Pass pad_id, to the generate (#2476)

Co-authored-by: Mark Obozov <[email protected]>

Author: krammnic

recipes/configs/qwen2_5/evaluation.yaml

@@ -1,7 +1,7 @@
 # Config for EleutherEvalRecipe in eleuther_eval.py
 #
 # To launch, run the following command from root torchtune directory:
-#    tune run eleuther_eval --config eleuther_evaluation tasks=["truthfulqa_mc2","hellaswag"]
+#    tune run eleuther_eval --config qwen2_5/evaluation
 
 output_dir: ./ # Not needed
 
@@ -11,7 +11,7 @@ model:
 
 checkpointer:
   _component_: torchtune.training.FullModelHFCheckpointer
-  checkpoint_dir: /tmp/Qwen2.5-0_5B-Instruct
+  checkpoint_dir: /tmp/Qwen2.5-0.5B-Instruct
   checkpoint_files: [
     model.safetensors,
   ]
@@ -21,8 +21,8 @@ checkpointer:
 # Tokenizer
 tokenizer:
   _component_: torchtune.models.qwen2_5.qwen2_5_tokenizer
-  path: /tmp/Qwen2.5-0_5B-Instruct/vocab.json
-  merges_file: /tmp/Qwen2.5-0_5B-Instruct/merges.txt
+  path: /tmp/Qwen2.5-0.5B-Instruct/vocab.json
+  merges_file: /tmp/Qwen2.5-0.5B-Instruct/merges.txt
   max_seq_len: null
 
 # Environment

recipes/eleuther_eval.py

@@ -408,12 +408,14 @@ def _model_generate(
             dtype=self._dtype,
             decoder_max_seq_len=self.max_length,
         ):
+
             toks, _ = generate(
                 self.model,
                 maybe_padded_context,
                 max_generated_tokens=self.max_gen_toks,
                 temperature=temperature,
                 top_k=None,
+                pad_id=self._tokenizer.pad_id,
                 stop_tokens=self._tokenizer.stop_tokens,
             )
         return toks[:bsz]