Upload README.md for normalized SALMon repo SpeechPPL/SALMon_Flow-SLM-1B-Extended

README.md

@@ -1,18 +1,15 @@
 ---
 dataset_info:
-- features:
+- config_name: bg_alignment
+  features:
   - name: task
     dtype: string
   - name: ind
     dtype: int64
   - name: positive_audio
-    dtype:
-      audio:
-        sampling_rate: null
+    dtype: audio
   - name: negative_audio
-    dtype:
-      audio:
-        sampling_rate: null
+    dtype: audio
   - name: prompt_audio
     dtype:
       audio:
@@ -30,9 +27,9 @@ dataset_info:
       audio:
         sampling_rate: 16000
   - name: positive_sample_tokenwise_loss
-    list: float32
+    sequence: float32
   - name: negative_sample_tokenwise_loss
-    list: float32
+    sequence: float32
   - name: code_frame_rate
     dtype: int64
   - name: code_depth
@@ -69,25 +66,18 @@ dataset_info:
   - name: train
     num_bytes: 86639819
     num_examples: 200
-    num_shards: 1
-    shards:
-    - train-00000-of-00001.parquet
   download_size: 86639819
   dataset_size: 86639819
-  config_name: bg_alignment
-- features:
+- config_name: bg_all_consistency
+  features:
   - name: task
     dtype: string
   - name: ind
     dtype: int64
   - name: positive_audio
-    dtype:
-      audio:
-        sampling_rate: null
+    dtype: audio
   - name: negative_audio
-    dtype:
-      audio:
-        sampling_rate: null
+    dtype: audio
   - name: audio_transition_s
     dtype: int64
   - name: prompt_audio
@@ -107,11 +97,11 @@ dataset_info:
       audio:
         sampling_rate: 16000
   - name: positive_sample_tokenwise_loss
-    list: float32
+    sequence: float32
   - name: negative_sample_tokenwise_loss
-    list: float32
+    sequence: float32
   - name: prompt_sample_tokenwise_loss
-    list: float32
+    sequence: float32
   - name: model_generated_continuation
     dtype:
       audio:
@@ -125,9 +115,9 @@ dataset_info:
   - name: ppl_sanity
     dtype: int64
   - name: positive_continuation_tokenwise_loss
-    list: float64
+    sequence: float64
   - name: negative_continuation_tokenwise_loss
-    list: float64
+    sequence: float64
   - name: model_generated_continuation1
     dtype:
       audio:
@@ -152,27 +142,18 @@ dataset_info:
   - name: train
     num_bytes: 1269408501
     num_examples: 200
-    num_shards: 3
-    shards:
-    - train-00000-of-00003.parquet
-    - train-00001-of-00003.parquet
-    - train-00002-of-00003.parquet
   download_size: 1269408501
   dataset_size: 1269408501
-  config_name: bg_all_consistency
-- features:
+- config_name: bg_domain_consistency
+  features:
   - name: task
     dtype: string
   - name: ind
     dtype: int64
   - name: positive_audio
-    dtype:
-      audio:
-        sampling_rate: null
+    dtype: audio
   - name: negative_audio
-    dtype:
-      audio:
-        sampling_rate: null
+    dtype: audio
   - name: audio_transition_s
     dtype: int64
   - name: prompt_audio
@@ -192,11 +173,11 @@ dataset_info:
       audio:
         sampling_rate: 16000
   - name: positive_sample_tokenwise_loss
-    list: float32
+    sequence: float32
   - name: negative_sample_tokenwise_loss
-    list: float32
+    sequence: float32
   - name: prompt_sample_tokenwise_loss
-    list: float32
+    sequence: float32
   - name: model_generated_continuation
     dtype:
       audio:
@@ -210,9 +191,9 @@ dataset_info:
   - name: ppl_sanity
     dtype: int64
   - name: positive_continuation_tokenwise_loss
-    list: float64
+    sequence: float64
   - name: negative_continuation_tokenwise_loss
-    list: float64
+    sequence: float64
   - name: model_generated_continuation1
