integrated water consumption calculation with new code on ecologits

data/electricity_mix.csv

@@ -1,4 +1,7 @@
 name,unit,source,WOR,EEE,ZWE,ZMB,ZAF,YEM,VNM,VEN,UZB,URY,USA,UKR,TZA,TWN,TTO,TUR,TUN,TKM,TJK,THA,TGO,SYR,SLV,SEN,SVK,SVN,SGP,SWE,SDN,SAU,RUS,SCG,ROU,QAT,PRY,PRT,POL,PAK,PHL,PER,PAN,OMN,NZL,NPL,NOR,NLD,NIC,NGA,NAM,MOZ,MYS,MEX,MLT,MNG,MMR,MKD,MDA,MAR,LBY,LVA,LUX,LTU,LKA,LBN,KAZ,KWT,KOR,PRK,KHM,KGZ,KEN,JPN,JOR,JAM,ITA,ISL,IRN,IRQ,IND,ISR,IRL,IDN,HUN,HTI,HRV,HND,HKG,GTM,GRC,GIB,GHA,GEO,GBR,GAB,FRA,FIN,ETH,ESP,ERI,EGY,EST,ECU,DZA,DOM,DNK,DEU,CZE,CYP,CUB,CRI,COL,CHN,CMR,CHL,CIV,CHE,COG,COD,CAN,BLR,BWA,BRA,BOL,BRN,BEN,BHR,BGR,BEL,BGD,BIH,AZE,AUS,AUT,ARG,AGO,ANT,ARM,ALB,ARE
 adpe,kg éq. Sb,ADEME Base IMPACTS ®,0.0000000737708,0.0000000642317,0.000000109502,0.000000162193,0.0000000862445,0.0000000163908,0.0000000945573,0.000000112811,0.000000103681,0.000000104586,0.0000000985548,0.0000000647907,0.000000132261,0.0000000578088,0.000000064552,0.0000000749765,0.0000000177021,0.000000131822,0.000000152621,0.0000000569593,0.000000134255,0.0000000268396,0.0000000472135,0.0000000470662,0.0000000606109,0.0000000992283,0.0000000198459,0.0000000777062,0.0000000955701,0.0000000134206,0.0000000960312,0.000000132772,0.0000000981761,0.00000001324,0.000000149181,0.0000000341863,0.000000101946,0.0000000842952,0.0000000595304,0.0000000952688,0.0000000790553,0.0000000374073,0.0000000720474,0.000000238273,0.000000127486,0.0000000329318,0.0000000414983,0.0000000621,0.000000128285,0.000000148382,0.000000044938,0.0000000578358,0.000000049475,0.000000176361,0.000000152699,0.000000119873,0.000000110674,0.0000000641089,0.0000000206592,0.000000153757,0.000000105692,0.0000000294596,0.0000000986932,0.0000000182134,0.000000135386,0.0000000141168,0.0000000518017,0.000000117457,0.0000000319202,0.000000181827,0.0000000958533,0.0000000596578,0.0000000147031,0.0000000196047,0.00000005439,0.0000000781905,0.0000000220304,0.0000000404306,0.000000100099,0.0000000610194,0.0000000219257,0.0000000610451,0.0000000644587,0.0000000937057,0.000000153989,0.0000000649373,0.0000000816213,0.0000000803251,0.0000000691645,0.0000000286211,0.000000156003,0.000000137999,0.0000000370973,0.000000113843,0.0000000485798,0.0000000805114,0.000000174161,0.0000000518326,0.0000000512406,0.000000033925,0.0000000990171,0.000000127168,0.0000000216438,0.0000000429285,0.0000000157411,0.0000000878733,0.0000000817565,0.0000000448771,0.0000000299542,0.0000000863908,0.000000122031,0.0000000851552,0.000000146313,0.000000105851,0.0000000949004,0.000000100467,0.000000265575,0.000000174647,0.0000000993179,0.0000000840478,0.0000000866014,0.00000010962,0.0000000969793,0.0000000185641,0.0000000239702,0.0000000135014,0.0000000823611,0.0000000337201,0.0000000394158,0.000000148007,0.000000092567,0.0000000790846,0.000000141124,0.0000000768612,0.000000124074,0.0000000449103,0.0000000854245,0.000000229556,0.0000000141548
 pe,MJ,ADPf / (1-%renewable_energy),9.988,12.873,16.122,1.798,11.732,16.250,11.238,15.163,17.367,107.570,11.358,12.936,9.305,11.439,14.289,16.150,12.902,23.300,19.165,10.414,21.978,16.989,13.012,14.516,11.680,12.146,10.477,11.026,29.629,14.058,13.200,14.242,15.585,11.916,0.020,14.153,13.178,16.175,11.120,8.211,16.364,22.306,24.731,0.396,4.952,8.511,24.696,11.279,468.150,0.206,12.268,11.775,19.374,15.114,14.132,19.120,18.429,11.702,19.116,8.249,10.128,21.043,12.116,12.341,13.260,12.753,10.199,32.793,34.655,15.380,68.996,10.718,13.677,14.799,12.656,0.013,15.022,20.372,20.363,10.023,10.706,11.603,11.784,20.167,18.548,15.762,,14.340,14.487,,10.097,10.425,13.579,28.341,11.289,11.275,36.133,12.090,13.289,10.195,16.334,20.908,16.376,12.412,16.824,16.260,12.517,13.118,17.317,45.996,7.312,14.119,10.807,11.348,14.783,11.782,34.147,0.097,11.987,13.194,19.642,9.031,11.587,15.689,14.337,14.036,14.375,10.776,12.935,21.705,12.831,16.908,11.036,10.049,16.972,,13.380,0.201,19.032
-gwp,kg éq. CO2,ADEME Base IMPACTS ®,0.590478,0.509427,0.842811,0.0141304,1.17562,1.06777,0.555572,0.497373,0.81178,0.296953,0.67978,0.646745,0.475635,0.845351,0.933059,0.706988,0.80722,1.38296,0.0426743,0.646174,0.545455,1.08778,0.473128,1.1195,0.309341,0.498523,0.655825,0.0464664,1.12472,0.913677,0.66131,1.07808,0.664245,0.722125,0.241601,0.571172,1.15075,0.748727,0.761317,0.284364,0.53403,1.41292,0.293397,0.0841323,0.023754,0.544803,0.941626,0.693123,0.357253,0.00880732,0.832206,0.739214,1.31149,1.47192,0.48193,1.24074,1.04213,0.933694,1.35361,0.234273,0.490016,0.154229,0.709185,0.883627,1.128,0.885084,0.599585,0.797361,1.41054,0.156039,0.589603,0.540891,0.781372,1.07345,0.621329,0.0194609,0.930385,1.48728,1.58299,0.901842,0.648118,0.875394,0.541558,1.3858,0.535759,0.692837,0.95888,0.645801,1.13127,0.977477,0.540126,0.132046,0.602137,0.732511,0.0813225,0.322068,0.251299,0.467803,1.13153,0.587775,1.51492,0.627714,1.02318,0.909252,0.633534,0.657374,0.799077,0.978041,1.28325,0.224471,0.214014,1.05738,0.456622,0.574415,0.928583,0.0448568,0.904053,0.0613102,0.238191,0.795697,2.12572,0.278647,0.585131,0.997577,1.19128,0.804567,0.827087,0.256537,0.795168,1.56025,0.726126,1.12326,0.245573,0.476811,0.412014,1.13161,0.252002,0.0207635,1.09088
+gwp,kg éq. CO2,ADEME Base IMPACTS ®,0.590478,0.509427,0.842811,0.0141304,1.17562,1.06777,0.555572,0.497373,0.81178,0.296953,0.67978,0.646745,0.475635,0.845351,0.933059,0.706988,0.80722,1.38296,0.0426743,0.646174,0.545455,1.08778,0.473128,1.1195,0.309341,0.498523,0.655825,0.0464664,1.12472,0.913677,0.66131,1.07808,0.664245,0.722125,0.241601,0.571172,1.15075,0.748727,0.761317,0.284364,0.53403,1.41292,0.293397,0.0841323,0.023754,0.544803,0.941626,0.693123,0.357253,0.00880732,0.832206,0.739214,1.31149,1.47192,0.48193,1.24074,1.04213,0.933694,1.35361,0.234273,0.490016,0.154229,0.709185,0.883627,1.128,0.885084,0.599585,0.797361,1.41054,0.156039,0.589603,0.540891,0.781372,1.07345,0.621329,0.0194609,0.930385,1.48728,1.58299,0.901842,0.648118,0.875394,0.541558,1.3858,0.535759,0.692837,0.95888,0.645801,1.13127,0.977477,0.540126,0.132046,0.602137,0.732511,0.0813225,0.322068,0.251299,0.467803,1.13153,0.587775,1.51492,0.627714,1.02318,0.909252,0.633534,0.657374,0.799077,0.978041,1.28325,0.224471,0.214014,1.05738,0.456622,0.574415,0.928583,0.0448568,0.904053,0.0613102,0.238191,0.795697,2.12572,0.278647,0.585131,0.997577,1.19128,0.804567,0.827087,0.256537,0.795168,1.56025,0.726126,1.12326,0.245573,0.476811,0.412014,1.13161,0.252002,0.0207635,1.09088
+water_wwf,L/kWh,World Resource Institute Guidance for Calculating Water Use Embedded in Purchased Electricity 2020 ,2216,2705.4,,,,,,,,,386,942,,,,1374,,,,833,,,,,1033,14807,,3777,,,2067,,454,,,1472,216,,,,,,1741,,8285,696,,,,,2472,469,87,,,,,,,7525,1393,3051,,,,,257,,,,,1855,,,3524,,,,2373,,212,1605,1525,,,,,,519,,,,167,,2464,2161,,295,,,765,,,,1374,1419,791,57,,,,606,,458,,8331,,,2850,,,893,,,,,379,621,,,,155,14145,1.291,,,,,0
+water_wcf,L/kWh,World Resource Institute Guidance for Calculating Water Use Embedded in Purchased Electricity 2020,5.04,4.48,,,,,,,,,3.14,2.12,,,,4.92,,,,2.88,,,,,5.79,2.99,,6.02,,,3.44,,7.34,,,9.58,2.5,,,,,,14.95,,6.66,3.44,,,,,1.67,5.3,1.14,,,,,,,2.54,12.79,5.9,,,,,1.89,,,,,2.31,,,4.84,,,,3.44,,1.48,2.27,3.67,,,,,,5.15,,,,2.35,,3.67,4.54,,6.21,,,2.2,,,,3.18,1.93,3.22,1.29,,,,6.02,,10.52,,5.68,,,8.18,,,18.58,,,,,4.62,2.61,,,,4.73,5.72,11.47,,,,,0
+

