Training in progress, step 500

ipu_config.json

@@ -7,7 +7,7 @@
   "execute_encoder_on_cpu_for_generation": false,
   "gradient_accumulation_steps": 32,
   "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 sha256:1fc61a0b2e8677824d0f412b1a2b34ec64a41e76efad409b23a77eb2c9b941b0
+oid sha256:37c8b1f11b6923d313b083ef49fb3126dece2213ab93b6fb6cb56b59220d5fdb
 size 248181489

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e0fed49aff71d16a99fb12b3fa788766a5b8d9f715f2619e190ab8fe986b64ca
+oid sha256:354a000e7b498833b02566e145ff807e39ae9f25840debc587c798c2f43ba390
 size 2735