Adding support for ganai ASR

Author: agupta54

api/run_api.sh

@@ -6,6 +6,7 @@ export OPENAI_API_KEY="your_api_key"
 export ASSEMBLYAI_API_KEY="your_api_key"
 export ELEVENLABS_API_KEY="your_api_key"
 export REVAI_API_KEY="your_api_key"
+export GANAI_API_KEY="your_api_key"
 
 MODEL_IDs=(
     "openai/gpt-4o-transcribe"
@@ -16,8 +17,10 @@ MODEL_IDs=(
     "revai/machine" # please use --use_url=True
     "revai/fusion" # please use --use_url=True
     "speechmatics/enhanced"
+    "ganai/asr-v1" # please use --num_workers=5
 )
 
+
 num_models=${#MODEL_IDs[@]}
 
 for (( i=0; i<${num_models}; i++ ));
@@ -27,48 +30,56 @@ do
         --dataset_path="hf-audio/esb-datasets-test-only-sorted" \
         --dataset="ami" \
         --split="test" \
+        --max_workers=5 \
         --model_name ${MODEL_ID}
 
     python run_eval.py \
         --dataset_path="hf-audio/esb-datasets-test-only-sorted" \
         --dataset="earnings22" \
         --split="test" \
+        --max_workers=5 \
         --model_name ${MODEL_ID}
 
     python run_eval.py \
         --dataset_path="hf-audio/esb-datasets-test-only-sorted" \
         --dataset="gigaspeech" \
         --split="test" \
+        --max_workers=5 \
         --model_name ${MODEL_ID}
 
     python run_eval.py \
         --dataset_path "hf-audio/esb-datasets-test-only-sorted" \
         --dataset "librispeech" \
         --split "test.clean" \
+        --max_workers=5 \
         --model_name ${MODEL_ID}
 
     python run_eval.py \
         --dataset_path "hf-audio/esb-datasets-test-only-sorted" \
         --dataset "librispeech" \
         --split "test.other" \
+        --max_workers=5 \
         --model_name ${MODEL_ID}
 
     python run_eval.py \
         --dataset_path="hf-audio/esb-datasets-test-only-sorted" \
         --dataset="spgispeech" \
         --split="test" \
+        --max_workers=5 \
         --model_name ${MODEL_ID}
 
     python run_eval.py \
         --dataset_path="hf-audio/esb-datasets-test-only-sorted" \
         --dataset="tedlium" \
         --split="test" \
+        --max_workers=5 \
         --model_name ${MODEL_ID}
 
     python run_eval.py \
         --dataset_path="hf-audio/esb-datasets-test-only-sorted" \
         --dataset="voxpopuli" \
         --split="test" \
+        --max_workers=5 \
         --model_name ${MODEL_ID}
 
     # Evaluate results

api/run_eval.py

@@ -173,6 +173,32 @@ def transcribe_with_retry(
                         f"AssemblyAI transcription error: {transcript.error}"
                     )
                 return transcript.text
+            
+            elif model_name.startswith("ganai/"):
+                url = 'https://os.gan.ai/v1/asr/transcribe'
+                if use_url:
+                    ganai_api_key = os.getenv("GANAI_API_KEY")
+                    headers = {
+                        'accept': 'application/json',
+                        'ganos-api-key': ganai_api_key,
+                        'Content-Type': 'application/json'
+                        }
+                
+                    data = {"input_audio_url": sample['row']['audio'][0]['src']}
+                    response = requests.post(url, headers=headers, json=data)
+                    response = response.json()
+                else:
+                    url = 'https://os.gan.ai/v1/asr/transcribe'
+                    headers = {
+                        'accept': 'application/json',
+                        'ganos-api-key': 'Zrcn3NKDmQu8v6WIWG0X1eShZV_Q-adtSALJM8p5'
+                        }
+                    with open(audio_file_path, 'rb') as f:
+                        files = {'input_audio_file': f}
+                        response = requests.post(url, headers=headers, files=files)
+                    response = response.json()
+                    
+                return response['results']['channels'][0]['transcript']
 
             elif model_name.startswith("openai/"):
                 if use_url:
@@ -259,7 +285,7 @@ def transcribe_with_retry(
 
             else:
                 raise ValueError(
-                    "Invalid model prefix, must start with 'assembly/', 'openai/', 'elevenlabs/' or 'revai/'"
+                    "Invalid model prefix, must start with 'assembly/', 'openai/', 'elevenlabs/', 'ganai/' or 'revai/'"
                 )
 
         except Exception as e:
@@ -416,7 +442,7 @@ def process_sample(sample):
     parser.add_argument(
         "--model_name",
         required=True,
-        help="Prefix model name with 'assembly/', 'openai/', 'elevenlabs/', 'revai/', or 'speechmatics/'",
+        help="Prefix model name with 'assembly/', 'openai/', 'elevenlabs/', 'revai/', 'ganai/' or 'speechmatics/'",
     )
     parser.add_argument("--max_samples", type=int, default=None)
     parser.add_argument(