src/calculator.py

@@ -21,7 +21,7 @@ def calculator_mode():
                 label = 'Provider',
                 options = [x for x in df['provider_clean'].unique()],
                 index = 7
-            )
+            ) #une liste de proviseurs à selectioner
 
         with col2:
             model = st.selectbox(
@@ -45,28 +45,28 @@ def calculator_mode():
         if df_filtered['warning_arch'].values[0] and df_filtered['warning_multi_modal'].values[0]:
             st.warning(WARNING_BOTH)
             
-    try:
+    #try:
 
-        impacts = llm_impacts(
-                        provider=provider_raw,
-                        model_name=model_raw,
-                        output_token_count=[x[1] for x in PROMPTS if x[0] == output_tokens][0],
-                        request_latency=100000
-                    )
+    impacts = llm_impacts(
+                    provider=provider_raw,
+                    model_name=model_raw,
+                    output_token_count=[x[1] for x in PROMPTS if x[0] == output_tokens][0],
+                    request_latency=100000
+                )
 
-        impacts, _, _ = format_impacts(impacts)
-        
-        with st.container(border=True):
+    impacts, _, _ = format_impacts(impacts)
+    
+    with st.container(border=True):
 
-            st.markdown('<h3 align = "center">Environmental impacts</h3>', unsafe_allow_html=True)
-            st.markdown('<p align = "center">To understand how the environmental impacts are computed go to the 📖 Methodology tab.</p>', unsafe_allow_html=True)
-            display_impacts(impacts)
+        st.markdown('<h3 align = "center">Environmental impacts</h3>', unsafe_allow_html=True)
+        st.markdown('<p align = "center">To understand how the environmental impacts are computed go to the 📖 Methodology tab.</p>', unsafe_allow_html=True)
+        display_impacts(impacts, provider, location="🌎 World")
+    
+    with st.container(border=True):
         
-        with st.container(border=True):
-            
-            st.markdown('<h3 align = "center">That\'s equivalent to ...</h3>', unsafe_allow_html=True)
-            st.markdown('<p align = "center">Making this request to the LLM is equivalent to the following actions :</p>', unsafe_allow_html=True)
-            display_equivalent(impacts)
+        st.markdown('<h3 align = "center">That\'s equivalent to ...</h3>', unsafe_allow_html=True)
+        st.markdown('<p align = "center">Making this request to the LLM is equivalent to the following actions :</p>', unsafe_allow_html=True)
+        display_equivalent(impacts, provider, location="🌎 World")
             
-    except Exception as e:
-        st.error('Could not find the model in the repository. Please try another model.')
+    # except Exception as e:
+    #     st.error('Could not find the model in the repository. Please try another model.')

src/content.py

@@ -72,7 +72,7 @@ ABOUT_TEXT = r"""
 The rapid evolution of generative AI is reshaping numerous industries and aspects of our daily lives. While these 
 advancements offer some benefits, they also **pose substantial environmental challenges that cannot be overlooked**. 
 Plus the issue of AI's environmental footprint has been mainly discussed at training stage but rarely at the inference 
-stage. That is an issue because **inference impacts for LLMs can largely overcome the training impacts when deployed 
+stage. That is an issue because **inference impacts for large langauge models (LLMs) can largely overcome the training impacts when deployed 
 at large scales**.
 At **[GenAI Impact](https://genai-impact.org/) we are dedicated to understanding and mitigating the environmental 
 impacts of generative AI** through rigorous research, innovative tools, and community engagement. Especially, in early 
@@ -146,38 +146,44 @@ Leaderboard.
 For general question on the project, please use the [GitHub thread](https://github.com/genai-impact/ecologits/discussions/45). 
 Otherwise use our contact form on [genai-impact.org/contact](https://genai-impact.org/contact/).
 """
-
+#TODO ajoute des mots sur water use. embodied vs per request use
+#TODO ajuster la partie sur embodied electricité. c'est pas du tout clair
 
