oobabooga runpod worker

matrix_pygmalion_bot/ai/koboldcpp.py

@@ -28,8 +28,7 @@ async def generate_sync(
     prompt: str,
     api_key: str,
     bot,
-    typing_fn,
-    api_endpoint = "pygmalion-6b"
+    typing_fn
 ):
     # Set the API endpoint URL
     endpoint = f"http://172.16.85.10:5001/api/latest/generate"

matrix_pygmalion_bot/ai/runpod.py

@@ -21,11 +21,10 @@ async def generate_sync(
     prompt: str,
     api_key: str,
     bot,
-    typing_fn,
-    api_endpoint = "pygmalion-6b"
+    typing_fn
 ):
     # Set the API endpoint URL
-    endpoint = f"https://api.runpod.ai/v2/{api_endpoint}/run"
+    endpoint = f"https://api.runpod.ai/v2/{bot.runpod_text_endpoint}/run"
 
     # Set the headers for the request
     headers = {
@@ -46,11 +45,49 @@ async def generate_sync(
         }
     }
 
+    input_data_oobabooga = {
+        "input": {
+            "data": [json.dumps([
+                prompt,
+                {
+                    'max_new_tokens': min(max_new_tokens, 2048),
+                    'do_sample': True,
+                    'temperature': bot.temperature,
+                    'top_p': 0.73,
+                    'typical_p': 1,
+                    'repetition_penalty': 1.1,
+                    'encoder_repetition_penalty': 1.0,
+                    'top_k': 0,
+                    'min_length': 0,
+                    'no_repeat_ngram_size': 0,
+                    'num_beams': 1,
+                    'penalty_alpha': 0,
+                    'length_penalty': 1,
+                    'early_stopping': False,
+                    'seed': -1,
+                    'add_bos_token': True,
+                    'custom_stopping_strings': [],
+                    'truncation_length': 2048,
+                    'ban_eos_token': False,
+                    'skip_special_tokens': True,
+                }
+            ])]
+        }
+    }
+
+    if bot.runpod_text_endpoint in ['pygmalion-6b', 'gpt-neo-2_7b', 'gpt-neo-1_3b']:
+        api_mode = "runpod"
+    else:
+        api_mode = "oobabooga"
+
     logger.info(f"sending request to runpod.io")
 
     # Make the request
     try:
-        r = requests.post(endpoint, json=input_data, headers=headers, timeout=180)
+        if api_mode == "runpod":
+            r = requests.post(endpoint, json=input_data, headers=headers, timeout=180)
+        else:
+            r = requests.post(endpoint, json=input_data_oobabooga, headers=headers, timeout=180)
     except requests.exceptions.RequestException as e:
         raise ValueError(f"<HTTP ERROR>")
     r_json = r.json()
@@ -62,7 +99,7 @@ async def generate_sync(
         TIMEOUT = 360
         DELAY = 5
         for i in range(TIMEOUT//DELAY):
-            endpoint = f"https://api.runpod.ai/v2/{api_endpoint}/status/{job_id}"
+            endpoint = f"https://api.runpod.ai/v2/{bot.runpod_text_endpoint}/status/{job_id}"
             r = requests.get(endpoint, headers=headers)
             r_json = r.json()
             logger.info(r_json)
@@ -79,7 +116,10 @@ async def generate_sync(
                     text = output[0]
                 else:
                     text = r_json["output"]
-                answer = text.removeprefix(prompt)
+                if api_mode == "runpod":
+                    answer = text.removeprefix(prompt)
+                else:
+                    answer = text["data"][0].removeprefix(prompt)
 #                lines = reply.split('\n')
 #                reply = lines[0].strip()
                 idx = answer.find(f"\nYou:")

matrix_pygmalion_bot/core.py

@@ -86,7 +86,6 @@ class Callbacks(object):
             )
         )
 
