Merge pull request #30 from boostcampaitech6/feat/29-add-recbole-gene…

…ral-model [FEAT] Add model in Recbole #29
boostcampaitech6 · Feb 21, 2024 · 13be27e · 13be27e
2 parents da9ab28 + 8738f23
commit 13be27e
Show file tree

Hide file tree

Showing 9 changed files with 188,457 additions and 188,371 deletions.
diff --git a/Recbole/Data_convert.ipynb b/Recbole/Data_convert.ipynb
@@ -53,11 +53,12 @@
     "os.makedirs(\"./dataset\", exist_ok=True)\n",
     "#inter file\n",
     "inter_file = train.copy()\n",
+    "inter_file['rating'] = 1.0\n",
     "inter_file = inter_file.rename(\n",
     "    columns={\n",
     "        \"user\":\"user_id:token\",\n",
     "        \"item\":\"item_id:token\",\n",
-    "        #\"rating\":\"rating:float\",\n",
+    "        \"rating\":\"rating:float\",\n",
     "        \"time\":\"timestamp:float\"\n",
     "    }\n",
     ")\n",
@@ -141,7 +142,7 @@
     "    inplace=True,\n",
     ")\n",
     "\n",
-    "item_file.to_csv(\"./dataset/ML.item\", sep='\\t', index=False)\n"
+    "item_file.to_csv(\"./dataset/ML/ML.item\", sep='\\t', index=False)\n"
    ]
   },
   {
@@ -175,7 +176,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.9.0"
+   "version": "3.9.18"
   }
  },
  "nbformat": 4,

diff --git a/Recbole/configs/Dataset/ML_sequence.yaml b/Recbole/configs/Dataset/ML_sequence.yaml
@@ -3,9 +3,13 @@ dataset: 'ML'
 USER_ID_FIELD: user_id
 ITEM_ID_FIELD: item_id
 TIME_FIELD: timestamp
-load_col:
-    inter: [user_id, item_id, timestamp]
 ITEM_LIST_LENGTH_FIELD: item_length
 LIST_SUFFIX: _list
 MAX_ITEM_LIST_LENGTH: 50
 repeatable : True
+RATING_FIELD: rating            # (str) Field name of rating feature.
+LABEL_FIELD: rating
+load_col: 
+    inter: ['user_id', 'item_id', 'timestamp', 'rating'] 
+    user: ['user_id']
+    item: ['item_id', 'director', 'class', 'writer', 'release_year']
diff --git a/Recbole/configs/Model/FM.yaml b/Recbole/configs/Model/FM.yaml
@@ -0,0 +1 @@
+embedding_size: 10              # (int) The embedding size of features.
diff --git a/Recbole/configs/Training/train.yaml b/Recbole/configs/Training/train.yaml
@@ -2,13 +2,13 @@
 epochs: 300                     # (int) The number of training epochs.
 train_batch_size: 2048          # (int) The training batch size.
 learner: adam                   # (str) The name of used optimizer.
-learning_rate: 0.001            # (float) Learning rate.
+learning_rate: 0.01            # (float) Learning rate.
 train_neg_sample_args:          # (dict) Negative sampling configuration for model training.
   distribution: uniform         # (str) The distribution of negative items.
-  sample_num: 1                 # (int) The sampled num of negative items.
-  alpha: 1.0                    # (float) The power of sampling probability for popularity distribution.
-  dynamic: False                # (bool) Whether to use dynamic negative sampling.
-  candidate_num: 0              # (int) The number of candidate negative items when dynamic negative sampling.
+  sample_num: 3                # (int) The sampled num of negative items.
+  alpha: 0.5                    # (float) The power of sampling probability for popularity distribution.
+  #dynamic: True                # (bool) Whether to use dynamic negative sampling.
+  #candidate_num: 500              # (int) The number of candidate negative items when dynamic negative sampling.
 eval_step: 1                    # (int) The number of training epochs before an evaluation on the valid dataset.
 stopping_step: 10               # (int) The threshold for validation-based early stopping.
 clip_grad_norm: ~               # (dict) The args of clip_grad_norm_ which will clip gradient norm of model. 

diff --git a/Recbole/configs/config_combination.txt b/Recbole/configs/config_combination.txt
@@ -2,4 +2,32 @@ Model_name / eval.yaml / Model.yaml / .... etc..
 ease ./configs/EVAL/eval_ease.yaml ./configs/Model/EASE.yaml
 recvae ./configs/EVAL/eval_rec_vae.yaml ./configs/Model/RecVae.yaml
 sasrec ./configs/EVAL/eval_sasrec.yaml ./configs/Model/Sasrec.yaml ./configs/Training/SasrecTrain.yaml ./configs/Dataset/ML_sequence.yaml
-lightgcn ./configs/EVAL/eval_lightgcn.yaml ./configs/Model/LightGCN.yaml
+lightgcn ./configs/EVAL/eval_lightgcn.yaml ./configs/Model/LightGCN.yaml ./configs/Training/train.yaml
+fm ./configs/Model/FM.yaml ./configs/Dataset/ML_sequence.yaml ./configs/Training/train.yaml
+pop
+nais
+neumf
+convncf
+dmf
+fism
+spectralcf
+gcmc
+ngcf 
+dgcf
+line
+multidae
+macridvae
+cdae
+enmf
+nncf
+ract
+slimelastic
+sgl
+admmslim
+nceplrec
+simplex
+ncl
+random
+
+diffrec
+ldiffrec
Original file line number	Diff line number	Diff line change
		@@ -0,0 +1 @@
		embedding_size: 10 # (int) The embedding size of features.