 METHODOLOGY_TEXT = r"""
 ## 📖 Methodology
 We have developed a methodology to **estimate the energy consumption and environmental impacts for an LLM inference** 
 based on request parameters and hypotheses on the data center location, the hardware used, the model architecture and
 more.
-In this section we will only cover the principles of the methodology related to the 🧮 **EcoLogits Calculator**. If 
-you wish to learn more on the environmental impacts modeling of an LLM request checkout the 
+In this section, we will only cover the principles of the methodology related to the 🧮 **EcoLogits Calculator**. If 
+you would like to learn more about the environmental impacts modeling of an LLM request, consider checking out the 
 🌱 [EcoLogits documentation page](https://ecologits.ai/methodology/).
 ### Modeling impacts of an LLM request
-The environmental impacts of an LLM inference are split into the **usage impacts** $I_{request}^u$ to account for 
-electricity consumption and the **embodied impacts** $I_{request}^e$ that relates to resource extraction, hardware 
-manufacturing and transportation. In general terms it can be expressed as follow:
+The environmental impacts of an LLM inference are split into **usage impacts** $I_{request}^u$ to account for 
+electricity consumption and **embodied impacts** $I_{request}^e$ that relate to resource extraction, hardware 
+manufacturing and transportation. It can be expressed as the following:
 $$ I_{request} = I_{request}^u  + I_{request}^e $$
 $$ I_{request} = E_{request}*F_{em}+\frac{\Delta T}{\Delta L}*I_{server}^e $$
-With,
+Where
 * $E_{request}$ the estimated energy consumption of the server and its cooling system.
 * $F_{em}$ the electricity mix that depends on the country and time.
 * $\frac{\Delta T}{\Delta L}$ the hardware usage ratio i.e. the computation time over the lifetime of the hardware.
 * $I_{server}^e$ the embodied impacts of the server.
-Additionally, to ⚡️ **direct energy consumption** the environmental impacts are expressed in **three dimensions 
-(multi-criteria impacts)** that are:
-* 🌍 **Global Warming Potential** (GWP): Potential impact on global warming in kgCO2eq (commonly known as GHG/carbon 
-emissions).
-* 🪨 **Abiotic Depletion Potential for Elements** (ADPe): Impact on the depletion of non-living resources such as 
-minerals or metals in kgSbeq.
-* ⛽️ **Primary Energy** (PE): Total energy consumed from primary sources in MJ.
+
+Additionally, the environmental impacts are expressed in **four dimensions 
+(multi-criteria impacts)**:
+* 🌍 **Global Warming Potential** (GWP): Potential impact on Climate Change in kilograms of CO2 equivalent, 
+or kgCO2eq, also commonly known as green house gases (GHG) or carbon emissions.
+* 🪨 **Abiotic Depletion Potential for Elements** (ADPe): Mesures the impact on the depletion of non-organic resources such as 
+minerals or metals in kilograms of antimony equivalent. This is to say the impact equating to that amount of antimony extracted.
+* ⛽️ **Primary Energy** (PE): Total energy consumed from primary sources in megajoules.
+* 💧 **Water Use** : Water consumption from this request. 
+
+Again, if you are interested in seeing the detailed calculations we employ, please consider checking out the methodology page on our website for [ecologits]https://ecologits.ai/latest/methodology/llm_inference/)!
+
 ### Principles, Data and Hypotheses
-We use a **bottom-up methodology** to model impacts, meaning that we will estimate the impacts of low-level physical 
-components to then estimate the impacts at software level (in that case an LLM inference). We also rely on **Life 
-Cycle Approach (LCA) proxies and approach** to model both usage and embodied phases with multi-criteria impacts. 
+We use a **bottom-up methodology** to model impacts, meaning that we estimate the impacts of low-level physical 
+components to then estimate the impacts at software level (in our case an LLM inference). We also rely on **Life 
+Cycle Approach (LCA) proxies** to model both usage and embodied phases with multi-criteria impacts. 
 If you are interested in this approach we recommend you to read the following [Boavizta](https://boavizta.org/) 
 resources.
 * [Digital & environment: How to evaluate server manufacturing footprint, beyond greenhouse gas emissions?](https://boavizta.org/en/blog/empreinte-de-la-fabrication-d-un-serveur) 
@@ -189,14 +195,16 @@ and latency based on the model architecture and number of output tokens.
 * [Boavizta API](https://github.com/Boavizta/boaviztapi) to estimate server embodied impacts and base energy 
 consumption.
 * [ADEME Base Empreinte®](https://base-empreinte.ademe.fr/) for electricity mix impacts per country.
+* [World Resource Institute](https://www.wri.org/research/guidance-calculating-water-use-embedded-purchased-electricity) for the water withdrawal factor (WWF) and water consumption factor (WCF) for different countries. 
+* [Various sources from companies' sustainability reports](https://docs.google.com/spreadsheets/d/1XvKNhqJJ3e0wlUlSOcv-HS8jIwjCJBEQYgNhkVwA7ac/edit?usp=sharing) for their on-site water usage efficiencies and power usage efficiencies.
 Finally here are the **main hypotheses** we have made to compute the impacts.
 * ⚠️ **We *"guesstimate"* the model architecture of proprietary LLMs when not disclosed by the provider.** 
 * Production setup: quantized models running on data center grade servers and GPUs such as A100.
 * Electricity mix does not depend on time (help us enhance EcoLogits and work on this [issue](https://github.com/genai-impact/ecologits/issues/42))
-* Ignore the following impacts: unused cloud resources, data center building, network and end-user devices... (for now)
+* We are ignoring the following impacts: unused cloud resources, data center building, network and end-user devices... (for now)
 ## Equivalents
 We have integrated impact equivalents to help people better understand the impacts and have reference points for 
-standard use cases and everyday activities.
+standard use cases in everyday activities.
 ### Request impacts
 These equivalents are computed based on the request impacts only.
 #### 🚶‍♂️‍➡️ Walking or 🏃‍♂️‍➡️ running distance
@@ -206,8 +214,8 @@ physical activity (for someone weighing 70kg):
 * 🚶‍♂️‍➡️ walking: $ 196\ kJ/km $ (speed of $ 3\ km/h $)
 * 🏃‍♂️‍➡️ running: $ 294\ kJ/km $ (speed of $ 10\ km/h $)
 We divide the request energy consumption by these values to compute the distance traveled. 
-#### 🔋 Electric Vehicle distance
-We compare the ⚡️ direct energy consumption with the energy consumer by a EV car. From 
+#### 🔋 Electric vehicle distance
+We compare the ⚡️ direct energy consumption with the energy consumer by an EV car. From 
 [selectra.info](https://selectra.info/energie/actualites/insolite/consommation-vehicules-electriques-france-2040) or 
 [tesla.com](https://www.tesla.com/fr_fr/support/power-consumption) we consider an average value of energy consumed per 
 kilometer of: $ 0.17\ kWh/km $.
@@ -217,43 +225,47 @@ We compare the 🌍 GHG emissions of the request and of streaming a video. From
 [impactco2.fr](https://impactco2.fr/outils/comparateur?value=1&comparisons=streamingvideo), we consider that 
 $ 1\ kgCO2eq $ is equivalent to $ 15.6\ h $ of streaming.
 We multiply that value by the GHG emissions of the request to get an equivalent in hours of video streaming.
+#### 🚰 Bottled waters
+We compare the water consumption to a standard 75cL water bottle. 
 ### Scaled impacts
 These equivalents are computed based on the request impacts scaled to a worldwide adoption use case. We imply that the
-same request is done 1% of the planet everyday for 1 year, and then compute impact equivalents.
+same request is done by 1% of the world population everyday for 1 year, and then compute impact equivalents.
 $$
-I_{scaled} = I_{request} * [1 \\% \ \text{of}\ 8B\ \text{people on earth}] * 365\ \text{days}
+I_{scaled} = I_{request} * [1 \% \ \text{of 8 billion people on earth}] * 365\ \text{days}
 $$
 #### Number of 💨 wind turbines or ☢️ nuclear plants
-We compare the ⚡️ direct energy consumption (scaled) by the energy production of wind turbines and nuclear power 
+We compare the ⚡️ scaled direct energy consumption by the energy production of wind turbines and nuclear power 
 plants. From [ecologie.gouv.fr](https://www.ecologie.gouv.fr/eolien-terrestre) we consider that a $ 2\ MW $ wind 
 turbine produces $ 4.2\ GWh $ a year. And from [edf.fr](https://www.edf.fr/groupe-edf/espaces-dedies/jeunes-enseignants/pour-les-jeunes/lenergie-de-a-a-z/produire-de-lelectricite/le-nucleaire-en-chiffres) 
 we learn that a $ 900\ MW $ nuclear power plant produces $ 6\ TWh $ a year.
 We divide the scaled energy consumption by these values to get the number of wind turbines or nuclear power plants 
 needed.
 #### Multiplier of 🇮🇪 Ireland electricity consumption
-We compare the ⚡️ direct energy consumption (scaled) by the electricity consumption of Ireland per year. From 
+We compare the ⚡️ scaled direct energy consumption by the electricity consumption of Ireland per year. From 
 [wikipedia.org](https://en.wikipedia.org/wiki/List_of_countries_by_electricity_consumption) we consider the Ireland 
 electricity consumption to be $ 33\ TWh $ a year for a population of 5M.
-We divide the scaled energy consumption by this value to get the equivalent number of "Ireland countries".
+We divide the scaled energy consumption by this value to get the equivalent number of "Irelands".
 #### Number of ✈️ Paris ↔ New York City flights
-We compare the 🌍 GHG emissions (scaled) of the request and of a return flight Paris ↔ New York City. From 
+We compare the 🌍 scaled GHG emissions of the request and of a return flight Paris ↔ New York City. From 
 [impactco2.fr](https://impactco2.fr/outils/comparateur?value=1&comparisons=&equivalent=avion-pny) we consider that a 
 return flight Paris → New York City → Paris for one passenger emits $ 1,770\ kgCO2eq $ and we consider an overall 
 average load of 100 passengers per flight.
 We divide the scaled GHG emissions by this value to get the equivalent number of return flights.
-
+#### Number of Olympic-sized swimming pools 🏊🏼
+We compare the scaled water consumption to the number of Olympic-sized swimming pools it can fill. According to the [Phinizy Center for Water Sciences](https://phinizycenter.org/olympic-swimming-pools/), 
+an Olympic-sized swimming pool holds about 2.5 million liters of water.
 #### If you are motivated to help us test and enhance this methodology [contact us](https://genai-impact.org/contact/)! 💪
 """
 
 CITATION_LABEL = "BibTeX citation for EcoLogits Calculator and the EcoLogits library:"
 CITATION_TEXT = r"""@misc{ecologits-calculator,
-  author={Samuel Rincé, Adrien Banse and Valentin Defour},
+  author={Samuel Rincé, Adrien Banse, Valentin Defour, and Chieh Hsu},
   title={EcoLogits Calculator},
   year={2025},
   howpublished= {\url{https://huggingface.co/spaces/genai-impact/ecologits-calculator}},
 }
 @software{ecologits,
-  author = {Samuel Rincé, Adrien Banse, Vinh Nguyen and Luc Berton},
+  author = {Samuel Rincé, Adrien Banse, Vinh Nguyen, Luc Berton, and Chieh Hsu},
   publisher = {GenAI Impact},
   title = {EcoLogits: track the energy consumption and environmental footprint of using generative AI models through APIs.},
 }"""

