Merge branch 'main' into gpu-support

serge-chat · Feb 25, 2024 · 38d2245 · 38d2245
2 parents 9dc8f42 + 26bf8a3
commit 38d2245
Show file tree

Hide file tree

Showing 2 changed files with 83 additions and 5 deletions.
diff --git a/README.md b/README.md
@@ -58,9 +58,10 @@ Instructions for setting up Serge on Kubernetes can be found in the [wiki](https
 | Category      | Models |
 |:-------------:|:-------|
 | **Alfred** | 40B-1023 |
+| **BioMistral | 7B |
 | **Code** | 13B, 33B |
 | **CodeLLaMA** | 7B, 7B-Instruct, 7B-Python, 13B, 13B-Instruct, 13B-Python, 34B, 34B-Instruct, 34B-Python |
-| **Gemma** | 2B, 7B |
+| **Gemma** | 2B, 2B-Instruct, 7B, 7B-Instruct |
 | **Falcon** | 7B, 7B-Instruct, 40B, 40B-Instruct |
 | **LLaMA 2**  | 7B, 7B-Chat, 7B-Coder, 13B, 13B-Chat, 70B, 70B-Chat, 70B-OASST |
 | **LLaMA Pro** | 8B, 8B-Instruct |
@@ -74,6 +75,7 @@ Instructions for setting up Serge on Kubernetes can be found in the [wiki](https
 | **Neural-Chat** | 7B-v3.3 | 
 | **Notus** | 7B-v1 |
 | **Notux** | 8x7b-v1 |
+| **Nous-Hermes 2** | Mistral-7B-DPO, Mixtral-8x7B-DPO, Mistral-8x7B-SFT |
 | **OpenChat** | 7B-v3.5-1210 |
 | **OpenLLaMA** | 3B-v2, 7B-v2, 13B-v2 |
 | **Orca 2** | 7B, 13B |

diff --git a/api/src/serge/data/models.json b/api/src/serge/data/models.json
@@ -15,6 +15,22 @@
             }           
         ]
     },
+    {   
+        "name": "BioMistral",
+        "models": [
+            {
+                "name": "BioMistral-7B",
+                "repo": "BioMistral/BioMistral-7B-GGUF",
+                "files": [
+                    {
+                        "name": "q4_K_M",
+                        "filename": "ggml-model-Q4_K_M.gguf",
+                        "disk_space": 4368439424.0
+                    }
+                ]
+            }           
+        ]
+    },    
     {
         "name": "Code",
         "models": [
@@ -200,26 +216,48 @@
         "models": [
             {
                 "name": "Gemma-2B",
-                "repo": "MaziyarPanahi/gemma-2b-GGUF",
+                "repo": "brittlewis12/gemma-2b-GGUF",
                 "files": [
                     {
                         "name": "q4_K_M",
                         "filename": "gemma-2b.Q4_K_M.gguf",
                         "disk_space": 1495245728.0
                     }
                 ]
-            },            
+            },
+            {
+                "name": "Gemma-2B-Instruct",
+                "repo": "brittlewis12/gemma-2b-it-GGUF",
+                "files": [
+                    {
+                        "name": "q4_K_M",
+                        "filename": "gemma-2b-it.Q4_K_M.gguf",
+                        "disk_space": 1630263008.0
+                    }
+                ]
+            },
             {
                 "name": "Gemma-7B",
-                "repo": "MaziyarPanahi/gemma-7b-GGUF",
+                "repo": "brittlewis12/gemma-7b-GGUF",
                 "files": [
                     {
                         "name": "q4_K_M",
                         "filename": "gemma-7b.Q4_K_M.gguf",
                         "disk_space": 5127231648.0
                     }
                 ]
-            }       
+            },
+            {
+                "name": "Gemma-7B-Instruct",
+                "repo": "brittlewis12/gemma-7b-it-GGUF",
+                "files": [
+                    {
+                        "name": "q4_K_M",
+                        "filename": "gemma-7b-it.Q4_K_M.gguf",
+                        "disk_space": 5329759200.0
+                    }
+                ]
+            }             
         ]
     },
     {
@@ -590,6 +628,44 @@
             }
         ]
     },
+    {   
+        "name": "Nous-Hermes-2",
+        "models": [
+            {
+                "name": "Nous-Hermes-2-Mistral-7B-DPO",
+                "repo": "NousResearch/Nous-Hermes-2-Mistral-7B-DPO-GGUF",
+                "files": [
+                    {
+                        "name": "q4_K_M",
+                        "filename": "Nous-Hermes-2-Mistral-7B-DPO.Q4_K_M.gguf",
+                        "disk_space": 4368450560.0
+                    }
+                ]
+            },
+            {
+                "name": "Nous-Hermes-2-Mistral-7B-DPO",
+                "repo": "TheBloke/Nous-Hermes-2-Mixtral-8x7B-DPO-GGUF",
+                "files": [
+                    {
+                        "name": "q4_K_M",
+                        "filename": "nous-hermes-2-mixtral-8x7b-dpo.Q4_K_M.gguf",
+                        "disk_space": 28446421792.0
+                    }
+                ]
+            },
+            {
+                "name": "Nous-Hermes-2-Mistral-7B-SFT",
+                "repo": "NousResearch/Nous-Hermes-2-Mixtral-8x7B-SFT-GGUF",
+                "files": [
+                    {
+                        "name": "q4_K_M",
+                        "filename": "Nous-Hermes-2-Mixtral-8x7B-SFT.Q4_K_M.gguf",
+                        "disk_space": 28446421760.0
+                    }
+                ]
+            }            
+        ]
+    },    
     {   
         "name": "OpenChat",
         "models": [