Merge pull request #4294 from BerriAI/litellm_add_ft_gpt

[feat] Add ft:gpt-4, ft:gpt-4o models

litellm/cost_calculator.py

@@ -191,6 +191,29 @@ def cost_per_token(
             * completion_tokens
         )
         return prompt_tokens_cost_usd_dollar, completion_tokens_cost_usd_dollar
+    elif "ft:gpt-4-0613" in model:
+        print_verbose(f"Cost Tracking: {model} is an OpenAI FinteTuned LLM")
+        # fuzzy match ft:gpt-4-0613:abcd-id-cool-litellm
+        prompt_tokens_cost_usd_dollar = (
+            model_cost_ref["ft:gpt-4-0613"]["input_cost_per_token"] * prompt_tokens
+        )
+        completion_tokens_cost_usd_dollar = (
+            model_cost_ref["ft:gpt-4-0613"]["output_cost_per_token"] * completion_tokens
+        )
+        return prompt_tokens_cost_usd_dollar, completion_tokens_cost_usd_dollar
+    elif "ft:gpt-4o-2024-05-13" in model:
+        print_verbose(f"Cost Tracking: {model} is an OpenAI FinteTuned LLM")
+        # fuzzy match ft:gpt-4o-2024-05-13:abcd-id-cool-litellm
+        prompt_tokens_cost_usd_dollar = (
+            model_cost_ref["ft:gpt-4o-2024-05-13"]["input_cost_per_token"]
+            * prompt_tokens
+        )
+        completion_tokens_cost_usd_dollar = (
+            model_cost_ref["ft:gpt-4o-2024-05-13"]["output_cost_per_token"]
+            * completion_tokens
+        )
+        return prompt_tokens_cost_usd_dollar, completion_tokens_cost_usd_dollar
+
     elif "ft:davinci-002" in model:
         print_verbose(f"Cost Tracking: {model} is an OpenAI FinteTuned LLM")
         # fuzzy match ft:davinci-002:abcd-id-cool-litellm

litellm/model_prices_and_context_window_backup.json

@@ -234,6 +234,30 @@
         "litellm_provider": "openai",
         "mode": "chat"
     },
+    "ft:gpt-4-0613": {
+        "max_tokens": 4096,
+        "max_input_tokens": 8192,
+        "max_output_tokens": 4096,
+        "input_cost_per_token": 0.00003,
+        "output_cost_per_token": 0.00006,
+        "litellm_provider": "openai",
+        "mode": "chat",
+        "supports_function_calling": true,
+        "source": "OpenAI needs to add pricing for this ft model, will be updated when added by OpenAI. Defaulting to base model pricing"
+    },
+    "ft:gpt-4o-2024-05-13": {
+        "max_tokens": 4096,
+        "max_input_tokens": 128000,
+        "max_output_tokens": 4096,
+        "input_cost_per_token": 0.000005,
+        "output_cost_per_token": 0.000015,
+        "litellm_provider": "openai",
+        "mode": "chat",
+        "supports_function_calling": true,
+        "supports_parallel_function_calling": true,
+        "supports_vision": true,
+        "source": "OpenAI needs to add pricing for this ft model, will be updated when added by OpenAI. Defaulting to base model pricing"
+    },
     "ft:davinci-002": {
         "max_tokens": 16384,
         "max_input_tokens": 16384,

litellm/tests/test_utils.py

@@ -24,6 +24,7 @@ from litellm.utils import (
     create_pretrained_tokenizer,
     create_tokenizer,
     function_to_dict,
+    get_llm_provider,
     get_max_tokens,
     get_supported_openai_params,
     get_token_count,
@@ -517,3 +518,33 @@ def test_duration_in_seconds():
     value = _duration_in_seconds(duration="1mo")
 
     assert value - expected_duration < 2
+
+
+def test_get_llm_provider_ft_models():
+    """
+    All ft prefixed models should map to OpenAI
+    gpt-3.5-turbo-0125 (recommended),
+    gpt-3.5-turbo-1106,
+    gpt-3.5-turbo-0613,
+    gpt-4-0613 (experimental)
+    gpt-4o-2024-05-13.
+    babbage-002, davinci-002,
+
+    """
+    model, custom_llm_provider, _, _ = get_llm_provider(model="ft:gpt-3.5-turbo-0125")
+    assert custom_llm_provider == "openai"
+
+    model, custom_llm_provider, _, _ = get_llm_provider(model="ft:gpt-3.5-turbo-1106")
+    assert custom_llm_provider == "openai"
+
+    model, custom_llm_provider, _, _ = get_llm_provider(model="ft:gpt-3.5-turbo-0613")
+    assert custom_llm_provider == "openai"
+
+    model, custom_llm_provider, _, _ = get_llm_provider(model="ft:gpt-4-0613")
+    assert custom_llm_provider == "openai"
+
+    model, custom_llm_provider, _, _ = get_llm_provider(model="ft:gpt-3.5-turbo-0613")
+    assert custom_llm_provider == "openai"
+
+    model, custom_llm_provider, _, _ = get_llm_provider(model="ft:gpt-4o-2024-05-13")
+    assert custom_llm_provider == "openai"

litellm/utils.py

@@ -4050,6 +4050,7 @@ def get_llm_provider(
         if (
             model in litellm.open_ai_chat_completion_models
             or "ft:gpt-3.5-turbo" in model
+            or "ft:gpt-4" in model  # catches ft:gpt-4-0613, ft:gpt-4o
             or model in litellm.openai_image_generation_models
         ):
             custom_llm_provider = "openai"

model_prices_and_context_window.json

@@ -234,6 +234,30 @@
         "litellm_provider": "openai",
         "mode": "chat"
     },
+    "ft:gpt-4-0613": {
+        "max_tokens": 4096,
+        "max_input_tokens": 8192,
+        "max_output_tokens": 4096,
+        "input_cost_per_token": 0.00003,
+        "output_cost_per_token": 0.00006,
+        "litellm_provider": "openai",
+        "mode": "chat",
+        "supports_function_calling": true,
+        "source": "OpenAI needs to add pricing for this ft model, will be updated when added by OpenAI. Defaulting to base model pricing"
+    },
+    "ft:gpt-4o-2024-05-13": {
+        "max_tokens": 4096,
+        "max_input_tokens": 128000,
+        "max_output_tokens": 4096,
+        "input_cost_per_token": 0.000005,
+        "output_cost_per_token": 0.000015,
+        "litellm_provider": "openai",
+        "mode": "chat",
+        "supports_function_calling": true,
+        "supports_parallel_function_calling": true,
+        "supports_vision": true,
+        "source": "OpenAI needs to add pricing for this ft model, will be updated when added by OpenAI. Defaulting to base model pricing"
+    },
     "ft:davinci-002": {
         "max_tokens": 16384,
         "max_input_tokens": 16384,