src/data/electricity_mix.csv

@@ -1,4 +1,7 @@
 name,unit,source,WOR,EEE,ZWE,ZMB,ZAF,YEM,VNM,VEN,UZB,URY,USA,UKR,TZA,TWN,TTO,TUR,TUN,TKM,TJK,THA,TGO,SYR,SLV,SEN,SVK,SVN,SGP,SWE,SDN,SAU,RUS,SCG,ROU,QAT,PRY,PRT,POL,PAK,PHL,PER,PAN,OMN,NZL,NPL,NOR,NLD,NIC,NGA,NAM,MOZ,MYS,MEX,MLT,MNG,MMR,MKD,MDA,MAR,LBY,LVA,LUX,LTU,LKA,LBN,KAZ,KWT,KOR,PRK,KHM,KGZ,KEN,JPN,JOR,JAM,ITA,ISL,IRN,IRQ,IND,ISR,IRL,IDN,HUN,HTI,HRV,HND,HKG,GTM,GRC,GIB,GHA,GEO,GBR,GAB,FRA,FIN,ETH,ESP,ERI,EGY,EST,ECU,DZA,DOM,DNK,DEU,CZE,CYP,CUB,CRI,COL,CHN,CMR,CHL,CIV,CHE,COG,COD,CAN,BLR,BWA,BRA,BOL,BRN,BEN,BHR,BGR,BEL,BGD,BIH,AZE,AUS,AUT,ARG,AGO,ANT,ARM,ALB,ARE
-adpe,kg éq. Sb,ADEME Base IMPACTS ®,0.0000000737708,0.0000000642317,0.000000109502,0.000000162193,0.0000000862445,0.0000000163908,0.0000000945573,0.000000112811,0.000000103681,0.000000104586,0.0000000985548,0.0000000647907,0.000000132261,0.0000000578088,0.000000064552,0.0000000749765,0.0000000177021,0.000000131822,0.000000152621,0.0000000569593,0.000000134255,0.0000000268396,0.0000000472135,0.0000000470662,0.0000000606109,0.0000000992283,0.0000000198459,0.0000000777062,0.0000000955701,0.0000000134206,0.0000000960312,0.000000132772,0.0000000981761,0.00000001324,0.000000149181,0.0000000341863,0.000000101946,0.0000000842952,0.0000000595304,0.0000000952688,0.0000000790553,0.0000000374073,0.0000000720474,0.000000238273,0.000000127486,0.0000000329318,0.0000000414983,0.0000000621,0.000000128285,0.000000148382,0.000000044938,0.0000000578358,0.000000049475,0.000000176361,0.000000152699,0.000000119873,0.000000110674,0.0000000641089,0.0000000206592,0.000000153757,0.000000105692,0.0000000294596,0.0000000986932,0.0000000182134,0.000000135386,0.0000000141168,0.0000000518017,0.000000117457,0.0000000319202,0.000000181827,0.0000000958533,0.0000000596578,0.0000000147031,0.0000000196047,0.00000005439,0.0000000781905,0.0000000220304,0.0000000404306,0.000000100099,0.0000000610194,0.0000000219257,0.0000000610451,0.0000000644587,0.0000000937057,0.000000153989,0.0000000649373,0.0000000816213,0.0000000803251,0.0000000691645,0.0000000286211,0.000000156003,0.000000137999,0.0000000370973,0.000000113843,0.0000000485798,0.0000000805114,0.000000174161,0.0000000518326,0.0000000512406,0.000000033925,0.0000000990171,0.000000127168,0.0000000216438,0.0000000429285,0.0000000157411,0.0000000878733,0.0000000817565,0.0000000448771,0.0000000299542,0.0000000863908,0.000000122031,0.0000000851552,0.000000146313,0.000000105851,0.0000000949004,0.000000100467,0.000000265575,0.000000174647,0.0000000993179,0.0000000840478,0.0000000866014,0.00000010962,0.0000000969793,0.0000000185641,0.0000000239702,0.0000000135014,0.0000000823611,0.0000000337201,0.0000000394158,0.000000148007,0.000000092567,0.0000000790846,0.000000141124,0.0000000768612,0.000000124074,0.0000000449103,0.0000000854245,0.000000229556,0.0000000141548
+adpe,kg eq. Sb,ADEME Base IMPACTS ®,0.0000000737708,0.0000000642317,0.000000109502,0.000000162193,0.0000000862445,0.0000000163908,0.0000000945573,0.000000112811,0.000000103681,0.000000104586,0.0000000985548,0.0000000647907,0.000000132261,0.0000000578088,0.000000064552,0.0000000749765,0.0000000177021,0.000000131822,0.000000152621,0.0000000569593,0.000000134255,0.0000000268396,0.0000000472135,0.0000000470662,0.0000000606109,0.0000000992283,0.0000000198459,0.0000000777062,0.0000000955701,0.0000000134206,0.0000000960312,0.000000132772,0.0000000981761,0.00000001324,0.000000149181,0.0000000341863,0.000000101946,0.0000000842952,0.0000000595304,0.0000000952688,0.0000000790553,0.0000000374073,0.0000000720474,0.000000238273,0.000000127486,0.0000000329318,0.0000000414983,0.0000000621,0.000000128285,0.000000148382,0.000000044938,0.0000000578358,0.000000049475,0.000000176361,0.000000152699,0.000000119873,0.000000110674,0.0000000641089,0.0000000206592,0.000000153757,0.000000105692,0.0000000294596,0.0000000986932,0.0000000182134,0.000000135386,0.0000000141168,0.0000000518017,0.000000117457,0.0000000319202,0.000000181827,0.0000000958533,0.0000000596578,0.0000000147031,0.0000000196047,0.00000005439,0.0000000781905,0.0000000220304,0.0000000404306,0.000000100099,0.0000000610194,0.0000000219257,0.0000000610451,0.0000000644587,0.0000000937057,0.000000153989,0.0000000649373,0.0000000816213,0.0000000803251,0.0000000691645,0.0000000286211,0.000000156003,0.000000137999,0.0000000370973,0.000000113843,0.0000000485798,0.0000000805114,0.000000174161,0.0000000518326,0.0000000512406,0.000000033925,0.0000000990171,0.000000127168,0.0000000216438,0.0000000429285,0.0000000157411,0.0000000878733,0.0000000817565,0.0000000448771,0.0000000299542,0.0000000863908,0.000000122031,0.0000000851552,0.000000146313,0.000000105851,0.0000000949004,0.000000100467,0.000000265575,0.000000174647,0.0000000993179,0.0000000840478,0.0000000866014,0.00000010962,0.0000000969793,0.0000000185641,0.0000000239702,0.0000000135014,0.0000000823611,0.0000000337201,0.0000000394158,0.000000148007,0.000000092567,0.0000000790846,0.000000141124,0.0000000768612,0.000000124074,0.0000000449103,0.0000000854245,0.000000229556,0.0000000141548
 pe,MJ,ADPf / (1-%renewable_energy),9.988,12.873,16.122,1.798,11.732,16.250,11.238,15.163,17.367,107.570,11.358,12.936,9.305,11.439,14.289,16.150,12.902,23.300,19.165,10.414,21.978,16.989,13.012,14.516,11.680,12.146,10.477,11.026,29.629,14.058,13.200,14.242,15.585,11.916,0.020,14.153,13.178,16.175,11.120,8.211,16.364,22.306,24.731,0.396,4.952,8.511,24.696,11.279,468.150,0.206,12.268,11.775,19.374,15.114,14.132,19.120,18.429,11.702,19.116,8.249,10.128,21.043,12.116,12.341,13.260,12.753,10.199,32.793,34.655,15.380,68.996,10.718,13.677,14.799,12.656,0.013,15.022,20.372,20.363,10.023,10.706,11.603,11.784,20.167,18.548,15.762,,14.340,14.487,,10.097,10.425,13.579,28.341,11.289,11.275,36.133,12.090,13.289,10.195,16.334,20.908,16.376,12.412,16.824,16.260,12.517,13.118,17.317,45.996,7.312,14.119,10.807,11.348,14.783,11.782,34.147,0.097,11.987,13.194,19.642,9.031,11.587,15.689,14.337,14.036,14.375,10.776,12.935,21.705,12.831,16.908,11.036,10.049,16.972,,13.380,0.201,19.032
-gwp,kg éq. CO2,ADEME Base IMPACTS ®,0.590478,0.509427,0.842811,0.0141304,1.17562,1.06777,0.555572,0.497373,0.81178,0.296953,0.67978,0.646745,0.475635,0.845351,0.933059,0.706988,0.80722,1.38296,0.0426743,0.646174,0.545455,1.08778,0.473128,1.1195,0.309341,0.498523,0.655825,0.0464664,1.12472,0.913677,0.66131,1.07808,0.664245,0.722125,0.241601,0.571172,1.15075,0.748727,0.761317,0.284364,0.53403,1.41292,0.293397,0.0841323,0.023754,0.544803,0.941626,0.693123,0.357253,0.00880732,0.832206,0.739214,1.31149,1.47192,0.48193,1.24074,1.04213,0.933694,1.35361,0.234273,0.490016,0.154229,0.709185,0.883627,1.128,0.885084,0.599585,0.797361,1.41054,0.156039,0.589603,0.540891,0.781372,1.07345,0.621329,0.0194609,0.930385,1.48728,1.58299,0.901842,0.648118,0.875394,0.541558,1.3858,0.535759,0.692837,0.95888,0.645801,1.13127,0.977477,0.540126,0.132046,0.602137,0.732511,0.0813225,0.322068,0.251299,0.467803,1.13153,0.587775,1.51492,0.627714,1.02318,0.909252,0.633534,0.657374,0.799077,0.978041,1.28325,0.224471,0.214014,1.05738,0.456622,0.574415,0.928583,0.0448568,0.904053,0.0613102,0.238191,0.795697,2.12572,0.278647,0.585131,0.997577,1.19128,0.804567,0.827087,0.256537,0.795168,1.56025,0.726126,1.12326,0.245573,0.476811,0.412014,1.13161,0.252002,0.0207635,1.09088
+gwp,kg eq. CO2,ADEME Base IMPACTS ®,0.590478,0.509427,0.842811,0.0141304,1.17562,1.06777,0.555572,0.497373,0.81178,0.296953,0.67978,0.646745,0.475635,0.845351,0.933059,0.706988,0.80722,1.38296,0.0426743,0.646174,0.545455,1.08778,0.473128,1.1195,0.309341,0.498523,0.655825,0.0464664,1.12472,0.913677,0.66131,1.07808,0.664245,0.722125,0.241601,0.571172,1.15075,0.748727,0.761317,0.284364,0.53403,1.41292,0.293397,0.0841323,0.023754,0.544803,0.941626,0.693123,0.357253,0.00880732,0.832206,0.739214,1.31149,1.47192,0.48193,1.24074,1.04213,0.933694,1.35361,0.234273,0.490016,0.154229,0.709185,0.883627,1.128,0.885084,0.599585,0.797361,1.41054,0.156039,0.589603,0.540891,0.781372,1.07345,0.621329,0.0194609,0.930385,1.48728,1.58299,0.901842,0.648118,0.875394,0.541558,1.3858,0.535759,0.692837,0.95888,0.645801,1.13127,0.977477,0.540126,0.132046,0.602137,0.732511,0.0813225,0.322068,0.251299,0.467803,1.13153,0.587775,1.51492,0.627714,1.02318,0.909252,0.633534,0.657374,0.799077,0.978041,1.28325,0.224471,0.214014,1.05738,0.456622,0.574415,0.928583,0.0448568,0.904053,0.0613102,0.238191,0.795697,2.12572,0.278647,0.585131,0.997577,1.19128,0.804567,0.827087,0.256537,0.795168,1.56025,0.726126,1.12326,0.245573,0.476811,0.412014,1.13161,0.252002,0.0207635,1.09088
+water_wwf,L/kWh,World Resource Institute Guidance for Calculating Water Use Embedded in Purchased Electricity 2020 ,2216,2705.4,,,,,,,,,386,942,,,,1374,,,,833,,,,,1033,14807,,3777,,,2067,,454,,,1472,216,,,,,,1741,,8285,696,,,,,2472,469,87,,,,,,,7525,1393,3051,,,,,257,,,,,1855,,,3524,,,,2373,,212,1605,1525,,,,,,519,,,,167,,2464,2161,,295,,,765,,,,1374,1419,791,57,,,,606,,458,,8331,,,2850,,,893,,,,,379,621,,,,155,14145,1.291,,,,,0
+water_wcf,L/kWh,World Resource Institute Guidance for Calculating Water Use Embedded in Purchased Electricity 2020,5.04,4.48,,,,,,,,,3.14,2.12,,,,4.92,,,,2.88,,,,,5.79,2.99,,6.02,,,3.44,,7.34,,,9.58,2.5,,,,,,14.95,,6.66,3.44,,,,,1.67,5.3,1.14,,,,,,,2.54,12.79,5.9,,,,,1.89,,,,,2.31,,,4.84,,,,3.44,,1.48,2.27,3.67,,,,,,5.15,,,,2.35,,3.67,4.54,,6.21,,,2.2,,,,3.18,1.93,3.22,1.29,,,,6.02,,10.52,,5.68,,,8.18,,,18.58,,,,,4.62,2.61,,,,4.73,5.72,11.47,,,,,0
+

