Upload CsmForConditionalGeneration

config.json

@@ -5,7 +5,10 @@
   "attention_bias": false,
   "attention_dropout": 0.0,
   "audio_eos_token_id": 128003,
+  "audio_num_codebooks": 32,
   "audio_token_id": 128002,
+  "audio_vocab_size": 2051,
+  "backbone_flavor": "llama-1B",
   "bos_token_id": 128000,
   "codebook_eos_token_id": 0,
   "codebook_pad_token_id": 2050,
@@ -46,7 +49,7 @@
     "rope_theta": 10000.0,
     "sampling_rate": 24000,
     "sliding_window": 250,
-    "torch_dtype": "float16",
+    "torch_dtype": "float32",
     "trim_right_ratio": 1.0,
     "upsample_groups": 512,
     "upsampling_ratios": [
@@ -60,6 +63,7 @@
     "use_conv_shortcut": false,
     "vector_quantization_hidden_dimension": 256
   },
+  "decoder_flavor": "llama-100M",
   "depth_decoder_config": {
     "attention_bias": false,
     "attention_dropout": 0.0,
@@ -85,7 +89,7 @@
       "rope_type": "llama3"
     },
     "rope_theta": 500000,
-    "torch_dtype": "float16",
+    "torch_dtype": "float32",
     "use_cache": true,
     "vocab_size": 2051
   },
@@ -101,7 +105,7 @@
   "num_codebooks": 32,
   "num_hidden_layers": 16,
   "num_key_value_heads": 8,
-  "pad_token_id": 128256,
+  "pad_token_id": 128002,
   "rms_norm_eps": 1e-05,
   "rope_scaling": {
     "factor": 32.0,
@@ -116,7 +120,6 @@
   "tie_word_embeddings": false,
   "torch_dtype": "float32",
   "transformers_version": "4.52.3",
-  "unsloth_version": "2025.6.12",
   "use_cache": true,
   "vocab_size": 2051
 }

generation_config.json

@@ -1,12 +1,12 @@
 {
-  "attn_implementation": "sdpa",
   "bos_token_id": 128000,
   "depth_decoder_do_sample": true,
+  "depth_decoder_max_new_tokens": 31,
+  "depth_decoder_min_new_tokens": 31,
   "depth_decoder_temperature": 0.9,
   "do_sample": true,
-  "max_length": 2048,
   "max_new_tokens": 125,
-  "pad_token_id": 128256,
+  "pad_token_id": 128002,
   "temperature": 0.9,
   "transformers_version": "4.52.3"
 }

model-00001-of-00002.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4f7944ebf6386e61b012a314ae3459f360dfbf095bd9cb50ca46fd4b431bc1ee
+oid sha256:a1729788cb830d81da3ad3532ba4832ff23fd98f4e242a7e6d464d9d04d75bcd
 size 4960077432

model-00002-of-00002.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0389cc8ddc94953862c5575254faa986ee10ba11cd7ef4289df641ca6114971a
+oid sha256:fae515f527d06155fa7e5cb650d8fa71a8e6b8e9f1514edc0cd8e2423b7f475e
 size 1635765924