Fix eval harness on next nvgpt (#117)

* Fix evaluation harness

* Fix few issues

launcher_scripts/nemo_launcher/collections/eval_harness/lm_eval/models/nemo_gpt3.py

@@ -24,6 +24,7 @@
 from nemo.collections.nlp.modules.common.megatron.megatron_init import fake_initialize_model_parallel
 from nemo.collections.nlp.modules.common.text_generation_utils import generate, get_computeprob_response
 from nemo.collections.nlp.parts.nlp_overrides import NLPDDPStrategy, NLPSaveRestoreConnector
+from nemo.collections.common.tokenizers.sentencepiece_tokenizer import SentencePieceTokenizer
 from nemo.utils import logging
 from nemo.utils.app_state import AppState
 from nemo.utils.get_rank import is_global_rank_zero
@@ -35,11 +36,11 @@
 
 
 class RequestDataset(Dataset):
-    def __init__(self, requests, tokenizer) -> None:
+    def __init__(self, requests, tokenizer, max_length=2048) -> None:
         super().__init__()
         self.requests = requests
         self.tokenizer = tokenizer
-        self.max_length = 2048
+        self.max_length = max_length
 
     def __len__(self):
         return len(self.requests)
@@ -48,6 +49,9 @@ def __getitem__(self, index):
         context, continuation = self.requests[index]
         context_enc = self.tokenizer.text_to_ids(context) if isinstance(context, str) else context
         continuation_enc = self.tokenizer.text_to_ids(continuation) if isinstance(continuation, str) else continuation
+        if isinstance(self.tokenizer, SentencePieceTokenizer):
+            continuation_enc = continuation_enc[1:]
+
         # sanity check
         assert len(context_enc) > 0
         assert len(continuation_enc) > 0
@@ -167,12 +171,8 @@ def __init__(self, args, truncate=False, batch_size=1):
         self.model.eval()
 
         self.max_length = self.model.cfg.get("max_position_embeddings")
-        assert self.tokenizer.text_to_ids("hello\n\nhello") == [
-            31373,
-            198,
-            198,
-            31373,
-        ], "Tokenizer text_to_ids is not working as expected."
+        self.pad_id = self.tokenizer.pad_id
+        self.eos_id = self.tokenizer.eos_id
 
         self.truncate = truncate
         self.batch_size = batch_size
@@ -202,7 +202,8 @@ def loglikelihood(self, requests):
     """
 
     def _loglikelihood(self, requests):
-        def pad_collate(batch, eos_id=50256):
+        def pad_collate(batch):
+            eos_id = self.eos_id
             tokens = [item[0] for item in batch]
             conti_lens = [item[1] for item in batch]
             lens = [len(token) - 1 for token in tokens]  # fake delete last token by reducing input len
@@ -241,7 +242,7 @@ def _collate(x):  # used to reorder request and remove duplications
             return -len(toks), tuple(toks)
 
         reord = utils.Reorderer(requests, _collate)
-        request_ds = RequestDataset(reord.get_reordered(), self.model.tokenizer)
+        request_ds = RequestDataset(reord.get_reordered(), self.model.tokenizer, self.max_length)
         request_dl = DataLoader(request_ds, collate_fn=pad_collate, batch_size=self.batch_size, shuffle=False)
 
         def logits_to_results(batch, response):
@@ -314,7 +315,7 @@ def loglikelihood_rolling(self, requests):
                     utils.make_disjoint_window,
                     utils.get_rolling_token_windows(
                         token_list=self.tokenizer.text_to_ids(string),
-                        prefix_token=50256,
+                        prefix_token=self.eos_id,
                         max_seq_len=self.max_length,
                         context_len=1,
                     ),