markers placed for measuring e2e time for 10 steps

Author: prishajain1

src/maxdiffusion/configs/ltx2_video.yml

@@ -67,7 +67,7 @@ flash_block_sizes: {
 dcn_context_parallelism: 1
 dcn_tensor_parallelism: 1
 ici_data_parallelism: 1
-ici_fsdp_parallelism: 1  
+ici_fsdp_parallelism: 8  
 ici_context_parallelism: -1 # recommended ICI axis to be auto-sharded
 ici_tensor_parallelism: 1
 enable_profiler: False
@@ -86,7 +86,7 @@ dataset_name: ''
 train_split: 'train'
 dataset_type: 'tfrecord'
 cache_latents_text_encoder_outputs: True
-per_device_batch_size: 0.125
+per_device_batch_size: 1.0
 compile_topology_num_slices: -1 
 quantization_local_shard_count: -1
 use_qwix_quantization: False