Training in progress, step 50

adapter_config.json

@@ -25,17 +25,12 @@
   "peft_type": "LORA",
   "peft_version": "0.18.0",
   "qalora_group_size": 16,
-  "r": 32,
+  "r": 64,
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
-    "gate_proj",
-    "v_proj",
-    "o_proj",
     "q_proj",
-    "k_proj",
-    "down_proj",
-    "up_proj"
+    "v_proj"
   ],
   "target_parameters": null,
   "task_type": "CAUSAL_LM",

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e57d5ae92c994fc60688e42b21c2704b821316eb018be6638ad66cda04b68d8a
-size 139512976
+oid sha256:a58cb67e09705a1599d4bb4443456d31b6a93360d9c9299d9580e92d3cb7a2af
+size 51395296

config.json

@@ -163,8 +163,11 @@
   "llm_dim": 2048,
   "lora_alpha": 32,
   "lora_dropout": 0.0,
-  "lora_r": 32,
-  "lora_target_modules": "all-linear",
+  "lora_r": 64,
+  "lora_target_modules": [
+    "v_proj",
+    "q_proj"
+  ],
   "max_new_tokens": 96,
   "model_dtype": "bfloat16",
   "model_type": "asr_model",
@@ -173,7 +176,6 @@
   "num_experts": 4,
   "num_experts_per_tok": 2,
   "pipeline_tag": "automatic-speech-recognition",
-  "pretrained_model_path": "mazesmazes/tiny-audio-glm",
   "projector_dropout": 0.0,
   "projector_hidden_dim": null,
   "projector_init_std": 0.02,

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:525af9a4adeb0e78775e525e9c9ab5462135766ffae4b0896b8c96a2c14dd578
+oid sha256:40bf28e411e4a834a9ade44f7ec62ae66830b16b3eeb54dc0564122ed560a875
 size 58732960

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4a9a23711201fd300282c14e57be4ead25e1ced687311c008a52a0795d9bc061
+oid sha256:7d2b97f0dd342aaafcd84bf11c5664c445f4b7720b9942bff45b06cd5c961158
 size 5201