more tests on the remote workers

runpod/runpod-worker-transformers/Dockerfile

@@ -53,9 +53,9 @@ RUN pip3 install --upgrade pip && \
     pip3 install torch torchvision torchaudio --extra-index-url=https://download.pytorch.org/whl/cu118 && \
     pip3 install bitsandbytes && \
     pip3 install safetensors && \
-    pip3 install sentencepiece triton && \
+    pip3 install sentencepiece && \
     pip3 install diffusers && \
-    pip3 install transformers accelerate xformers triton && \
+    pip3 install git+https://github.com/huggingface/transformers accelerate xformers triton && \
     pip3 install huggingface-hub && \
     pip3 install runpod && \
     pip3 cache purge
@@ -86,14 +86,15 @@ RUN pip3 install --upgrade pip && \
     pip3 install torch torchvision torchaudio --extra-index-url=https://download.pytorch.org/whl/cu118 && \
     pip3 install bitsandbytes && \
     pip3 install safetensors && \
-    pip3 install sentencepiece triton && \
+    pip3 install sentencepiece && \
     pip3 install diffusers && \
-    pip3 install transformers accelerate xformers triton && \
+    pip3 install git+https://github.com/huggingface/transformers accelerate xformers triton && \
+#    pip3 install rwkv && \
     pip3 install huggingface-hub && \
     pip3 install runpod && \
     pip3 cache purge
 
-RUN mkdir /workspace/repositories
+RUN mkdir -p /workspace/repositories && mkdir -p /root/.cache/huggingface
 COPY --from=builder /workspace/repositories /workspace/repositories/
 COPY --from=builder /root/.cache/huggingface /root/.cache/huggingface
 

runpod/runpod-worker-transformers/runpod_infer.py

@@ -169,9 +169,9 @@ def load_quantized(model_name, wbits, groupsize, device):
         from safetensors.torch import load_file as safe_load
         if device == -1:
             device = "cpu"
-        model.load_state_dict(safe_load(str(pt_path), device))
+        model.load_state_dict(safe_load(str(pt_path), map_location=device), strict = False)
     else:
-        model.load_state_dict(torch.load(str(pt_path)))
+        model.load_state_dict(torch.load(str(pt_path)), strict = False)
     model.seqlen = 2048
     print('Done.')
 

runpod/serverless-automatic/Dockerfile

@@ -10,8 +10,11 @@ WORKDIR /workspace
 # PFG: https://civitai.com/api/download/models/1316
 # Hassanblend: https://civitai.com/api/download/models/4635
 # Deliberate v2: https://civitai.com/api/download/models/15236
+# ChilloutMix: https://civitai.com/api/download/models/11745
+# AbyssOrangeMix3: https://civitai.com/api/download/models/11811 and VAE https://civitai.com/api/download/models/11811?type=VAE
+# OrangeChillMix: https://civitai.com/api/download/models/13549
 
-RUN wget -O /workspace/stable-diffusion-webui/models/Stable-diffusion/model.safetensors https://civitai.com/api/download/models/4635
+RUN wget -O /workspace/stable-diffusion-webui/models/Stable-diffusion/model.safetensors https://civitai.com/api/download/models/11745
 ## Extra downloads (for Hassanblend)
 RUN wget -O /workspace/stable-diffusion-webui/models/Stable-diffusion/model.vae.safetensors https://huggingface.co/stabilityai/sd-vae-ft-mse-original/resolve/main/vae-ft-mse-840000-ema-pruned.safetensors