Merge branch 'main' into tgi-rocm

README.md

@@ -1,6 +1,8 @@
 <div align="center">
-
-![image](https://github.com/huggingface/text-generation-inference/assets/3841370/38ba1531-ea0d-4851-b31a-a6d4ddc944b0)
+  
+<a href="https://www.youtube.com/watch?v=jlMAX2Oaht0">
+  <img width=560 width=315 alt="Making TGI deployment optimal" src="https://huggingface.co/datasets/Narsil/tgi_assets/resolve/main/thumbnail.png">
+</a>
 
 # Text Generation Inference
 
@@ -138,6 +140,10 @@ this will impact performance.
 `text-generation-inference` is instrumented with distributed tracing using OpenTelemetry. You can use this feature
 by setting the address to an OTLP collector with the `--otlp-endpoint` argument.
 
+### Architecture
+
+![image](https://github.com/huggingface/text-generation-inference/assets/3841370/38ba1531-ea0d-4851-b31a-a6d4ddc944b0)
+
 ### Local install
 
 You can also opt to install `text-generation-inference` locally.

docs/source/basic_tutorials/preparing_model.md

@@ -4,7 +4,7 @@ Text Generation Inference improves the model in several aspects.
 
 ## Quantization
 
-TGI supports [bits-and-bytes](https://github.com/TimDettmers/bitsandbytes#bitsandbytes), [GPT-Q](https://arxiv.org/abs/2210.17323) and [AWQ](https://arxiv.org/abs/2306.00978) quantization. To speed up inference with quantization, simply set `quantize` flag to `bitsandbytes`, `gptq` or `awq` depending on the quantization technique you wish to use. When using GPT-Q quantization, you need to point to one of the models [here](https://huggingface.co/models?search=gptq) when using AWQ quantization, you need to point to one of the models [here](https://huggingface.co/models?search=awq). To get more information about quantization, please refer to [quantization guide](./../conceptual/quantization.md)
+TGI supports [bits-and-bytes](https://github.com/TimDettmers/bitsandbytes#bitsandbytes), [GPT-Q](https://arxiv.org/abs/2210.17323) and [AWQ](https://arxiv.org/abs/2306.00978) quantization. To speed up inference with quantization, simply set `quantize` flag to `bitsandbytes`, `gptq` or `awq` depending on the quantization technique you wish to use. When using GPT-Q quantization, you need to point to one of the models [here](https://huggingface.co/models?search=gptq) when using AWQ quantization, you need to point to one of the models [here](https://huggingface.co/models?search=awq). To get more information about quantization, please refer to [quantization guide](./../conceptual/quantization)
 
 
 ## RoPE Scaling

server/text_generation_server/models/flash_causal_lm.py

@@ -670,7 +670,7 @@ class FlashCausalLM(Model):
                 self.device,
             )
             _, batch = self.generate_token(batch)
-        except Exception as e:
+        except torch.cuda.OutOfMemoryError as e:
             raise RuntimeError(
                 f"Not enough memory to handle {len(batch.input_ids)} prefill tokens. "
                 f"You need to decrease `--max-batch-prefill-tokens`"