    "loss": 0.5242
+    "loss": 0.1696
   },
   {
     "step": 850,
-    "loss": 0.4592
+    "loss": 0.1418
   },
   {
     "step": 860,
-    "loss": 0.5213
+    "loss": 0.1621
   },
   {
     "step": 870,
-    "loss": 0.4535
+    "loss": 0.1361
   },
   {
     "step": 880,
-    "loss": 0.5557
+    "loss": 0.1692
   },
   {
     "step": 890,
-    "loss": 0.5018
+    "loss": 0.1589
   },
   {
     "step": 900,
-    "loss": 0.5108
+    "loss": 0.1663
   },
   {
     "step": 910,
-    "loss": 0.4915
+    "loss": 0.1743
   },
   {
     "step": 920,
-    "loss": 0.4874
+    "loss": 0.1606
   },
   {
     "step": 930,
-    "loss": 0.5796
+    "loss": 0.1948
   },
   {
     "step": 940,
-    "loss": 0.4931
+    "loss": 0.1563
   },
   {
     "step": 950,
-    "loss": 0.4407
+    "loss": 0.1554
   },
   {
     "step": 960,
-    "loss": 0.4776
+    "loss": 0.1452
   },
   {
     "step": 970,
-    "loss": 0.4795
+    "loss": 0.1534
   },
   {
     "step": 980,
-    "loss": 0.4901
+    "loss": 0.1431
   },
   {
     "step": 990,
-    "loss": 0.5025
+    "loss": 0.154
   },
   {
     "step": 1000,
-    "loss": 0.4933
-  },
-  {
-    "step": 1010,
-    "loss": 0.5486
-  },
-  {
-    "step": 1020,
-    "loss": 0.52
-  },
-  {
-    "step": 1030,
-    "loss": 0.4893
-  },
-  {
-    "step": 1040,
-    "loss": 0.4921
-  },
-  {
-    "step": 1050,
-    "loss": 0.4784
-  },
-  {
-    "step": 1060,
-    "loss": 0.4854
-  },
-  {
-    "step": 1070,
-    "loss": 0.5374
-  },
-  {
-    "step": 1080,
-    "loss": 0.4991
-  },
-  {
-    "step": 1090,
-    "loss": 0.5092
-  },
-  {
-    "step": 1100,
-    "loss": 0.5146
-  },
-  {
-    "step": 1110,
-    "loss": 0.5591
-  },
-  {
-    "step": 1120,
-    "loss": 0.521
-  },
-  {
-    "step": 1130,
-    "loss": 0.4831
-  },
-  {
-    "step": 1140,
-    "loss": 0.4785
-  },
-  {
-    "step": 1150,
-    "loss": 0.5337
-  },
-  {
-    "step": 1160,
-    "loss": 0.4591
-  },
-  {
-    "step": 1170,
-    "loss": 0.5437
-  },
-  {
-    "step": 1180,
-    "loss": 0.5602
-  },
-  {
-    "step": 1190,
-    "loss": 0.4957
-  },
-  {
-    "step": 1200,
-    "loss": 0.5573
+    "loss": 0.1428
   }
 ]