-        api_endpoint = "pygmalion-6b"
         await self.client.room_read_markers(room.room_id, event.event_id, event.event_id)
         # Ignore messages when disabled
         if "disabled" in self.bot.room_config[room.room_id] and self.bot.room_config[room.room_id]["disabled"] == True and not event.body.startswith('!start'):
@@ -225,9 +224,6 @@ class Callbacks(object):
             new_answer = event.body.removeprefix('!replace').strip()
             await self.bot.send_message(self.client, room.room_id, new_answer, reply_to=chat_history_item.relates_to_event)
             return
-        elif event.body.startswith('!2'):
-            chat_message.updateText( event.body.removeprefix('!2').strip() )
-            api_endpoint = "ynznznpn6qz6yh"
         elif event.body.startswith('!'):
             await self.bot.send_message(self.client, room.room_id, "<ERROR> UNKNOWN COMMAND")
             return
@@ -258,7 +254,7 @@ class Callbacks(object):
 #        print("")
         try:
             typing = lambda : self.client.room_typing(room.room_id, True, 15000)
-            answer = await self.bot.text_ai.generate_sync(full_prompt, self.bot.runpod_api_key, self.bot, typing, api_endpoint)
+            answer = await self.bot.text_ai.generate_sync(full_prompt, self.bot.runpod_api_key, self.bot, typing)
             answer = answer.strip()
             await self.client.room_typing(room.room_id, False)
             if not (self.bot.translate is None):
@@ -308,6 +304,7 @@ class ChatBot(object):
         self.service_image = "other"
         self.model = "other"
         self.runpod_api_key = None
+        self.runpod_text_endpoint = "pygmalion-6b"
         self.text_ai = None
         self.image_ai = None
 
@@ -552,6 +549,8 @@ async def main() -> None:
             bot.model = config[section]['model']
         if config.has_option('DEFAULT', 'runpod_api_key'):
             bot.runpod_api_key = config['DEFAULT']['runpod_api_key']
+        if config.has_option('DEFAULT', 'runpod_text_endpoint'):
+            bot.runpod_text_endpoint = config['DEFAULT']['runpod_text_endpoint']
         if config.has_option('DEFAULT', 'stablehorde_api_key'):
             bot.stablehorde_api_key = config['DEFAULT']['stablehorde_api_key']
         await bot.read_conf2(section)

runpod/runpod-worker-oobabooga-api/Dockerfile

@@ -7,8 +7,8 @@ FROM ${DEV_IMAGE} as builder
 
 # https://github.com/runpod/containers/blob/main/oobabooga/Dockerfile
 
-# run; DOCKER_BUILDKIT=1 docker build --build-arg MODEL_NAME="PygmalionAI/pygmalion-350m" -t magn418/runpod-oobabooga-pygmalion:test .
-
+# DOCKER_BUILDKIT=1 docker build --build-arg MODEL_NAME="PygmalionAI/pygmalion-350m" -t magn418/runpod-oobabooga-pygmalion:test .
+# docker builder prune
 
 ARG MODEL_NAME="PygmalionAI/pygmalion-350m"
 ENV MODEL_NAME=${MODEL_NAME}
@@ -83,10 +83,14 @@ RUN cd /workspace/text-generation-webui/ && mkdir repositories && cd repositorie
     git clone --branch cuda --single-branch https://github.com/qwopqwop200/GPTQ-for-LLaMa.git && \
     (cd GPTQ-for-LLaMa && python3 setup_cuda.py bdist_wheel -d .)
 # && python3 setup_cuda.py install
+
+
+FROM builder AS modeldownloader
 RUN cd /workspace/text-generation-webui && python3 download-model.py ${MODEL_NAME}
 #RUN git lfs install && \
 #    git clone --depth 1 https://huggingface.co/${MODEL_NAME}
 
+
 FROM ${BASE_IMAGE}
 #ENV TORCH_CUDA_ARCH_LIST="3.5;5.0;6.0;6.1;7.0;7.5;8.0;8.6+PTX"
 RUN mkdir -p /workspace
@@ -116,6 +120,7 @@ RUN --mount=type=cache,target=/root/.cache,sharing=locked pip3 install --upgrade
 
 #RUN mkdir -p /workspace/text-generation-webui/repositories
 COPY --from=builder /workspace/text-generation-webui /workspace/text-generation-webui
+COPY --from=modeldownloader /workspace/text-generation-webui/models /workspace/text-generation-webui/models
 
 RUN --mount=type=cache,target=/root/.cache,sharing=locked \
     cd /workspace/text-generation-webui && pip3 install -r requirements.txt && \