Training in progress, step 1000

config.json

@@ -0,0 +1,32 @@
+{
+  "act_format": "linear",
+  "act_on": false,
+  "act_type": "associative",
+  "architectures": [
+    "InnerLoopARMTForCausalLM"
+  ],
+  "attend_to_previous_input": false,
+  "base_model_config": null,
+  "base_model_name": "meta-llama/Llama-3.2-1B",
+  "constant_depth": false,
+  "correction": true,
+  "d_mem": 64,
+  "dtype": "bfloat16",
+  "freeze_mem": false,
+  "gating": false,
+  "layers_attr": "model.layers",
+  "max_hop": 4,
+  "model_type": "armt",
+  "n_heads": 1,
+  "noisy_halting": false,
+  "num_mem_tokens": 32,
+  "segment_alignment": "left",
+  "segment_size": 1024,
+  "sliding_window": true,
+  "time_penalty": 0.0,
+  "transformers_version": "4.57.3",
+  "use_denom": true,
+  "use_sink": true,
+  "wrap_layers": null,
+  "wrap_pos": false
+}

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:23edd802ab9b87d8cda92f29e0647f2a02ed27b9f578df52bc1e9cc1295afacf
+size 2089174366

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7c1b34bdd8b7d94940328e08cd9974c876eec460126e0c8ca633a3609f7e46d3
+size 6904