src/electricity_mix.py

@@ -2,6 +2,8 @@ from csv import DictReader
 import pandas as pd
 
 PATH = "src/data/electricity_mix.csv"
+#ATTENTION ! Il a deux fichiers avec le même nom, un dans data/electricity_mix.csv et l'autre dans src/data/electricity_mix.csv
+#vérifie que je modifie le bon fichier
 
 COUNTRY_CODES = [
     ("🌎 World", "WOR"),
@@ -145,7 +147,8 @@ COUNTRY_CODES = [
     ("🇦🇪 United Arab Emirates", "ARE")
 ]
 
-
+#faut que j'acquris plus de données sur ça. on manque des données sur beacoup de pays 
+#pour les pays qui manquent des centres de données, comme beaucoup en Afrique, qui mène leurs requêts ? 
 def find_electricity_mix(code: str):
     # TODO: Maybe more optimal to construct database at the beginning of the app
     #       in the same fashion as find_model
@@ -153,7 +156,7 @@ def find_electricity_mix(code: str):
     with open(PATH) as fd:
         csv = DictReader(fd)
         for row in csv:
-            res += [float(row[code])]
+            res += [row[code]]
     return res
 
 def dataframe_electricity_mix(countries: list):
@@ -172,4 +175,5 @@ def dataframe_electricity_mix(countries: list):
                         index='country',
                         values='value')
 
-    return df
+    return df
+

src/expert.py

@@ -32,6 +32,8 @@ def expert_mode():
                 key = 1
             )
 
+        provider_selected = st.session_state[1].lower()
+
         with col2:
             model_exp = st.selectbox(
                 label = 'Model',
@@ -79,15 +81,40 @@ def expert_mode():
 
         location = st.selectbox('Location', [x[0] for x in COUNTRY_CODES])
 
-        col4, col5, col6 = st.columns(3)
+        col4, col5 = st.columns(2)
 
         with col4:
-            mix_gwp = st.number_input('Electricity mix - GHG emissions [kgCO2eq / kWh]', find_electricity_mix([x[1] for x in COUNTRY_CODES if x[0] ==location][0])[2], format="%0.6f")
+            try:
+                mix_gwp = st.number_input('Electricity mix - GHG emissions [kgCO2eq / kWh]', float(find_electricity_mix([x[1] for x in COUNTRY_CODES if x[0] ==location][0])[2]), format="%0.6f")
             #disp_ranges = st.toggle('Display impact ranges', False)
+
+            except: 
+                mix_gwp = st.number_input('Electricity mix - GHG emissions [kgCO2eq / kWh]', float(find_electricity_mix(["WOR"][0])[2]), format="%0.6f")
+                st.warning(f"Lacking data on {location}, showing global average data.")
+
         with col5:
-            mix_adpe = st.number_input('Electricity mix - Abiotic resources [kgSbeq / kWh]', find_electricity_mix([x[1] for x in COUNTRY_CODES if x[0] ==location][0])[0], format="%0.13f")
+            try:
+                mix_adpe = st.number_input('Electricity mix - Abiotic resources [kgSbeq / kWh]', float(find_electricity_mix([x[1] for x in COUNTRY_CODES if x[0] ==location][0])[0]), format="%0.13f")
+            except:
+                mix_adpe = st.number_input('Electricity mix - Abiotic resources [kgSbeq / kWh]', float(find_electricity_mix(["WOR"][0])[0]), format="%0.13f")
+                st.warning(f"Lacking data on {location}, showing global average data.")
+
+        col6, col7 = st.columns(2)
+
         with col6:
-            mix_pe = st.number_input('Electricity mix - Primary energy [MJ / kWh]', find_electricity_mix([x[1] for x in COUNTRY_CODES if x[0] ==location][0])[1], format="%0.3f")
+            try: 
+                mix_pe = st.number_input('Electricity mix - Primary energy [MJ / kWh]', float(find_electricity_mix([x[1] for x in COUNTRY_CODES if x[0] ==location][0])[1]), format="%0.3f")
+            except: 
+                mix_pe = st.number_input('Electricity mix - Primary energy [MJ / kWh]', float(find_electricity_mix(["WOR"][0])[1]), format="%0.3f")
+                st.warning(f"Lacking data on {location}, showing global average data.")
+
+        with col7:
+            try: 
+                mix_water = st.number_input('Electricity mix - Water consumption factor [L / kWh]', float(find_electricity_mix([x[1] for x in COUNTRY_CODES if x[0] ==location][0])[4]), format="%0.3f")
+            except: 
+                mix_water = st.number_input('Electricity mix - Water consumption factor [L / kWh]', float(find_electricity_mix(["WOR"][0])[4]), format="%0.3f")
+                st.warning(f"Lacking data on {location}, showing global average data.")
+
 
     impacts = compute_llm_impacts(model_active_parameter_count=active_params,
                 model_total_parameter_count=total_params,
@@ -95,7 +122,9 @@ def expert_mode():
                 request_latency=100000,
                 if_electricity_mix_gwp=mix_gwp,
                 if_electricity_mix_adpe=mix_adpe,
-                if_electricity_mix_pe=mix_pe
+                if_electricity_mix_pe=mix_pe,
+                if_electricity_mix_wcf=mix_water,
+                provider = provider_selected
             )
     
     impacts, usage, embodied = format_impacts(impacts)
@@ -104,7 +133,7 @@ def expert_mode():
 
         st.markdown('<h3 align="center">Environmental Impacts</h2>', unsafe_allow_html = True)
 
