XFE Git
XFE Studio Git
Git 首页 全局搜索
XFE 主站 文档 NuGet
公开
关注 0 Fork 0 Star 1
返回提交历史

XFEstudio/gpt4free

new gpt-4 beating opensource models

2fad27b2
abc <98614666+xtekky@users.noreply.github.com>
提交于

代码差异

4 个文件 +27 -18
Modified README.md +8 -7
@@ -281,13 +281,14 @@ set G4F_PROXY=http://host:port
281 281 | [beta.theb.ai](https://beta.theb.ai) | `g4f.Provider.Theb` | ✔️ | ✔️ | ✔️ | ![Unknown](https://img.shields.io/badge/Unknown-grey) | ❌ |
282 282 | [you.com](https://you.com) | `g4f.Provider.You` | ✔️ | ✔️ | ✔️ | ![Unknown](https://img.shields.io/badge/Unknown-grey) | ❌ |
283 283
284 ## New OpenSource Models
285 While we wait for gpt-5, here is a list of new models that are at least better than gpt-3.5-turbo. Some rival gpt-4. Expect this list to grow.
286
287 | Website | Provider | parameters |
288 | ------ | ------- | ------ |
289 | [mixtral-8x22b](https://huggingface.co/mistral-community/Mixtral-8x22B-v0.1) | `g4f.Provider.DeepInfra` | 176B / 44b active |
290 | [dbrx-instruct](https://www.databricks.com/blog/introducing-dbrx-new-state-art-open-llm) | `g4f.Provider.DeepInfra` | 132B / 36B active|
284 ## Best OpenSource Models
285 While we wait for gpt-5, here is a list of new models that are at least better than gpt-3.5-turbo. **Some are better than gpt-4**. Expect this list to grow.
286
287 | Website | Provider | parameters | better than |
288 | ------ | ------- | ------ | ------ |
289 | [mixtral-8x22b](https://huggingface.co/mistral-community/Mixtral-8x22B-v0.1) | `g4f.Provider.DeepInfra` | 176B / 44b active | gpt-3.5-turbo |
290 | [dbrx-instruct](https://www.databricks.com/blog/introducing-dbrx-new-state-art-open-llm) | `g4f.Provider.DeepInfra` | 132B / 36B active| gpt-3.5-turbo |
291 | [command-r+](https://txt.cohere.com/command-r-plus-microsoft-azure/) | `g4f.Provider.HuggingChat` | 104B | gpt-4-0613 |
291 292
292 293
293 294 ### GPT-3.5
Modified g4f/Provider/HuggingChat.py +7 -7
@@ -14,13 +14,12 @@ class HuggingChat(AsyncGeneratorProvider, ProviderModelMixin):
14 14 working = True
15 15 default_model = "mistralai/Mixtral-8x7B-Instruct-v0.1"
16 16 models = [
17 "mistralai/Mixtral-8x7B-Instruct-v0.1",
18 "google/gemma-7b-it",
19 "meta-llama/Llama-2-70b-chat-hf",
20 "NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO",
21 "codellama/CodeLlama-34b-Instruct-hf",
22 "mistralai/Mistral-7B-Instruct-v0.2",
23 "openchat/openchat-3.5-0106",
17 "HuggingFaceH4/zephyr-orpo-141b-A35b-v0.1",
18 'CohereForAI/c4ai-command-r-plus',
19 'mistralai/Mixtral-8x7B-Instruct-v0.1',
20 'google/gemma-1.1-7b-it',
21 'NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO',
22 'mistralai/Mistral-7B-Instruct-v0.2'
24 23 ]
25 24 model_aliases = {
26 25 "openchat/openchat_3.5": "openchat/openchat-3.5-0106",
@@ -48,6 +47,7 @@ class HuggingChat(AsyncGeneratorProvider, ProviderModelMixin):
48 47 **kwargs
49 48 ) -> AsyncResult:
50 49 options = {"model": cls.get_model(model)}
50
51 51 system_prompt = "\n".join([message["content"] for message in messages if message["role"] == "system"])
52 52 if system_prompt:
53 53 options["preprompt"] = system_prompt
Modified g4f/Provider/PerplexityLabs.py +3 -2
@@ -19,13 +19,14 @@ class PerplexityLabs(AsyncGeneratorProvider, ProviderModelMixin):
19 19 "sonar-small-online", "sonar-medium-online", "sonar-small-chat", "sonar-medium-chat", "mistral-7b-instruct",
20 20 "codellama-70b-instruct", "llava-v1.5-7b-wrapper", "llava-v1.6-34b", "mixtral-8x7b-instruct",
21 21 "gemma-2b-it", "gemma-7b-it"
22 "mistral-medium", "related"
22 "mistral-medium", "related", "dbrx-instruct"
23 23 ]
24 24 model_aliases = {
25 25 "mistralai/Mistral-7B-Instruct-v0.1": "mistral-7b-instruct",
26 26 "mistralai/Mixtral-8x7B-Instruct-v0.1": "mixtral-8x7b-instruct",
27 27 "codellama/CodeLlama-70b-Instruct-hf": "codellama-70b-instruct",
28 "llava-v1.5-7b": "llava-v1.5-7b-wrapper"
28 "llava-v1.5-7b": "llava-v1.5-7b-wrapper",
29 'databricks/dbrx-instruct': "dbrx-instruct"
29 30 }
30 31
31 32 @classmethod
Modified g4f/models.py +9 -2
@@ -165,7 +165,7 @@ mistral_7b_v02 = Model(
165 165 mixtral_8x22b = Model(
166 166 name = "HuggingFaceH4/zephyr-orpo-141b-A35b-v0.1",
167 167 base_provider = "huggingface",
168 best_provider = DeepInfra
168 best_provider = RetryProvider([HuggingChat, DeepInfra])
169 169 )
170 170
171 171 # Misc models
@@ -269,7 +269,13 @@ pi = Model(
269 269 dbrx_instruct = Model(
270 270 name = 'databricks/dbrx-instruct',
271 271 base_provider = 'mistral',
272 best_provider = DeepInfra
272 best_provider = RetryProvider([DeepInfra, PerplexityLabs])
273 )
274
275 command_r_plus = Model(
276 name = 'CohereForAI/c4ai-command-r-plus',
277 base_provider = 'mistral',
278 best_provider = HuggingChat
273 279 )
274 280
275 281 class ModelUtils:
@@ -324,6 +330,7 @@ class ModelUtils:
324 330 'claude-3-sonnet': claude_3_sonnet,
325 331
326 332 # other
333 'command-r+': command_r_plus,
327 334 'dbrx-instruct': dbrx_instruct,
328 335 'lzlv-70b': lzlv_70b,
329 336 'airoboros-70b': airoboros_70b,