Fix typo & improve comment (#43)

Author: li-plus

bloom-inference-scripts/bloom-accelerate-inference.py

@@ -149,11 +149,11 @@ def generate():
         generated = generate()
         total_new_tokens_generated += sum(new_tokens for _, _, new_tokens in generated)
     torch.cuda.synchronize()
-    througput = (time.time() - t0) / (total_new_tokens_generated)
+    throughput = (time.time() - t0) / (total_new_tokens_generated)
     print_rank0(
         f"""
 *** Performance stats:
-Throughput per token including tokenize: {througput*1000:.2f} msecs
+Throughput per token including tokenize: {throughput*1000:.2f} msecs
 Start to ready to generate: {t_ready - t_start:.3f} secs
 Tokenize and generate {total_new_tokens_generated} (bs={args.batch_size}) tokens: {t_generate_span:.3f} secs
 Start to finish: {t_ready - t_start + t_generate_span:.3f} secs

bloom-inference-scripts/bloom-ds-inference.py

@@ -282,11 +282,11 @@ def generate():
         generated = generate()
         total_new_tokens_generated += sum(new_tokens for _, _, new_tokens in generated)
     torch.cuda.synchronize()
-    througput = (time.time() - t0) / (total_new_tokens_generated)
+    throughput = (time.time() - t0) / (total_new_tokens_generated)
     print_rank0(
         f"""
 *** Performance stats:
-Throughput per token including tokenize: {througput*1000:.2f} msecs
+Throughput per token including tokenize: {throughput*1000:.2f} msecs
 Start to ready to generate: {t_ready - t_start:.3f} secs
 Tokenize and generate {total_new_tokens_generated} (bs={args.batch_size}) tokens: {t_generate_span:.3f} secs
 Start to finish: {t_ready - t_start + t_generate_span:.3f} secs

bloom-inference-scripts/bloom-ds-zero-inference.py

@@ -1,8 +1,8 @@
 # usage:
-# deepspeed --num_gpus 8 bloom-ds-inference.py --name bigscience/bloom
+# deepspeed --num_gpus 8 bloom-ds-zero-inference.py --name bigscience/bloom
 #
 # to run benchmarks:
-# deepspeed --num_gpus 8 bloom-ds-inference.py --name bigscience/bloom --benchmark
+# deepspeed --num_gpus 8 bloom-ds-zero-inference.py --name bigscience/bloom --benchmark
 #
 
 
@@ -212,11 +212,11 @@ def generate():
     torch.cuda.synchronize()
     # note that we actually generate world_size unique streams (though the benchmark feeds the same inputs)
     total_new_tokens_generated *= world_size
-    througput = (time.time() - t0) / (total_new_tokens_generated)
+    throughput = (time.time() - t0) / (total_new_tokens_generated)
     print_rank0(
         f"""
 *** Performance stats:
-Throughput per token including tokenize: {througput*1000:.2f} msecs
+Throughput per token including tokenize: {throughput*1000:.2f} msecs
 Start to ready to generate: {t_ready - t_start:.3f} secs
 Tokenize and generate {total_new_tokens_generated} (bs={args.batch_size}) tokens: {t_generate_span:.3f} secs
 Start to finish: {t_ready - t_start + t_generate_span:.3f} secs