-        display_impacts(impacts)
+        display_impacts(impacts, provider_exp, location)
 
     with st.expander('⚖️ Usage vs Embodied'):
 
@@ -121,9 +150,11 @@ def expert_mode():
             names = ['usage', 'embodied'],
             title = 'GHG emissions',
             color_discrete_sequence=["#00BF63", "#0B3B36"],
-            width = 100
+            width = 400
             )
-            fig_gwp.update_layout(showlegend=False, title_x=0.5)
+            fig_gwp.update_layout(
+                showlegend=False, 
+                title_x=0.25)
 
             st.plotly_chart(fig_gwp)
 
@@ -133,10 +164,11 @@ def expert_mode():
                 names = ['usage', 'embodied'],
                 title = 'Abiotic depletion',
                 color_discrete_sequence=["#0B3B36","#00BF63"],
-                width = 100)
+                width = 400
+                )
             fig_adpe.update_layout(
                 showlegend=False,
-                title_x=0.5)
+                title_x=0.25)
             
             st.plotly_chart(fig_adpe)
 
@@ -146,8 +178,11 @@ def expert_mode():
                 names = ['usage', 'embodied'],
                 title = 'Primary energy',
                 color_discrete_sequence=["#00BF63", "#0B3B36"],
-                width = 100)
-            fig_pe.update_layout(showlegend=False, title_x=0.5)
+                width = 400
+                )
+            fig_pe.update_layout(
+                showlegend=False, 
+                title_x=0.25)
 
             st.plotly_chart(fig_pe)
 
@@ -165,23 +200,59 @@ def expert_mode():
 
             df_comp = dataframe_electricity_mix(countries_to_compare)
 
+
+            impact_metrices = [
+                "Abiotic Depletion Potential for Elements (Kilograms of antimony equivalent)",
+                "Primary energy demand (Megajoules)",
+                "Global Warming Potential (Kilograms of CO2 equivalent)",
+                "Water Withdrawal Factor (Liters/Kilowatt-hour)",
+                "Water Consumption Factor (Liters/Kilowatt-hour)"
+            ]
+
+            impact_labels = {
+                impact_metrices[0] : "adpe (kg eq. Sb)",
+                impact_metrices[1]  : "pe (MJ)",
+                impact_metrices[2]  : "gwp (kg eq. CO2)",
+                impact_metrices[3] :  "water_wwf (L/kWh)",
+                impact_metrices[4] : "water_wcf (L/kWh)" 
+            }
+
             impact_type = st.selectbox(
                 label='Select an impact type to compare',
-                options=[x for x in df_comp.columns if x!='country'],
-                index=1)
+                options=[x for x in impact_labels],  # les noms affichés à l'utilisateur
+                index=1
+            )
+
+            impact_type_used = impact_labels[impact_type]
 
-            df_comp.sort_values(by = impact_type, inplace = True)
+            df_comp.sort_values(by = impact_type_used, inplace = True)
             
             fig_2 = px.bar(
                 df_comp,
                 x = df_comp.index,
-                y = impact_type,
-                text = impact_type,
-                color = impact_type
+                y = impact_type_used,
+                text = impact_type_used,
+                color = impact_type_used
             )
             
             st.plotly_chart(fig_2)
 
+            impact_explanations = {
+                impact_metrices[0]:
+                    "Measures the depletion of non-renewable resources (metals, minerals), expressed in kg of antimony equivalent (Sb). \
+                    The enviornmental impact equates to that amount of antimony being extracted.",
+                impact_metrices[1]:
+                    "Total amount of primary energy used (both renewable and non-renewable), expressed in megajoules (MJ).",
+                impact_metrices[2]:
+                    "Potential contribution to climate change over a 100-year time horizon, expressed in kilograms of CO₂ equivalent.",
+                impact_metrices[3]:
+                    "Total volume of water withdrawn throughout the process, expressed in liters (L).",
+                impact_metrices[4]:
+                    "Volume of water actually consumed (not returned to the source), expressed in liters (L)."
+            }
+
+            st.info(impact_explanations[impact_type])
+
         except:
 
             st.warning("Can't display chart with no values.")

src/impacts.py

@@ -7,9 +7,16 @@ from src.utils import (
     format_energy_eq_physical_activity,
     format_gwp_eq_airplane_paris_nyc,
     format_gwp_eq_streaming,
+    format_water_eq_bottled_waters,
+    format_water_eq_olympic_sized_swimming_pool,
     PhysicalActivity,
-    EnergyProduction
+    EnergyProduction,
+    AI_COMPANY_TO_DATA_CENTER_PROVIDER,
+    PROVIDER_PUE,
+    PROVIDER_WUE_ONSITE
 )
+from src.electricity_mix import COUNTRY_CODES, find_electricity_mix, dataframe_electricity_mix
+
 
 ############################################################################################################
 
@@ -20,40 +27,83 @@ def get_impacts(model, active_params, total_params, mix_ghg, mix_adpe, mix_pe):
 ############################################################################################################
 
 
-def display_impacts(impacts):
+def display_impacts(impacts, provider, location):
 
     st.divider()
 
     col_energy, col_ghg, col_adpe, col_pe, col_water = st.columns(5)
 
     with col_energy:
-        st.markdown('<h4 align="center">⚡️ Energy</h4>', unsafe_allow_html = True)
+        st.markdown("""
+        <div style="text-align: center;">
+        <div style="font-size: 30px;">⚡️</div>
+        <div style="font-size: 25px;">Energy</div>
+        </div>
+        """, unsafe_allow_html = True)
         st.latex(f'\Large {impacts.energy.magnitude:.3g} \ \large {impacts.energy.units}')
-        st.markdown(f'<p align="center"><i>Evaluates the electricity consumption<i></p>', unsafe_allow_html = True)
+        st.markdown("""
+        <div style="height: 10px;"></div>            
+        <div style="text-align: center;"><i>Evaluates the electricity consumption<i>
+        </div>
+        """, unsafe_allow_html = True)
 
     with col_ghg:
-        st.markdown('<h4 align="center">🌍️ GHG Emissions</h4>', unsafe_allow_html = True)
+        st.markdown("""
+        <div style="text-align: center;">
+        <div style="font-size: 30px;">🌍️</div>
+        <div style="font-size: 18px;">GHG Emissions</div>
+        </div>
+        """, unsafe_allow_html = True)
         st.latex(f'\Large {impacts.gwp.magnitude:.3g} \ \large {impacts.gwp.units}')
-        st.markdown(f'<p align="center"><i>Evaluates the effect on global warming<i></p>', unsafe_allow_html = True)
+        st.markdown("""
+        <div style="text-align: center;"><i>Evaluates the effect on climate change<i>
+        </div>
+        """, unsafe_allow_html = True)
 
     with col_adpe:
-        st.markdown('<h4 align="center">🪨 Abiotic Resources</h4>', unsafe_allow_html = True)
+        st.markdown("""
+        <div style="text-align: center;">
+        <div style="font-size: 30px;">🪨</div>
+        <div style="font-size: 16px;">Abiotic Resources</div>
+        </div>
+        """, unsafe_allow_html = True)
         st.latex(f'\Large {impacts.adpe.magnitude:.3g} \ \large {impacts.adpe.units}')
-        st.markdown(f'<p align="center"><i>Evaluates the use of metals and minerals<i></p>', unsafe_allow_html = True)
+        st.markdown("""
+        <div style="text-align: center;"><i>Evaluates the use of metals and minerals<i>
+        </div>
+        """, unsafe_allow_html = True)
 
     with col_pe:
-        st.markdown('<h4 align="center">⛽️ Primary Energy</h4>', unsafe_allow_html = True)
+        st.markdown("""
+        <div style="text-align: center;">
+        <div style="font-size: 30px;">⛽️</div>
+        <div style="font-size: 18px;">Primary Energy</div>
+        </div>
+        """, unsafe_allow_html = True)
         st.latex(f'\Large {impacts.pe.magnitude:.3g} \ \large {impacts.pe.units}')
-        st.markdown(f'<p align="center"><i>Evaluates the use of energy resources<i></p>', unsafe_allow_html = True)
+        st.markdown("""
+        <div style="height: 10px;"></div>
+        <div style="text-align: center;"><i>Evaluates the use of energy resources<i>
+        </div>
+        """, unsafe_allow_html = True)
+
+    with col_water: 
+        st.markdown("""
+        <div style="text-align: center;">
+        <div style="font-size: 30px;">🚰</div>
+        <div style="font-size: 25px;">Water</div>
+        </div>
+        """, unsafe_allow_html = True)
+        st.latex(f'\Large {impacts.water.magnitude:.3g} \ \large {impacts.water.units}')
+        st.markdown("""
+        <div style="text-align: center;"><i>Evaluates the use of water<i>
+        </div>
+        """, unsafe_allow_html = True)
 