     dtype:
       audio:
@@ -237,27 +218,18 @@ dataset_info:
   - name: train
     num_bytes: 1270504334
     num_examples: 200
-    num_shards: 3
-    shards:
-    - train-00000-of-00003.parquet
-    - train-00001-of-00003.parquet
-    - train-00002-of-00003.parquet
   download_size: 1270504334
   dataset_size: 1270504334
-  config_name: bg_domain_consistency
-- features:
+- config_name: gender_consistency
+  features:
   - name: task
     dtype: string
   - name: ind
     dtype: int64
   - name: positive_audio
-    dtype:
-      audio:
-        sampling_rate: null
+    dtype: audio
   - name: negative_audio
-    dtype:
-      audio:
-        sampling_rate: null
+    dtype: audio
   - name: audio_transition_s
     dtype: int64
   - name: prompt_audio
@@ -277,11 +249,11 @@ dataset_info:
       audio:
         sampling_rate: 16000
   - name: positive_sample_tokenwise_loss
-    list: float32
+    sequence: float32
   - name: negative_sample_tokenwise_loss
-    list: float32
+    sequence: float32
   - name: prompt_sample_tokenwise_loss
-    list: float32
+    sequence: float32
   - name: model_generated_continuation
     dtype:
       audio:
@@ -295,9 +267,9 @@ dataset_info:
   - name: ppl_sanity
     dtype: int64
   - name: positive_continuation_tokenwise_loss
-    list: float64
+    sequence: float64
   - name: negative_continuation_tokenwise_loss
-    list: float64
+    sequence: float64
   - name: model_generated_continuation1
     dtype:
       audio:
@@ -322,27 +294,18 @@ dataset_info:
   - name: train
     num_bytes: 1315571898
     num_examples: 200
-    num_shards: 3
-    shards:
-    - train-00000-of-00003.parquet
-    - train-00001-of-00003.parquet
-    - train-00002-of-00003.parquet
   download_size: 1315571898
   dataset_size: 1315571898
-  config_name: gender_consistency
-- features:
+- config_name: rir_consistency
+  features:
   - name: task
     dtype: string
   - name: ind
     dtype: int64
   - name: positive_audio
-    dtype:
-      audio:
-        sampling_rate: null
+    dtype: audio
   - name: negative_audio
-    dtype:
-      audio:
-        sampling_rate: null
+    dtype: audio
   - name: audio_transition_s
     dtype: int64
   - name: prompt_audio
@@ -362,11 +325,11 @@ dataset_info:
       audio:
         sampling_rate: 16000
   - name: positive_sample_tokenwise_loss
-    list: float32
+    sequence: float32
   - name: negative_sample_tokenwise_loss
-    list: float32
+    sequence: float32
   - name: prompt_sample_tokenwise_loss
-    list: float32
+    sequence: float32
   - name: model_generated_continuation
     dtype:
       audio:
@@ -380,9 +343,9 @@ dataset_info:
   - name: ppl_sanity
     dtype: int64
   - name: positive_continuation_tokenwise_loss
-    list: float64
+    sequence: float64
   - name: negative_continuation_tokenwise_loss
-    list: float64
+    sequence: float64
   - name: model_generated_continuation1
     dtype:
       audio:
@@ -407,27 +370,18 @@ dataset_info:
   - name: train
     num_bytes: 1300026144
     num_examples: 200
-    num_shards: 3
-    shards:
-    - train-00000-of-00003.parquet
-    - train-00001-of-00003.parquet
-    - train-00002-of-00003.parquet
   download_size: 1300026144
   dataset_size: 1300026144
-  config_name: rir_consistency
-- features:
+- config_name: sentiment_alignment
+  features:
   - name: task
     dtype: string
   - name: ind
     dtype: int64
   - name: positive_audio
-    dtype:
-      audio:
-        sampling_rate: null
+    dtype: audio
   - name: negative_audio
-    dtype:
-      audio:
-        sampling_rate: null
