update readme for 1.1 (#1)

- update readme for 1.1 (94fd7768eb7044be6c6d7ebf70a8dbaec7b8b1e9)


Co-authored-by: Ewan Pinnington <[email protected]>

README.md

@@ -8,20 +8,17 @@ language:
 library_name: anemoi
 ---
 
-# AIFS Single - v1.0
+# AIFS Single - v1.1
 
 <!-- Provide a quick summary of what the model is/does. -->
 
 Here, we introduce the **Artificial Intelligence Forecasting System (AIFS)**, a data driven forecast
 model developed by the European Centre for Medium-Range Weather Forecasts (ECMWF). 
 
-The release of AIFS Single v1.0 marks the first operationally supported AIFS model. Version 1
-supersedes the existing experimental version, [0.2.1 AIFS-single](https://huggingface.co/ecmwf/aifs-single-0.2.1). 
-The new version, 1.0, brings changes to the AIFS Single model, including among many others:
-
-- Improved performance for upper-level atmospheric variables (AIFS Single still uses 13 pressure-levels, so this improvement mainly refers to 50 and 100 hPa)
-- Improved skill for total precipitation.
-- Additional output variables, including 100 meter winds, snow-fall, surface solar-radiation and land variables such as soil-moisture and soil-temperature.
+The release of AIFS Single v1.1 represents a slight modification to the AIFS model. Version 1.1
+supersedes the existing operational version, [1.1.0 AIFS-single](https://huggingface.co/ecmwf/aifs-single-1.0). 
+The new version, 1.1, brings minor changes to the v1.0 model. These changes mainly correspond to  the removal of 
+spurious rainfall points caused by incorrect soil moisture loss weighting during training of the v1.0 model.
 
 <div style="display: flex; justify-content: center;">
   <img src="assets/radiation_cloudcover.gif" alt="AIFS 10 days Forecast" style="width: 50%;"/>
@@ -33,7 +30,7 @@ are available to the public under ECMWF’s open data policy (https://www.ecmwf.
 Note that due to the non-determinism of GPUs, users will be unable to exactly reproduce an official AIFS forecast 
 when running AIFS Single themselves.
 
-For more details please refer to https://confluence.ecmwf.int/display/FCST/Implementation+of+AIFS+Single+v1
+For more information on this update please see the [confluence page](https://confluence.ecmwf.int/display/FCST/Implementation+of+AIFS+Single+v1)
 
 ## Data Details
 
@@ -102,7 +99,7 @@ There are no changes in resolution compared to previous version AIFS Single v0.2
 
 | | Component | Horizontal Resolution [kms] | Vertical Resolution [levels] |
 |---|:---:|:---:|:---:|
-| Atmosphere | AIFS-single v1.0 | ~ 31 |  13 |
+| Atmosphere | AIFS-single v1.1 | ~ 31 |  13 |
 
 
 ### Model Sources
@@ -138,17 +135,17 @@ step-by-step workflow is specified to run the AIFS using the HuggingFace model:
 🚨  **Note** we train AIFS using `flash_attention` (https://github.com/Dao-AILab/flash-attention).
 The use of 'Flash Attention' package also imposes certain requirements in terms of software and hardware. Those can be found under #Installation and Features in https://github.com/Dao-AILab/flash-attention
 
-🚨 **Note** the `aifs_single_v1.0.ckpt` checkpoint just contains the model’s weights. 
+🚨 **Note** the `aifs_single_v1.1.ckpt` checkpoint just contains the model’s weights. 
 That file does not contain any information about the optimizer states, lr-scheduler states, etc.
 
-## How to train AIFS Single v1.0
+## How to train AIFS Single v1.1
 
 To train this model you can use the configuration files included in this repository and the following Anemoi packages:
 
 ```
-anemoi-training==0.3.1
-anemoi-models==0.4.0
-anemoi-graphs==0.4.4
+anemoi-training==0.4.0
+anemoi-models==0.5.0
+anemoi-graphs==0.5.2
 ```
 and run the pretraining stage as follows,
 
@@ -197,7 +194,7 @@ the forcing variables, like orography, are min-max normalised.
 
 ### Training Procedure
 
-Based on the different experiments we have made - the final training recipe for AIFS Single v1.0 has deviated slightly
+Based on the different experiments we have made - the final training recipe for AIFS Single v1.1 has deviated slightly
 from the one used for AIFS Single v0.2.1 since we found that we could get a well trained model by skipping the ERA5
 rollout and directly doing the rollout on the operational-analysis (extended) dataset. When we say 'extended' we refer 
 to the fact that for AIFS Single v0.2.1 we used just operational-analysis data from 2019 to 2021, while in this new 
@@ -252,7 +249,7 @@ variables. For verification, each system is compared against the operational ECM
 are initialised. In addition, the forecasts are compared against radiosonde observations of geopotential, temperature
 and windspeed, and SYNOP observations of 2 m temperature, 10 m wind and 24 h total precipitation. The definition
 of the metrics, such as ACC (ccaf), RMSE (rmsef) and forecast activity (standard deviation of forecast anomaly,
-sdaf) can be found in e.g Ben Bouallegue et al. ` [2024].
+sdaf) can be found in e.g Ben Bouallegue et al. ` [2024]. No significant changes in skill wer found in the v1.1 model fix.
 
 ### AIFS Single v1.0 vs AIFS Single v0.2.1 (2023)