-    with col_water:
-        st.markdown('<h4 align="center">🚰 Water</h4>', unsafe_allow_html = True)
-        st.latex('\Large Upcoming...')
-        st.markdown(f'<p align="center"><i>Evaluates the use of water<i></p>', unsafe_allow_html = True)
 
 ############################################################################################################
 
-def display_equivalent(impacts):
+def display_equivalent(impacts, provider, location):
 
     st.divider()
 
@@ -61,39 +111,69 @@ def display_equivalent(impacts):
     
     streaming_eq = format_gwp_eq_streaming(impacts.gwp)
 
-    col1, col2, col3 = st.columns(3)
+    col1, col2, col3, col4 = st.columns(4)
 
     with col1:
         physical_activity, distance = format_energy_eq_physical_activity(impacts.energy)
         if physical_activity == PhysicalActivity.WALKING:
-            physical_activity = "🚶 " + physical_activity.capitalize()
+            physical_activity_emoji = "🚶 " 
+            physical_activity = physical_activity.capitalize()
         if physical_activity == PhysicalActivity.RUNNING:
-            physical_activity = "🏃 " + physical_activity.capitalize()
-
-        st.markdown(f'<h4 align="center">{physical_activity}</h4>', unsafe_allow_html = True)
+            physical_activity_emoji = "🏃 " 
+            physical_activity = physical_activity.capitalize()
+
+        st.markdown(f"""
+        <div style="text-align: center;">
+        <div style="font-size: 30px;">{physical_activity_emoji}</div>
+        <div style="font-size: 25px;">{physical_activity}</div>
+        </div>
+        """, unsafe_allow_html = True)
         st.latex(f'\Large {distance.magnitude:.3g} \ \large {distance.units}')
         st.markdown(f'<p align="center"><i>Based on energy consumption<i></p>', unsafe_allow_html = True)
 
     with col2:
         ev_eq = format_energy_eq_electric_vehicle(impacts.energy)
-        st.markdown(f'<h4 align="center">🔋 Electric Vehicle</h4>', unsafe_allow_html = True)
+        st.markdown(f"""
+        <div style="text-align: center;">
+        <div style="font-size: 30px;">🔋</div>
+        <div style="font-size: 22px;">Electric Vehicle</div>
+        </div>
+        """, unsafe_allow_html = True)
         st.latex(f'\Large {ev_eq.magnitude:.3g} \ \large {ev_eq.units}')
         st.markdown(f'<p align="center"><i>Based on energy consumption<i></p>', unsafe_allow_html = True)
 
     with col3:
         streaming_eq = format_gwp_eq_streaming(impacts.gwp)
-        st.markdown(f'<h4 align="center">⏯️ Streaming</h4>', unsafe_allow_html = True)
+        st.markdown(f"""
+        <div style="text-align: center;">
+        <div style="font-size: 30px;">⏯️</div>
+        <div style="font-size: 25px;">Streaming</div>
+        </div>
+        """, unsafe_allow_html = True)
         st.latex(f'\Large {streaming_eq.magnitude:.3g} \ \large {streaming_eq.units}')
         st.markdown(f'<p align="center"><i>Based on GHG emissions<i></p>', unsafe_allow_html = True)
     
+    with col4:
+        #water = water_impact(impacts, provider, location)
+        water_eq = format_water_eq_bottled_waters(impacts.water)
+        st.markdown(f"""
+        <div style="text-align: center;">
+        <div style="font-size: 30px;">🚰</div>
+        <div style="font-size: 25px;">Bottled Waters</div>
+        </div>
+        """, unsafe_allow_html = True)
+        st.latex(f'\Large {water_eq.magnitude:.3g} \ \large {"bottles"}')
+        st.markdown(f'<p align="center"><i>Based on water consumption, measured in 0.75 L bottles.<i></p>', unsafe_allow_html = True)
+    
+
     st.divider()
     
     st.markdown('<h3 align="center">What if 1% of the planet does this request everyday for 1 year ?</h3>', unsafe_allow_html = True)
     st.markdown('<p align="center">If this use case is largely deployed around the world, the equivalent impacts would be the impacts of this request x 1% of 8 billion people x 365 days in a year.</p>', unsafe_allow_html = True)
 
-    col4, col5, col6 = st.columns(3)
+    col5, col6, col7, col8 = st.columns(4)
 
-    with col4:
+    with col5:
 
         electricity_production, count = format_energy_eq_electricity_production(impacts.energy)
         if electricity_production == EnergyProduction.NUCLEAR:
@@ -102,15 +182,21 @@ def display_equivalent(impacts):
         if electricity_production == EnergyProduction.WIND:
             emoji = "💨️ "
             name = "Wind turbines"
-        st.markdown(f'<h4 align="center">{emoji} {count.magnitude:.0f} {name} (yearly)</h4>', unsafe_allow_html = True)
+        st.markdown(f'<h4 align="center">{emoji} {count.magnitude:.0f} {name} <span style="font-size: 12px">\n (yearly ⚡️ production)</span></h2></h4>', unsafe_allow_html = True)
         st.markdown(f'<p align="center"><i>Based on energy consumption<i></p>', unsafe_allow_html = True)
         
-    with col5:
+    with col6:
         ireland_count = format_energy_eq_electricity_consumption_ireland(impacts.energy)
-        st.markdown(f'<h4 align="center">🇮🇪 {ireland_count.magnitude:.3f} x Ireland <span style="font-size: 12px">(yearly ⚡️ cons.)</span></h2></h4>', unsafe_allow_html = True)
+        st.markdown(f'<h4 align="center">🇮🇪 {ireland_count.magnitude:.3f} x Irelands <span style="font-size: 12px">\n (yearly ⚡️ consumption)</span></h2></h4>', unsafe_allow_html = True)
         st.markdown(f'<p align="center"><i>Based on energy consumption<i></p>', unsafe_allow_html = True)
 
-    with col6:
+    with col7:
         paris_nyc_airplane = format_gwp_eq_airplane_paris_nyc(impacts.gwp)
         st.markdown(f'<h4 align="center">✈️ {round(paris_nyc_airplane.magnitude):,} Paris ↔ NYC</h4>', unsafe_allow_html = True)
-        st.markdown(f'<p align="center"><i>Based on GHG emissions<i></p>', unsafe_allow_html = True)
+        st.markdown(f'<p align="center"><i>Based on GHG emissions<i></p>', unsafe_allow_html = True)
+
+    with col8:
+        olympic_swimming_pool = format_water_eq_olympic_sized_swimming_pool(impacts.water)
+        st.markdown(f'<div style="text-align: center; font-size: 20px; font-weight: bold;">🏊🏼 {round(olympic_swimming_pool.magnitude):,} Olympic-sized swimming pools</h4>', unsafe_allow_html = True)
+        st.markdown(f'<p align="center"><i>Based on water consumption<i></p>', unsafe_allow_html = True)
+

src/models.py

@@ -15,9 +15,13 @@ def clean_models_data(df, with_filter = True):
         'cohere': 'Cohere',
         'microsoft': 'Microsoft',
         'mistral-community': 'Mistral Community',
-        'databricks': 'Databricks'
+        'databricks': 'Databricks',
+        "azureopenai" : "Microsoft", #présent dans ecologits mais pas ici. traité comme Openai
+        "huggingfacehub" : "AWS" #présent dans ecologits mais pas ici. traité comme Openai
     }
 
+
+
     models_to_keep = MAIN_MODELS
     
     df.drop('type', axis=1, inplace=True)

src/utils.py

@@ -2,7 +2,7 @@ from dataclasses import dataclass
 from enum import Enum
 
 from ecologits.model_repository import models
-from ecologits.impacts.modeling import Impacts, Energy, GWP, ADPe, PE
+from ecologits.impacts.modeling import Impacts, Energy, GWP, ADPe, PE, Water
 #from ecologits.tracers.utils import llm_impacts
 from pint import UnitRegistry, Quantity
 
@@ -27,6 +27,9 @@ u.define('km = kilometer')
 u.define('s = second')
 u.define('min = minute')
 u.define('h = hour')
+u.define('L = liter')
+u.define('mL = milliliter')
+#u.define('bottled waters = bottled waters')
 q = u.Quantity
 