+    dtype: audio
   - name: prompt_audio
     dtype:
       audio:
@@ -445,9 +399,9 @@ dataset_info:
       audio:
         sampling_rate: 16000
   - name: positive_sample_tokenwise_loss
-    list: float32
+    sequence: float32
   - name: negative_sample_tokenwise_loss
-    list: float32
+    sequence: float32
   - name: code_frame_rate
     dtype: int64
   - name: code_depth
@@ -484,25 +438,18 @@ dataset_info:
   - name: train
     num_bytes: 46471520
     num_examples: 200
-    num_shards: 1
-    shards:
-    - train-00000-of-00001.parquet
   download_size: 46471520
   dataset_size: 46471520
-  config_name: sentiment_alignment
-- features:
+- config_name: sentiment_consistency
+  features:
   - name: task
     dtype: string
   - name: ind
     dtype: int64
   - name: positive_audio
-    dtype:
-      audio:
-        sampling_rate: null
+    dtype: audio
   - name: negative_audio
-    dtype:
-      audio:
-        sampling_rate: null
+    dtype: audio
   - name: audio_transition_s
     dtype: int64
   - name: prompt_audio
@@ -522,11 +469,11 @@ dataset_info:
       audio:
         sampling_rate: 16000
   - name: positive_sample_tokenwise_loss
-    list: float32
+    sequence: float32
   - name: negative_sample_tokenwise_loss
-    list: float32
+    sequence: float32
   - name: prompt_sample_tokenwise_loss
-    list: float32
+    sequence: float32
   - name: model_generated_continuation
     dtype:
       audio:
@@ -540,9 +487,9 @@ dataset_info:
   - name: ppl_sanity
     dtype: int64
   - name: positive_continuation_tokenwise_loss
-    list: float64
+    sequence: float64
   - name: negative_continuation_tokenwise_loss
-    list: float64
+    sequence: float64
   - name: model_generated_continuation1
     dtype:
       audio:
@@ -567,27 +514,18 @@ dataset_info:
   - name: train
     num_bytes: 1311269667
     num_examples: 200
-    num_shards: 3
-    shards:
-    - train-00000-of-00003.parquet
-    - train-00001-of-00003.parquet
-    - train-00002-of-00003.parquet
   download_size: 1311269667
   dataset_size: 1311269667
-  config_name: sentiment_consistency
-- features:
+- config_name: speaker_consistency
+  features:
   - name: task
     dtype: string
   - name: ind
     dtype: int64
   - name: positive_audio
-    dtype:
-      audio:
-        sampling_rate: null
+    dtype: audio
   - name: negative_audio
-    dtype:
-      audio:
-        sampling_rate: null
+    dtype: audio
   - name: audio_transition_s
     dtype: int64
   - name: prompt_audio
@@ -607,11 +545,11 @@ dataset_info:
       audio:
         sampling_rate: 16000
   - name: positive_sample_tokenwise_loss
-    list: float32
+    sequence: float32
   - name: negative_sample_tokenwise_loss
-    list: float32
+    sequence: float32
   - name: prompt_sample_tokenwise_loss
-    list: float32
+    sequence: float32
   - name: model_generated_continuation
     dtype:
       audio:
@@ -625,9 +563,9 @@ dataset_info:
   - name: ppl_sanity
     dtype: int64
   - name: positive_continuation_tokenwise_loss
-    list: float64
+    sequence: float64
   - name: negative_continuation_tokenwise_loss
-    list: float64
+    sequence: float64
   - name: model_generated_continuation1
     dtype:
       audio:
@@ -652,14 +590,8 @@ dataset_info:
   - name: train
     num_bytes: 1316303506
     num_examples: 200
-    num_shards: 3
-    shards:
-    - train-00000-of-00003.parquet
-    - train-00001-of-00003.parquet
-    - train-00002-of-00003.parquet
   download_size: 1316303506
   dataset_size: 1316303506
-  config_name: speaker_consistency
 ---
 
 # SALMon Normalized Dataset