final fixes

.github/workflows/slow-tests.yml

@@ -32,12 +32,6 @@ jobs:
           source activate trl
           pip install -e ".[test]" --no-deps
           pip install pytest-reportlog parameterized
-      
-      - name: Run common tests on single GPU
-        if: always()
-        run: |
-          source activate trl
-          make tests_common_gpu
 
       - name: Run slow SFT tests on single GPU
         if: always()
@@ -80,12 +74,6 @@ jobs:
           source activate trl
           pip install -e ".[test]" --no-deps
           pip install pytest-reportlog parameterized
-      
-      - name: Run common tests on single GPU
-        if: always()  
-        run: |
-          source activate trl
-          make tests_common_gpu
 
       - name: Run slow SFT tests on single GPU
         if: always()
@@ -103,12 +91,14 @@ jobs:
         if: always()
         run: |
           source activate trl
+          pip install deepspeed
           make run_sft_examples
       
       - name: Run end-to-end DPO examples tests on multi GPU
         if: always()
         run: |
           source activate trl
+          pip install deepspeed
           make run_dpo_examples
       
       - name: Generate Reports

scripts/log_reports.py

@@ -109,8 +109,7 @@ def main(slack_channel_name=None):
         from slack_sdk import WebClient
 
         if len(message) > MAX_LEN_MESSAGE:
-            print(f"Truncating long message from {len(message)} to {MAX_LEN_MESSAGE}")
-            message = message[:MAX_LEN_MESSAGE] + "..."
+            message = f"There are {total_num_failed} failed tests in total ! Cannot display the entire summary - please check the action results directly"
 
         if len(message) != 0:
             md_report = {

setup.py

@@ -68,7 +68,7 @@
     "tyro>=0.5.11",
 ]
 EXTRAS = {
-    "test": ["parameterized", "pytest", "pytest-xdist", "accelerate", "pytest-cov", "pytest-xdist", "wandb"],
+    "test": ["parameterized", "pytest", "pytest-xdist", "accelerate", "pytest-cov", "pytest-xdist"],
     "peft": ["peft>=0.4.0"],
     "diffusers": ["diffusers>=0.18.0"],
     "deepspeed": ["deepspeed>=0.9.5"],

tests/slow/test_dpo_slow.py

@@ -25,13 +25,7 @@
 from trl import DPOTrainer, is_peft_available
 
 from ..testing_utils import require_bitsandbytes, require_peft, require_torch_gpu, require_torch_multi_gpu
-from .testing_constants import (
-    DPO_GEN_DURING_EVAL,
-    DPO_LOSS_TYPES,
-    DPO_PRECOMPUTE_LOGITS,
-    GRADIENT_CHECKPOINTING_KWARGS,
-    MODELS_TO_TEST,
-)
+from .testing_constants import DPO_LOSS_TYPES, DPO_PRECOMPUTE_LOGITS, GRADIENT_CHECKPOINTING_KWARGS, MODELS_TO_TEST
 
 
 if is_peft_available():
@@ -57,10 +51,8 @@ def tearDown(self):
         torch.cuda.empty_cache()
         gc.collect()
 
-    @parameterized.expand(
-        list(itertools.product(MODELS_TO_TEST, DPO_LOSS_TYPES, DPO_GEN_DURING_EVAL, DPO_PRECOMPUTE_LOGITS))
-    )
-    def test_dpo_bare_model(self, model_id, loss_type, gen_during_eval, pre_compute_logits):
+    @parameterized.expand(list(itertools.product(MODELS_TO_TEST, DPO_LOSS_TYPES, DPO_PRECOMPUTE_LOGITS)))
+    def test_dpo_bare_model(self, model_id, loss_type, pre_compute_logits):
         """
         A test that tests the simple usage of `DPOTrainer` using a bare model in full precision.
         """
@@ -90,7 +82,6 @@ def test_dpo_bare_model(self, model_id, loss_type, gen_during_eval, pre_compute_
                 tokenizer=tokenizer,
                 train_dataset=self.dataset,
                 eval_dataset=self.dataset,
-                generate_during_eval=gen_during_eval,
                 loss_type=loss_type,
                 precompute_ref_log_probs=pre_compute_logits,
                 max_length=self.max_length,

tests/slow/testing_constants.py

@@ -22,4 +22,3 @@
 
 DPO_LOSS_TYPES = ["sigmoid", "ipo", "kto_pair"]
 DPO_PRECOMPUTE_LOGITS = [True, False]
-DPO_GEN_DURING_EVAL = [True, False]