 @dataclass
@@ -35,6 +38,7 @@ class QImpacts:
     gwp: Quantity
     adpe: Quantity
     pe: Quantity
+    water: Quantity
 
 
 class PhysicalActivity(str, Enum):
@@ -69,6 +73,8 @@ EV_ENERGY_EQ = q("0.17 kWh / km")
 # From https://impactco2.fr/outils/comparateur?value=1&comparisons=streamingvideo
 STREAMING_GWP_EQ = q("15.6 h / kgCO2eq")
 
+BOTTLED_WATERS_EQ = q("0.75 L")
+
 # From https://ourworldindata.org/population-growth
 ONE_PERCENT_WORLD_POPULATION = 80_000_000
 
@@ -90,6 +96,57 @@ IRELAND_POPULATION_MILLION = 5
 # From https://impactco2.fr/outils/comparateur?value=1&comparisons=&equivalent=avion-pny
 AIRPLANE_PARIS_NYC_GWP_EQ = q("177000 kgCO2eq")
 
+# From https://www.patagoniaalliance.org/wp-content/uploads/2014/08/How-much-water-does-an-Olympic-sized-swimming-pool-hold.pdf
+OLYMPIC_SWIMMING_POOL = q("2500000 L") #2.5 million 
+
+# From https://docs.google.com/spreadsheets/d/1uj8yA601uBtJ7GSf7k96Lv1NoQBfsCnVmTCII2HgZvo/edit?gid=0#gid=0
+# Google : https://www.gstatic.com/gumdrop/sustainability/google-2025-environmental-report.pdf
+# Meta: https://sustainability.atmeta.com/wp-content/uploads/2024/08/Meta-2024-Sustainability-Report.pdf
+# Microsoft: https://azure.microsoft.com/en-us/blog/how-microsoft-measures-datacenter-water-and-energy-use-to-improve-azure-cloud-sustainability/
+# OVHCloud: https://corporate.ovhcloud.com/en/sustainability/environment/
+# Scaleway: https://www-uploads.scaleway.com/Impact_Report2024_A4_EN_e63efcae20.pdf
+# AWS: https://sustainability.aboutamazon.com/2023-report
+# Equinix: https://www.equinix.com/resources/infopapers/2023-corporate-sustainability-report
+PROVIDER_WUE_ONSITE = { #Water use efficiency on-site, as opposed to off-site generated energy 
+    "Google" : 0.916,
+    "Meta": 0.18,    # L/kWh, 2023
+    "Microsoft": 0.49, #2022
+    "OVHCloud": 0.37, #2024
+    "Scaleway": 0.216, #2023
+    "AWS" : 0.18, #2023
+    "Equinix" : 1.07 #2023
+}
+
+
+# Google https://www.gstatic.com/gumdrop/sustainability/google-2025-environmental-report.pdf
+# Meta https://sustainability.atmeta.com/data-centers/#:~:text=Meta's%20operational%20data%20centers%2C%20on,Effectiveness%20(WUE)%20of%200.20.
+# Microsoft https://azure.microsoft.com/en-us/blog/how-microsoft-measures-datacenter-water-and-energy-use-to-improve-azure-cloud-sustainability/
+# OVHCloud https://corporate.ovhcloud.com/en/sustainability/environment/
+# Scaleway https://www-uploads.scaleway.com/Impact_Report2024_A4_EN_e63efcae20.pdf
+# AWS https://sustainability.aboutamazon.com/products-services/aws-cloud
+# Equinix https://www.equinix.com/content/dam/eqxcorp/en_us/documents/resources/infopapers/ip_2023_sustainability_en.pdf
+PROVIDER_PUE = { #Power use efficiency 
+    "Google" : 1.09,
+    "Meta" : 1.09,	
+    "Microsoft" : 1.18,	
+    "OVHCloud" : 1.26,	
+    "Scaleway" : 1.37,	
+    "AWS" : 1.15,	
+    "Equinix" : 1.42	
+}
+
+AI_COMPANY_TO_DATA_CENTER_PROVIDER = { #A list that draws the connection between AI companies and their data center providers 
+    "anthropic"	: "Google",
+    "mistralai"	: "OVHCloud",
+    "cohere"	: "AWS", 
+    "databricks" : "Microsoft", 
+    "meta"	: "Meta",
+    "google"	: "Google",
+    "microsoft"	: "Microsoft", 
+    "openai" : "Microsoft"
+} 
+
+
 #####################################################################################
 ### IMPACTS FORMATING
 #####################################################################################
@@ -115,6 +172,12 @@ def format_pe(pe: PE) -> Quantity:
         val = val.to("kJ")
     return val
 
+def format_water(water: Water) -> Quantity:
+    val = q(water.value, water.unit)
+    if val < q("1 L"):
+        val = val.to("mL")
+    return val
+
 def format_impacts(impacts: Impacts) -> QImpacts:
 
     try:
@@ -122,18 +185,21 @@ def format_impacts(impacts: Impacts) -> QImpacts:
         impacts.gwp.value = (impacts.gwp.value.max + impacts.gwp.value.min)/2
         impacts.adpe.value = (impacts.adpe.value.max + impacts.adpe.value.min)/2
         impacts.pe.value = (impacts.pe.value.max + impacts.pe.value.min)/2
+        impacts.water.value = (impacts.water.value.max + impacts.water.value.min)/2
         return QImpacts(
             energy=format_energy(impacts.energy),
             gwp=format_gwp(impacts.gwp),
             adpe=format_adpe(impacts.adpe),
-            pe=format_pe(impacts.pe)
+            pe=format_pe(impacts.pe),
+            water=format_water(impacts.water)
         ), impacts.usage, impacts.embodied
     except: #when no range
         return QImpacts(
             energy=format_energy(impacts.energy),
             gwp=format_gwp(impacts.gwp),
             adpe=format_adpe(impacts.adpe),
-            pe=format_pe(impacts.pe)
+            pe=format_pe(impacts.pe),
+            water=format_water(impacts.water)
         ), impacts.usage, impacts.embodied
 
 def split_impacts_u_e(impacts: Impacts) -> QImpacts:
@@ -153,7 +219,8 @@ def format_impacts_expert(impacts: Impacts, display_range: bool) -> QImpacts:
             energy=format_energy(impacts.energy),
             gwp=format_gwp(impacts.gwp),
             adpe=format_adpe(impacts.adpe),
-            pe=format_pe(impacts.pe)
+            pe=format_pe(impacts.pe),
+            water=format_water(impacts.water)
         ), impacts.usage, impacts.embodied
     
     else:
@@ -165,7 +232,8 @@ def format_impacts_expert(impacts: Impacts, display_range: bool) -> QImpacts:
             energy=format_energy(energy),
             gwp=format_gwp(gwp),
             adpe=format_adpe(adpe),
-            pe=format_pe(pe)
+            pe=format_pe(pe),
+            water=format_water(impacts.water)
         ), impacts.usage, impacts.embodied
 
 #####################################################################################
@@ -199,6 +267,11 @@ def format_gwp_eq_streaming(gwp: Quantity) -> Quantity:
         streaming_eq = streaming_eq.to("s")
     return streaming_eq
 
+def format_water_eq_bottled_waters(water: Quantity) -> Quantity:
+    water = water.to("L")
+    bottled_water_eq = water / BOTTLED_WATERS_EQ
+    return bottled_water_eq
+
 def format_energy_eq_electricity_production(energy: Quantity) -> tuple[EnergyProduction, Quantity]:
     electricity_eq = energy * ONE_PERCENT_WORLD_POPULATION * DAYS_IN_YEAR
     electricity_eq = electricity_eq.to("TWh")
@@ -216,4 +289,15 @@ def format_energy_eq_electricity_consumption_ireland(energy: Quantity) -> Quanti
 def format_gwp_eq_airplane_paris_nyc(gwp: Quantity) -> Quantity:
     gwp_eq = gwp * ONE_PERCENT_WORLD_POPULATION * DAYS_IN_YEAR
     gwp_eq = gwp_eq.to("kgCO2eq")
-    return gwp_eq / AIRPLANE_PARIS_NYC_GWP_EQ####################################################################################### MODELS PARAMETER####################################################################################
+    return gwp_eq / AIRPLANE_PARIS_NYC_GWP_EQ
+
+def format_water_eq_olympic_sized_swimming_pool(water: Quantity) -> Quantity:
+    water_eq = water * ONE_PERCENT_WORLD_POPULATION * DAYS_IN_YEAR
+    water_eq = water_eq.to("L")
+    return water_eq / OLYMPIC_SWIMMING_POOL
+
+# def format_water_eq_olympic_sized_swimming_pool(water):
+#     water_eq = water * ONE_PERCENT_WORLD_POPULATION * DAYS_IN_YEAR
+#     return water_eq / OLYMPIC_SWIMMING_POOL
+
+####################################################################################### MODELS PARAMETER####################################################################################