modified the params

megatron/text_generation_utils.py

@@ -190,7 +190,7 @@ def generate_samples_input_from_file(model):
             raw_text = None
             context_count += 1
 
-def generate_samples_eval(model, context, max_gen_length, eos_token_id, do_sample):
+def generate_samples_eval(model, context, max_gen_length, eos_token_id):
     # Generate samples for lm evaluation
     # NEED TO THINK ABOUT eos token
 
@@ -202,17 +202,8 @@ def generate_samples_eval(model, context, max_gen_length, eos_token_id, do_sampl
 
     context_tokens = tokenizer.tokenize(context)
     args.out_seq_length = max_gen_length + len(context_tokens)
-    args.recompute = True #set this default value
     args.eos_id = eos_token_id
 
-    if not do_sample:
-        args.greedy = True
-    else:
-        # set similar to huggngface
-        args.top_p = 1.0
-        args.temperature = 1.0
-        args.top_k = 50
-
     with torch.no_grad():
         token_stream = get_token_stream(model, [context_tokens])
         for counter, decode_tokens in enumerate(token_stream):