Training in progress, step 500

ipu_config.json

@@ -7,7 +7,7 @@
   "execute_encoder_on_cpu_for_generation": false,
   "gradient_accumulation_steps": 16,
   "inference_device_iterations": 5,
-  "inference_replication_factor": 1,
+  "inference_replication_factor": 4,
   "ipus_per_replica": 4,
   "layers_per_ipu": [
     0,
@@ -22,8 +22,8 @@
   "optimum_version": "1.5.1",
   "output_mode": "final",
   "recompute_checkpoint_every_layer": true,
-  "replicated_tensor_sharding": false,
-  "replication_factor": 1,
+  "replicated_tensor_sharding": true,
+  "replication_factor": 4,
   "seed": 42,
   "sharded_execution_for_inference": false,
   "transformers_version": "4.20.1"

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:36320590637ba5df4c21bf5789b9554670d88a8f89c10dc2ca72bd738d2763c0
+oid sha256:b6ea9c5196058fad77a012b87e30e6429992729394461ce69d760675a2d418ae
 size 249361645

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1ff886cafb02ffc57877d4603c284b215d327454952326bb64c3dadf5b056737
+oid sha256:38a6acd9e7e69dc8bc762ace4138e9906131074cd0ac6f7d24493c963f7af192
 size 2735